Power Management for Fuel Cell and Battery

Hybrid Unmanned Aerial Vehicle Applications

by

Thomas Strele

A Thesis Presented in Partial Fulfillment

of the Requirements for the Degree
Master of Science

Approved November 2016 by the

Graduate Supervisory Committee:

Changho Nam, Chair

Scott Pollat
Arunachala Kannan

ARIZONA STATE UNIVERSITY

December 2016
 ABSTRACT

As electric powered unmanned aerial vehicles enter a new age of commercial viability,

market opportunities in the small UAV sector are expanding. Extending UAV flight time through a

combination of fuel cell and battery technologies enhance the scope of potential applications. A

brief survey of UAV history provides context and examples of modern day UAVs powered by fuel

cells are given. Conventional hybrid power system management employs DC-to-DC converters to

control the power split between battery and fuel cell. In this study, a transistor replaces the DC-to-

DC converter which lowers weight and cost. Simulation models of a lithium ion battery and a

proton exchange membrane fuel cell are developed and integrated into a UAV power system

model. Flight simulations demonstrate the operation of the transistor-based power management

scheme and quantify the amount of hydrogen consumed by a 5.5 kg fixed wing UAV during a six

hour flight. Battery power assists the fuel cell during high throttle periods but may also augment

fuel cell power during cruise flight. Simulations demonstrate a 60 liter reduction in hydrogen

consumption when battery power assists the fuel cell during cruise flight. Over the full duration of

the flight, averaged efficiency of the power system exceeds 98%. For scenarios where inflight

battery recharge is desirable, a constant current battery charger is integrated into the UAV power

system. Simulation of inflight battery recharge is performed. Design of UAV hybrid power systems

must consider power system weight against potential flight time. Data from the flight simulations

are used to identify a simple formula that predicts flight time as a function of energy stored

onboard the modeled UAV. A small selection of commercially available batteries, fuel cells, and

compressed air storage tanks are listed to characterize the weight of possible systems. The

formula is then used in conjunction with the weight data to generate a graph of power system

weight versus potential flight times. Combinations of the listed batteries, fuel cells, and storage

tanks are plotted on the graph to evaluate various hybrid power system configurations.

i
 DEDICATION

This work is dedicated to my parents.

ii
 ACKNOWLEDGMENTS

I would like to thank Dr. Changho Nam for his guidance and support during this project.

iii
 TABLE OF CONTENTS

Page

LIST OF TABLES .................................................................................................................................. vii

LIST OF FIGURES ............................................................................................................................... viii

LIST OF SYMBOLS AND ACRONYMS ................................................................................................xi

CHAPTER

1 INTRODUCTION ................. ................................................................................................... 1

1.1 UAV Definition ................................................................................................ 1

1.2 UAV History .................................................................................................... 1

1.3 UAV Market………………………………………………………………………3

1.4 UAV Hybrid Power Systems……………………………………………………5

1.5 UAV Power Management……………………………………………………….7

1.6 Research Objectives…………….……………………………………………..10

2 LITERATURE SURVEY ............ ............................................................................................ 14

2.1 Fuel Cell Powered UAV Examples .............................................................. 14

2.2 Hybrid Powered UAV Examples .................................................................. 17

2.3 Lithium Polymer Battery Description…………………………...…………….21

2.4 Lithium Polymer Battery Ratings…………………...………...………………24

2.5 Lithium Ion Battery Model……………………………..………………………25

2.6 Proton Exchange Membrane Fuel Cell Description………...………………26

2.7 Proton Exchange Membrane Fuel Cell Polarization Curve……..…………33

2.8 The Electric Double Layer………………………………………..…………...39

2.9 Hydrogen Storage………………….……………..………………...…………40

2.10 Hybrid Power Systems Management…………….……..………………...…43

2.11 ASU APM Circuit…………………………………………..…………………..48

3 SIMULATION COMPONENTS ............................................................................................. 51

3.1 Simscape and Simulink ................................................................................ 51

3.2 Lithium Ion Battery Component Model ........................................................ 52

iv
CHAPTER Page

3.3 Battery Charger Model…………………………….…………………………...57

3.4 PEMFC Component Model……………………………………………………59

3.5 Balance of Plant Model………………………………………………………...71

3.6 Motor Propeller Model………………………………………………………….72

4 SYSTEM INTEGRATION ........... .......................................................................................... 78

4.1 The Current Assist and Observer Circuit ..................................................... 78

4.2 Hybrid Power Management Controller ........................................................ 80

4.3 Hybrid Power System Model…………………………………………………..84

4.4 System Level Simulation……………………………………………………....85

4.5 Inflight Battery Recharge………………………………………………………95

4.6 Flight Time and Hybrid Power System Weight………………………………98

5 DISCUSSION AND CONCLUSION .................................................................................... 108

5.1 Research Results ....................................................................................... 108

5.2 Future Research ......................................................................................... 112

5.3 Conclusion…………………….……………………………………………….114

REFERENCES....... .......................................................................................................................... 116

APPENDIX

A SIMSCAPE CODE FOR LITHIUM ION BATTERY COMPONENT ................................. 122

B LIPO BATTERY DISCHARGE CURVES AUTOMATION SCRIPT……..………………….125

C BATTERY DISCHARGE COMPARISON MODEL…………………………………………...127

D SIMSCAPE CODE FOR NERNST POTENTIAL SUBCOMPONENT…………..….………129

E MATLAB CALCULATION OF LEE MODEL ERROR OF NERNST POTENTIAL…………132

F SIMSCAPE CODE FOR ACTIVATION RESISTOR SUBCOMPONENT………………...135

G MATLAB CODE TO CALCULATE CELL RESISTANCE……………….….………………..138

H SIMSCAPE CODE FOR OHMIC RESISTOR SUBCOMPONENT…………………………140

I SIMSCAPE CODE FOR CONCENTRATION RESISTOR SUBCOMPONENT…………...143

J SIMSCAPE CODE FOR DOUBLE LAYER CAPACITOR SUBCOMPONENT…………….145

v
APPENDIX Page

K SIMSCAPE CODE FOR FLOW RATE SUBCOMPONENT……………………..…………147

L SIMSCAPE CODE FOR PEMFC COMPONENT (TOP LEVEL)…………………………..149

M MATLAB CODE FOR DYNAMIC RESPONSE TEST………………………………………152

N SIMSCAPE CODE FOR MPSS COMPONENTS………………………………………….154

O OBSERVER_1 DIAGRAM……………………………………………………………………157

P OBSERVER_2 DIAGRAM……………………………………………………………………..159

Q HYBRID POWER SYSTEM MODEL CONFIGURATION PARAMETERS……………….161

R FLIGHT PROFILE AUTOMATION SCRIPT………………………………………………....163

S MATLAB SCRIPT TO GENERATE FIGURE 45…………………………………………….166

T MATLAB SCRIPT TO GENERATE FIGURE 46…………………………………………….169

U MATLAB SCRIPT TO GENERATE FIGURE 47…………………………………………….172

V MATLAB SCRIPT TO GENERATE FIGURE 48…………………………………………….174

W CALCULATE CHANGE IN TOTAL ENERGIES BETWEEN SIMULATION RUNS………177

X THRUST TEST BENCH DATA……………………………………………………………......182

Y SIMULATION OF SIX HOUR FLIGHT FOR DIFFERENT BATTERY CAPACITIES…….187

Z MATLAB SCRIPT TO CALCULATE TAKEOFF AND LANDING ENERGY………………190

AA MATLAB SCRIPT TO PLOT FLIGHT TIME VERSUS POWER SYSTEM WEIGHT ........ 192

vi
 LIST OF TABLES

Table Page

1. Energy and Power Densities .............................................................................................. 6

2. Battery Type Comparison ................................................................................................. 22

3. LiPo 4S 8Ah Battery Model Parameter Values ................................................................ 55

4. Gibbs Free Energy for Hydrogen Reaction ...................................................................... 62

5. AXI 4120/14 GDCM with 13 x 8 APC Propeller Specification ......................................... 74

6. Simulation Total Energies for Hybrid = 0 and 1.5 ............................................................ 91

7. Hydrogen Consumed and Battery Final SOC for Different Battery Capacities .............. 93

8. LiPo 4S Battery Weights ................................................................................................... 93

9. Comparison of Simulation Data and Flight Time Approximation Formula .................... 101

10. MaxAmps LiPo 4S Battery Family Specifications ........................................................ 102

11. Horizon Fuel Cell Weights ............................................................................................ 102

12. Ninja Pro v2 Tank Weights and Storage Capacities .................................................... 103

vii
 LIST OF FIGURES

Figure Page

1. Hornet UAV ....................................................................................................................... 15

2. Spider Lion UAV ................................................................................................................ 15

3. Ion Tiger UAV .................................................................................................................... 16

4. Grey-faced Buzzard UAV .................................................................................................. 18

5. FAUCON H2 (left), Stalker XE (right) ............................................................................... 19

6. Piper Cub (left), ASU Composite UAV (right) ................................................................... 20

7. ASU Composite UAV CAD Assembly (left), ASU Composite UAV Disassembly (right) 21

8. ASU Composite UAV Test Flight, December 2012 ......................................................... 21

9. Battery Type Volumetric Energy Density versus Specific Energy Density ..................... 23

10. Membrane Electrode Assembly ...................................................................................... 29

11. Complete Fuel Cell System ............................................................................................ 32

12. PEMFC Polarization Curve ............................................................................................. 38

13. Ninja Pro v2 90/4500 Compressed Air Tank .................................................................. 43

14. APM Circuit Concepts using DC-to-DC Converters ...................................................... 44

15. ASU APM Circuit ............................................................................................................. 50

16. LiPo 4S 8 Ah Battery Discharge Circuit .......................................................................... 54

17. LiPo 4S 8 Ah Battery Discharge Curves ........................................................................ 56

18. Battery Discharge Comparison at 0.5C Discharge Rate ............................................... 56

19. Battery Recharge Circuit ................................................................................................. 57

20. Battery Recharge Simulation .......................................................................................... 58

21. Battery SOC during Recharge ........................................................................................ 58

22. Battery Voltage during Recharge ................................................................................... 59

23. Battery Current during Recharge .................................................................................... 59

24. Electrical Circuit Equivalent Model of PEMFC ............................................................... 60

25. Percent Error of Lee Model Approximation of Nernst Potential ..................................... 63

26. Percent Error of Lee Model for Different Hydrogen Pressures ...................................... 64
viii
Figure Page

27. MEA Resistance versus Current Density for Different Relative Humidity Values ........ 66

28. Cell Resistance versus Current Density and Temperature ........................................... 68

29. PEMFC Test Simulation .................................................................................................. 69

30. PEMFC Model Polarization and Power Curves ............................................................. 69

31. PEMFC Model Parameters ............................................................................................. 70

32. PEMFC Dynamic Response for Range of Double Layer Capacitance ......................... 70

33. BOP Circuit ...................................................................................................................... 71

34. BOP_Load Control Signal ............................................................................................... 71

35. MPSS Block Diagram ..................................................................................................... 74

36. Propeller Speed and Motor Power versus Throttle ........................................................ 75

37. Thrust Test Bench ........................................................................................................... 76

38. Thrust Test Bench Data .................................................................................................. 76

39. CAOC Concept ................................................................................................................ 79

40. HPMC Circuit ................................................................................................................... 81

41. Battery and Hydrogen Coarse Fuel Gages .................................................................... 82

42. Setpoint Flow Chart ......................................................................................................... 83

43. Hybrid Power System Model .......................................................................................... 85

44. Simulation Input Signals ................................................................................................. 86

45. Fuel Cell and Battery Voltage and Current versus Time (hybrid = 0 and 1.5) .............. 88

46. Electrical Power versus Time (hybrid = 0 and 1.5) ........................................................ 88

47. Battery SOC and Hydrogen Consumption versus Time ................................................ 89

48. Q1 Power Dissipation versus Time ................................................................................ 90

49. Battery SOC and Hydrogen Consumed for Different Battery Capacities ...................... 93

50. Hybrid Power System Performance for 11 Ah Battery .................................................. 94

51. Hybrid Power System Performance for 11 Ah Battery Expanded View ........................ 95

52. CTRL_Block Flow Chart ................................................................................................. 96

53. Input Signals for Inflight Battery Recharge Simulation .................................................. 97

ix
Figure Page

54. Fuel Cell and Battery Voltage and Current during Inflight Recharge ............................ 98

55. Battery SOC and Current during Inflight Battery Recharge ........................................... 98

56. Power Consumed During Six Hour Flight (Wh/min) ...................................................... 99

57. Flight Time versus Weight for Power System Configurations ..................................... 105

58. Hydrogen Phase Diagram ............................................................................................ 113

59. Battery Discharge Comparison Model……………………………………………………128

60. Observer_1 Diagram ………………………………………………………………………158

61. Observer_2 Diagram ………………………………………………………………………160

62. Hybrid Power System Model Configuration Parameters ………………………………162

63. Oscilloscope Capture of ESC Voltage Signals Sent to Motor ………………………...183

64. Oscilloscope Capture of ESC Voltage and Current Motor Signals (zoomed in) …….183

65. Motor Measurements with DC Power Supply and No Propeller ................................184

66. Motor Measurements with DC Power Supply with Propeller ………………………….185

67. Motor Measurements with Motorcycle Battery with Propeller ………………………...186

x
 LIST OF SYMBOLS AND ACRONYMS

Symbol Page

1. Active Power Management: APM ....................................................................................... 7

2. Alternative Power Source: APS .......................................................................................... 5

3. Arizona State University: ASU .......................................................................................... 17

4. Balance of Plant: BOP ........................................................................................................ 7

5. Battery Capacity (Ah): CBAT ............................................................................................. 100

6. Brushless DC Motor: BDCM ............................................................................................... 8

7. Current Assist and Observer Circuit: CAOC .................................................................... 78

8. Department of Defense: DOD ............................................................................................. 1

9. Department of Energy (United States): DOE ................................................................... 40

10. Electromagnetic Interference: EMI ................................................................................. 45

11. Electronic Speed Controller: ESC .................................................................................... 8

12. Energy Available from Hybrid Power System: EHPS ........................................................ 99

13. Energy Consumed During Takeoff and Landing: ETOL .................................................. 99

14. Energy Delivered from Battery: EBAT ............................................................................ 100

15. Energy Delivered from Fuel Cell: EFC ........................................................................... 100

16. Finite Element Analysis: FEA ....................................................................................... 112

17. Fuel Cell Technologies Office: FCTO .............................................................................. 15

18. Gas Diffusion Layer: GDL ............................................................................................... 28

19. Global Positioning System: GPS .................................................................................... 14

20. High-Altitude, Long-Endurance: HALE ............................................................................. 3

21. Higher Heating Value: HHV ............................................................................................ 94

22. Hybrid Power Management Controller: HPMC .............................................................. 80

23. Internal Combustion Engine: ICE ..................................................................................... 2

24. Intelligence, Surveillance, Reconnaissance: ISR ............................................................. 3

25. Joint Unmanned Combat Aircraft System: J-UCAS ........................................................ 2

26. Liquid Hydrogen: LH2 .................................................................................................... 112

xi
Symbol Page

27. Liquid Petroleum Gas: LPG ............................................................................................ 18

28. Liters of Hydrogen Available: LHA ................................................................................ 100

29. Lithium Ion Battery: Li-Ion ............................................................................................... 11

30. Lithium Polymer Battery: LiPo ........................................................................................ 19

31. Lower Heating Value: LHV ............................................................................................. 94

32. Membrane Electrode Assembly: MEA ........................................................................... 28

33. Metal Oxide Field Effect Transistor: MOSFET ............................................................... 48

34. Microprocessor Block: MB .............................................................................................. 81

35. Motor Propeller Subsystem: MPSS ................................................................................ 73

36. Naval Research Laboratory: NRL ................................................................................... 15

37. Nickel Cadmium Battery: Ni-Cd ...................................................................................... 21

38. Nickel Metal-Hydride Battery: Ni-MH .............................................................................. 21

39. Operational Amplifier: OpAmp ........................................................................................ 54

40. Passive Power Management: PPM .................................................................................. 7

41. Physical Network Approach: PNA .................................................................................. 51

42. Power Management System: PMS .................................................................................. 7

43. Proton Exchange Membrane Fuel Cell: PEMFC ............................................................. 6

44. Revolutions per Minute: RPM ......................................................................................... 75

45. Small Unmanned Aerial System: sUAS ........................................................................... 3

46. Solid Oxide Fuel Cell: SOFC .......................................................................................... 18

47. Standard Liters per Minute: SLPM ................................................................................. 42

48. Standard Temperature and Pressure: STP ................................................................... 42

49. State of Charge: SOC ....................................................................................................... 9

50. Tetrahydro Furan: THF ................................................................................................... 24

51. Thrust Test Bench: TTB .................................................................................................. 75

52. Unmanned Aerial Vehicle: UAV ........................................................................................ 1

53. Voltage versus Current Curve: VI-curve .......................................................................... 9

xii
 CHAPTER 1

INTRODUCTION

1.1 UAV Definition

Unmanned Aerial Vehicles (UAV) may be loosely described as an aircraft without a pilot

onboard. A UAV may be piloted remotely via radio control or fly autonomously using an onboard

autopilot system. In some cases, both forms of flight control are present. That is to say radio

control is present in addition to an onboard autopilot system. Refining the definition of a UAV, a

propulsion system must also be present onboard the aircraft (“DOD dictionary”, 2007). A

propulsion system may consist of a motor with a propeller or a jet engine. This second

requirement excludes flying projectiles such as bullets and artillery shells. Typically, the term UAV

implies a powered unmanned aircraft, either fixed wing or rotary wing. According to the U.S.

Department of Defense (DOD), a UAV is defined as,

A powered, aerial vehicle that does not carry a human operator, uses aerodynamic forces
to provide vehicle lift, can fly autonomously or be piloted remotely, can be expendable or
recoverable, and can carry a lethal or nonlethal payload. Ballistic or semi-ballistic
vehicles, cruise missiles, and artillery projectiles are not considered unmanned aerial
vehicles. (“DOD dictionary”, 2007)

Note that all ballistic or semi-ballistic rockets and missiles are excluded from the UAV definition.

For the special case of cruise missiles (non-ballistic), where aerodynamic forces provide

significant lift, the DOD definition explicitly excludes them from the UAV definition. There are

many examples of both fixed wing and rotary wing UAVs, but this paper will focus only on fixed

wing UAVs.

1.2 UAV History

UAVs have been around since before the first manned powered airplane flight of the

Wright brothers in 1903 (“Wright Brothers”, n.d.). The first recorded UAV flight took place in 1896

when Samuel Langley launched an unmanned, steam powered aircraft from a house boat on the

Potomac River. The aircraft, named the “Aerodrome No. 5”, flew for about 90 seconds and

covered a distance of 1005 m (“Langley Aerodrome Number 6”, n.d.). The Aerodrome weighed

11.4 kg and had a wingspan of 4.2 m. The engine consisted of a single cylinder steam engine

producing 1 horsepower or 745.7 W. Further development of UAVs were closely tied to

1
navigational and radio control technologies. During the period of 1917 to 1918, both the U.S.

Navy and U.S. Army funded UAV programs to deliver ordinance to remote targets. The Navy

“Aerial Torpedo” and the Army “Kettering Bug” were tested with mixed results, but were never

used in combat (Keane, 2013). By 1924, the Navy Research Lab demonstrated the successful

flight of a radio controlled F-5L aircraft. The significance of this test flight was that the F-5L aircraft

was remotely controlled during all phases of flight (Keane, 2013). The Navy later sponsored

development of carrier based UAVs under the Joint Unmanned Combat Aircraft System program

(J-UCAS). With of over a decade of development, the J-UCAS program produced the TDR-1

assault drone (Gundlach, 2014). Considered the Navy’s first operational UAV, the TDR-1 flew

combat missions during World War II. With a 14.6 m wingspan and weighing 2676 kg, the TDR-1

could carry up to 907 kg of ordinance to a target 684 km away. With a cruising speed of 225

km/h, the flight endurance works out to be 3.04 hours. The TDR-1 was powered by twin Lycoming

O-435 internal combustion engines (ICE) each weighing 196.4 kg and producing 220 horsepower

or 164.1 kW (Interstate TDR, 2016). On board avionics systems included a television transmitter,

radar altimeter, and radio control receiver (Lee, 2013). A significant factor to the success of the

TDR-1 can be credited to the improved navigation and radio control electronics.

Today, there are a variety of fixed wing UAVs in both military and civilian service.

Introduced in 1996, the General Atomics RQ-1 Predator UAV provided aerial reconnaissance

capabilities to the U.S. military. In 2002, the addition of missile carrying capacity changed the

designation from RQ-1 to MQ-1. Military designation convention assigns the letter R for

reconnaissance, the letter M for multi-role, and the letter Q for unmanned. Further enhancements

to the Predator were released in 2005 changing the designation to MQ-1B which is still in service

today (“MQ-1B Predator”, 2015). With a wingspan of 16.8 m and a weight of 512 kg, the MQ-1B

has a range of 1427 km and a maximum flight endurance of 40 hours (“MQ-1 Predator/MQ-9

Reaper”, 2016). Powered by an ICE, the Rotax 914F four cylinder engine develops 115

horsepower or an equivalent 85.8 kW of electrical power. A second example of a modern day

UAV, a civilian UAV, is the Altair. As a joint development effort between NASA and General

Atomics, it is a modified version of a Predator B. Designed for high altitude scientific and
2
commercial research, it can reach a maximum altitude of 15.8 km. With a weight of 3175 kg and a

wingspan of 26.2 m, the Altair has a range of 12,441 km and a flight endurance of 32 hours

(Gibbs, 2015). Powered by the Honeywell TPE 331-10T turboprop engine, the power plant

develops 940 horsepower (700.9 kW) with a weight of only 174.6 kg (“TPE331-10 Turboprop

Engine”, 2006). The third example of a modern UAV is perhaps one of the most famous, the

Northrop Grumman RQ-4B Global Hawk. As the military designator indicates, it is an unmanned

reconnaissance aircraft. More specifically, it is a High-Altitude, Long-Endurance (HALE) aerial

reconnaissance system. Able to reach altitudes of 19.8 km, it is perhaps the highest flying UAV to

date. The Global Hawk weighs 6781 kg and has a wingspan of 39.9 m (“Northrop Grumman

Facts”, 2008). The Rolls Royce AE3007H turbofan engine produces 9.5 million newtons of thrust

with a power plant weight of 746 kg (“AE3007”, n.d.). The flight endurance is reported to be 35

hours with a range of 22780 km (“Northrop Grumman Facts”, 2008). The on board sensor system

combines electro-optical sensing, infra-red sensing, and synthetic aperture radar (“Northrop

Grumman RQ-4 Global Hawk”, 2016). Global Hawk is a paragon of how far UAV technology has

come since the early days of UAV flight.

1.3 UAV Market

The current day market for UAVs is expanding, but the sector of small Unmanned Aerial

Systems (sUAS) has the largest growth potential. The designation of UAS instead of UAV

includes not only the airplane platform, but also the control and analysis equipment needed to

perform the mission objectives. The classification of “small” UAS is generally accepted as UAV

weight under 25 kg. The UAV is not valued on the novelty of a remote controlled aircraft, but

rather on the value of the mission it can perform. In other words, the value of the UAS is in the

capabilities it provides to the customer. For example, military UAVs perform intelligence,

surveillance, and reconnaissance (ISR) missions. The airplane platform includes imaging and

navigation equipment to record imaging and location data. The full UAS system includes the UAV

plus radio control equipment and analysis computers to interpret the data. The value of the UAS

is in the ability to gather imaging data with associated locations of a remote area. Similarly, the

value of a civilian UAS may be to provide thermal imaging data of crops over a group of large
3
fields for better farming. Within the aerospace industry, the small UAS sector is small, valued at

1.25 billion U.S. dollars in 2003 (NASA, 2006). However, the small UAS sector has also been

called the most dynamic of the aerospace industry due to its high growth potential. In 2015, ABI

research predicted the sUAS sector alone will exceed 8.4 billion dollars by 2018 (Manufacturing

Close-Up, 2015). Enabling this explosive growth, are advancements in electronics. Smaller,

lighter, and more capable systems can now be manufactured at a price low enough to interest

private sector customers. Previously, sUAS costs have been too high for many civilian

applications, but there is a trend of declining costs. Although sUAS price points are still above

conventional costs, sUAS solutions are capturing the attention of potential users. The expectation

is that sUAS solutions will soon be favorable for a wide variety of civilian applications.

As the cost of sUAS decrease, the variety of applications broadens and the user base

expands. During the early years of UAV development, demonstration of the technology was the

goal. The users were researchers and financiers. Although research and development costs were

high, they were acceptable to demonstrate proof of concept. After reliable service had been

proven, military and government sponsored programs were the only users; the only ones who

could afford the costs. For these end users, UAS applications were mainly military, namely ISR

and weaponry. As the technology matured, scientific programs were funded, such as the Altair

program. Continued development and innovation have now decreased the costs to the point

where commercial applications are possible. Scientific and civilian flight missions that are

considered dull, dirty, or dangerous are good candidates for sUAS applications. Areas of interest

such as earth science, land management, communications, and homeland security are being

considered (NASA, 2016). To continue the trend of increased UAS mission capability at a lower

cost, further improvements to the technology are being researched. Topics such as cost, flight

endurance, flight autonomy, sensor systems, and application data analysis are areas which will

have a significant impact on sUAS market growth.

4
1.4 UAV Hybrid Power Systems

Within the sUAS category, UAVs weighing 10 kg or less, often use electric motors as the

power plant instead of ICEs. There are many advantages to using an electric motor for small

scale UAVs. Electric motors have higher efficiency than their ICE counterparts. Electric motors

have efficiencies in the range of 80% to 90% while small ICE have efficiencies in the 10% to 20%

range. Since there are fewer moving parts, electric motors are more reliable and easier to

maintain. Electric motors have no emissions; zero carbon footprint. Also, electric motors have low

noise, vibration and thermal signature. The disadvantage of electric motors is electric power

storage. Due to limitations in battery technology, flight endurance is limited. With battery power

only, typical flight times are three hours or less. There is substantial interest in the UAV

community to extend the endurance of electric powered flight.

Two approaches have been taken to improve flight endurance for electric powered, small

scale UAVs. The first approach is to mount photovoltaic cells on the aircraft to generate electric

power during flight. The second approach is to use fuel cells to convert stored hydrogen into

electric power. In this paper, the focus will be on the fuel cell approach. For many cases, the

power delivered by the photovoltaic cells or the fuel cell is not enough for the aircraft to takeoff or

climb. To compensate, lithium polymer batteries are combined with the alternative power source

(APS) to create a hybrid power system. In other words, battery power is used to augment power

delivery from the photovoltaic cells or fuel cell during the takeoff or climb segments of flight. An

example of a small scale, electric powered UAV using a hybrid power system is the FAUCON H2

from EnergyOr Technologies. With a wingspan of 3 m and a weight of 9 kg, the FAUCON H2

completed a flight of 10 hours and 4 minutes. Using a hybrid fuel cell and battery power system,

the flight lasted four times longer than using batteries alone (“EnergyOr fuel cell powered UAV

reaches 10 h flight endurance”, 2011). A second example is a demonstrator aircraft constructed

at the University of Johannesburg in 2009. A small Piper Cub model airplane was powered by a

custom 100 W fuel cell. The Piper Cub had a wingspan of 2.31 m and a weight of 5.35 kg. The

main objective of the project was to demonstrate that a small 100 W fuel cell could sufficiently

power the aircraft during level flight. To have enough power during takeoff and climb, lithium
5
polymer batteries were used. The radio control transmitter had one channel dedicated to

controlling a relay inside the aircraft to switch between batteries or the fuel cell. During takeoff

and climb, the batteries powered the electric motor. When the level flight or cruising portion of the

flight began, the power source was switched from batteries to the fuel cell. Although the flight

lasted only 9 minutes, the project goal was successfully demonstrated. That is, a 100 W fuel cell

can deliver sufficient power to maintain a small scale UAV in level flight (Furrutter, 2009).

Using fuel cells in combination with batteries in UAVs has a synergistic or complimentary

quality. Hydrogen fuel cells have a relatively high energy density but low power density. On the

other hand, batteries have a relatively low energy density but have a high power density. The fuel

cell can enhance flight endurance while the battery delivers high power when needed. Fuel cells

extend the flight endurance due to the high energy density of hydrogen. Compared to gasoline,

compressed hydrogen gas has more than 2.5 times the specific energy. In other words, for the

same amount of fuel weight, compressed hydrogen stores more than 2.5 times the energy than

gasoline. Additionally, fuel cells have an operational efficiency of roughly 3 times that of ICEs.

Ideally, the superior energy density of compressed hydrogen and the higher efficiency of fuel cells

can result in a 7.5 times or more increase in flight endurance. It is the potential increase in flight

endurance that has attracted interest and research in hybrid fuel cell and battery power systems

for small scale UAV applications. Actual flight endurance improvement is lower than the ideal

value due to practical engineering issues such as weight, power management, fuel cell system

design, hydrogen storage, and other system level design issues.

Table 1

Energy and Power Densities (Zhang, 2014)

Energy Source Specific Energy (MJ/kg) Energy Density (MJ/L) Specific Power (W/kg)
Gasoline 45.7 34.6
Hydrogen (300 bar) 119.7 2.48 59.17
LiPo Battery 0.543 1.44 2320

Even though fuel cells have the ability to extend flight endurance, from a practical

engineering point of view, there are still advantages and disadvantages to their use. Starting with

the advantages, fuel cells operate at a higher efficiency than ICEs. Typical proton exchange

6
membrane fuel cells (PEMFC) operate from 40% to 60% efficiencies while small ICEs have

efficiencies from 10% to 20%. With fuel cells, there are no moving parts, no vibration, and no

combustion; just direct energy conversion. Operation is quiet and dependable. There are no

carbon emissions, just water and heat as byproducts. The disadvantages include higher cost,

sensitivity to fuel contaminants, low power density, and a limited operational lifetime. Also, the

fuel cell requires supporting equipment for operation called balance of plant (BOP). The BOP

equipment consists of a fuel cell controller, a thermal management system, a water management

system, and possibly an air intake compressor; all of which consume power. Part of the fuel cell

output power may be used to run the BOP equipment, but a battery is needed to power the BOP

equipment during fuel cell startup.

1.5 UAV Power Management

Power management of hybrid fuel cell and battery power systems in UAV applications

controls the power flow to and from both power sources. The goal of the power management

system (PMS) is to maximize flight endurance while keeping both fuel cell and battery within their

safe operational limits. Research in this area is ongoing and many different approaches have

been reported in literature. In general, the different power management approaches may be

classified into two broad categories, active and passive. Passive power management (PPM) are

cases where the fuel cell and battery are selected to be roughly the same voltage and are

connected directly in parallel. The combined power is then sent to the electric motor to power the

aircraft. The advantage of PPM is that there are no extra electronics to carry along in the UAV;

weight is kept to a minimum. The drawback of PPM, is that there is no control of power flow in the

system. There is no mechanism to ensure the fuel cell and battery operate within their limits. If

there is voltage mismatch between the fuel cell and battery, then energy may be wasted by

unmanaged power flow between the two sources. In these cases, the two power sources are said

to be in contention. The second category, active power management (APM), uses electronics to

control power flow in the system. The fuel cell and battery are kept within operational limits. The

downside to APM is that there are supervisory electronics to carry in the UAV, thus increasing the

weight and consuming power. An APM system may be as simple as using a relay to control which
7
source delivers power to the motor. The Piper Cub example cited earlier is a good example of a

simple APM approach. The beneficial feature of the Piper Cub APM approach was that the weight

and power consumption of the APM was minimal. Only a relay and one channel from the radio

control were used. The disadvantage was that the remote operator had to know when to switch

between power sources. An APM system may be very complicated such as using multiple DC-to-

DC converters under microprocessor control. In these cases, control of power flow can be closely

managed and optimal results attained. The cost for such control is extra weight, cost, and power

consumption.

The PPM electrical architecture centers on an electrical bus which serves to interconnect

all power sources and power loads. The bus is simply two wires; one positive and one negative.

The negative bus wire is considered ground for the power system; the electrical reference

potential to which all other power system voltages are referenced to. The positive bus wire is at

the power system voltage. For example, a 12 V power system means that the positive bus wire is

at plus 12 V relative to the negative bus wire. The point is that a voltage is always the electrical

potential difference between two points; in this case the bus positive wire relative to the bus

negative wire. The PEMFC positive terminal, the battery positive terminal, and the electronic

speed controller (ESC) positive input terminal are connected to the positive bus wire. Similarly,

the negative terminals of the three components are connected to the negative bus wire. In effect,

the two power sources and the power load are connected in parallel which constrains the voltage

across them to be equal. The power load is the electronic speed controller. The ESC is an

electronic device that drives the brushless DC motor (BDCM) to a constant speed. It receives its

input power from the bus and speed commands from a separate electronic input connected to the

radio control receiver. When set to zero, the ESC draws no power from the bus and sends no

power to the BDCM. In reality, the ESC still draws a small amount of power from the bus at a zero

setpoint; enough power to run its internal electronics (quiescent power). At this point, there is zero

electrical load on the power bus (neglecting the load quiescent power). Ideally, the PEMFC and

battery currents should be zero; unfortunately, they are not. A small mismatch in the PEMFC

voltage and the battery voltage results in a large current between them. The current is a result of
8
the parallel connection between the power sources which constrains the terminal voltages to be

equal. The end result is that the current between the power sources will increase until the power

source voltages are the same. Depending on the magnitude of the power source voltage

mismatch, the current flow between them can be large which results in a significant waste of

energy. The two power sources are in contention. When designing a PPM system, the voltage

mismatch between the PEMFC and battery pack is minimized in order to minimize wasted

energy. As the ESC setpoint value increases, electrical power is drawn from the bus to the ESC

which in turn sends electrical power to the BDCM. The load power drawn from the bus continues

to increase until the BDCM is rotating at the ESC setpoint. The power sources now deliver power

to the ESC and to each other. As the load on each power source increases, the operating point

on the voltage versus current curve (VI-curve) for each power source changes. A power system

balance is reached where the load receives its required power and the two sources split the

power load according to the equilibrium point reached on their respective VI-curves. At higher

load currents, the power sources have an opportunity to equalize their terminal voltages by

establishing a ratio of current delivered to the load. If the source VI-curves are somewhat close,

the power exchange between them can become small since their terminal voltages can equalize

by splitting the load current. In practice, the PPM scheme is used and the energy wasted at low

load currents is tolerated. The advantage of the PPM scheme is that no supervisory power

management electronics are used, thus minimizing complexity, cost, and weight. The

disadvantages of the PPM scheme is that energy is wasted by power exchange between the

sources and there is no control over the battery state of charge (SOC) and the fuel cell operating

point.

The APM electrical architecture is different from the PPM case in that it controls power

flow from the sources. The goal is to minimize wasted energy at the power sources and maximize

the energy delivered to the load. There are many ways to implement an APM system. The Piper

Cub example is a simple case, where a relay is used to connect either the battery or the fuel cell

to the bus, but never both at the same time. During takeoff and initial climb of the flight profile,

only the battery is connected to the bus, delivering high power to the load. When the cruise phase
9
of the flight profile is reached, the relay switches over to the fuel cell. The battery is disconnected

from the bus and the fuel cell is connected to the bus. The fuel cell then delivers a reduced power

level to the load, but no power flows from the fuel cell to the battery. In this manner, the relay

electrically separates the two power sources and there is never contention between the sources.

A common feature of power management strategies reported in literature is the use of

one or more DC-to-DC converters as a key component in APM electrical architecture. For a

power source connected to the electrical bus through a DC-to-DC converter, the amount of power

delivered to the bus can be controlled by the output voltage of the converter. For example,

consider a hybrid fuel cell and battery power system. The fuel cell is connected to the bus through

a DC-to-DC converter while the battery is connected directly to the bus. The APM controller can

adjust the DC-to-DC converter output voltage which effectively varies the fuel cell voltage

presented to the bus. If the controller keeps track of the battery SOC, it can then use this

information to maintain the battery SOC within safe limits. At the same time, adjusting the DC-to-

DC converter output voltage effectively limits the load current of the fuel cell. A disadvantage to

the use of DC-to-DC converters between a power source and the bus is the weight of the

converter. High power DC-to-DC converters are heavy, and significantly increase the weight of an

APM system. In UAV applications, excessive weight due to the APM system detracts from aircraft

performance.

1.6 Research Objectives

In conventional electric powered UAVs, a battery is the preferred electrical energy

storage device. Current research investigates the application of alternative energy sources such

as fuel cells or photovoltaic arrays to UAV power systems. In this study, photovoltaic power is not

pursued and the focus is on fuel cell and battery energy sources as applied to UAV power

systems. Batteries alone can supply enough electric power for UAV flight, but have a limited flight

time; typically under 3 hours. Fuel cells present an alternative to batteries and have the potential

for longer flight times. Unfortunately, to handle all electric power requirements for UAV flight, a

fuel cell must be sized so that its power rating can handle the maximum power requirement. For

UAV flight, power consumption can be three to five times higher during takeoff than during regular
10
cruising conditions, but the higher power is only required for a short period of time; roughly 2% to

4% of the total flight time. The result is that the fuel cell size is much larger and heavier than

necessary for 96% to 98% of the flight time. The combination of fuel cell and battery for UAV

power systems allows the fuel cell to be properly sized for the majority of flight time while the

battery can assist with power delivery during takeoff. A hybrid power system configuration

consisting of a fuel cell combined with a battery can significantly improve flight endurance and still

meet all power requirements for UAV flight. The first research goal of this study is to create

simulation models of the battery, fuel cell, and electrical load. The simulation environment

employed for this work is Matlab and the simulation components are written in the Simscape

language. Simscape runs in the Simulink simulator which is part of Matlab. Simscape is a new

addition to the Simulink toolbox and is used to create simulation models of physical systems.

Active power management of a hybrid power system controls the electrical load sharing

between the fuel cell and battery. The power required by the load must be met at all times during

the flight profile and the fuel cell and battery must be maintained within their safe operating limits.

In many active power management approaches, a DC-to-DC converter is used in the main power

path to the load. DC-to-DC converters with large power ratings are heavy and thus detrimental to

overall UAV performance. An active power management system that does not use a DC-to-DC

converter in the main power path will lower power system weight and thus improve overall UAV

performance. Typical of UAV hybrid power sources, the battery is of the Lithium Ion (Li-Ion) type,

or more specifically, Lithium Polymer (LiPo). For such batteries, the SOC must be kept within

lower and upper limits or damage to the battery may result. In cases where the SOC remains

outside the safe operating zone, the battery may spontaneously ignite and cause a hazardous fire

condition. A popular type of fuel cell used in UAV hybrid power systems is the PEMFC. The

current drawn from the PEMFC must be controlled so that the maximum current rating is never

exceeded. If current drawn from the PEMFC exceeds the maximum rating, then the fuel cell will

shut down as a safeguard against permanent damage. The second research goal of this study is

to develop an active power management scheme that does not use a DC-to-DC converter in the

main power path of the hybrid power system and maintains both battery and fuel cell within their
11
safe operating zones. A simulation model of this hybrid power management controller is then

created in Simscape.

To create a complete model of the UAV power system, the components are combined into a

working simulation model. The battery, fuel cell, electrical load, and power management controller

models are combined in a top level Simscape simulation model to evaluate the flight performance

of the power system. For a given flight plan, a power usage profile or flight profile can be

generated. To determine the amount of hydrogen fuel required for the given flight profile, the

simulation is run. Simulation results can then assess power system performance by inspecting

battery SOC, fuel cell current, and the power split between fuel cell and battery during flight. A

feature of the hybrid power management controller is the ability to change the fuel cell current

during flight. When this feature is used during cruise conditions, the battery may contribute part of

the load current thus lowering the amount of current the fuel cell supplies. Lower current from the

fuel cell extends the hydrogen fuel supply and therefore the flight time as well. This feature of the

controller is named current assist and can be used to extend flight endurance of the UAV. The

third research goal of this study is to determine the amount of hydrogen fuel necessary to

complete a six hour flight profile. The hydrogen consumption is to be determined with current

assist turned off and also with current assist active. Reviewing battery SOC and fuel cell current

during flight will establish if the controller has successfully maintained the battery and fuel cell

within their safe zones of operation.

During short flights, a UAV may have excess hydrogen available. For these scenarios,

recharging the battery during flight may be advantageous. For example, if a UAV is trying to climb

to maximum altitude, the battery may be a limiting factor. Using excess hydrogen to recharge the

battery inflight may allow higher altitudes to be reached. Another benefit of inflight battery

recharge is that during ground maintenance, a recharged battery may shorten service time or

allow for shorter turnaround times for back to back missions. A battery charger model is

developed and integrated into the UAV power system model. The fourth research goal of this

study is to demonstrate inflight battery recharge by simulation of the given flight profile with the

12
battery charger active. The battery SOC should have an upper limit and the battery charge should

cease when this limit is reached.

For UAV hybrid power system design, selection of the fuel cell, battery, and hydrogen storage

tank will determine the UAV flight endurance and power system weight. Although the total weight

of the power system can be determined during the design phase, the total flight time may be

difficult to determine. For a given flight profile, simulation results can be used to derive a simple

formula that applies to the particular UAV case. The formula can then be used to evaluate power

system weight versus potential flight time to help choose the best combination of components.

The fifth and final research goal of this study is to derive such a formula for the ASU Composite

UAV using a given six hour flight profile. The resulting formula is then used to create a plot of

power system weight versus flight time for combinations of commercially available fuel cells,

batteries, and hydrogen storage tanks.

13
 CHAPTER 2

LITERATURE SURVEY

2.1 Fuel Cell Powered UAV Examples

For electric powered UAVs, using hydrogen fuel cells to extend the total flight time has

become an area of possible disruptive technology. Within the aerospace community, an

increased interest in the development of experimental UAVs using fuel cell power systems has

occurred. For example, the UAV manufacturer AeroVironment Inc. developed a micro air vehicle

named the Wasp. With a wingspan of only 9 inches and a weight of 7 ounces, it carried a global

positioning system (GPS) module and two cameras for aerial reconnaissance missions. The

Wasp was powered by a lithium ion battery and set an endurance record of 100 minutes during

its initial flight in 2002 (Fuentes, 2005). By 2003, the first fuel cell powered UAV was developed

by AeroVironment. Based on the earlier Wasp design, the Hornet replaced the lithium ion battery

with hydrogen fuel cells. The fuel cells had the potential of doubling the flight endurance (Moffitt,

2010). A unique design feature of the Hornet was that the fuel cells were integrated directly into

the structure of the wings. The fuel cells had two purposes, to serve as part of the wing structure

and to supply power to the electric motor and onboard electronics. The intended benefit of

exposing the fuel cells at the top wing surface was to have plenty of airflow to supply oxygen to

the chemical reactions inside the fuel cells (Jefferson, 2002). Unfortunately, the airflow over the

fuel cells proved to be excessive during the Hornet’s maiden flight in 2003. The flight time was

reduced to only five minutes due to failure of the fuel cells. The problem was that a specific range

of water content inside the fuel cell membranes needed to be maintained, but the excessive

airflow caused the fuel cell membranes to become dehydrated (Moffitt, 2010).

14
Figure 1. Hornet UAV (www.avinc.com/images/uploads/general/7/hornet2-1_bg.jpg)

The U.S. Naval Research Laboratory (NRL) has an ongoing program to develop fuel cell

powered UAVs. Founded in 1923 by T. Edison, the Navy’s corporate research laboratory started

out with only two divisions, radio and sound (“NRL History”, 2016). Early accomplishments of the

NRL include the successful flight of the radio controlled aircraft F-5L (Keane, 2013). Since the

early years, NRL has grown to encompass 18 general areas of research with 21 divisions (“NRL

Fact Book”, 2014). The alternative energy section (6113) of the NRL chemistry division is one

group researching fuel cell powered UAVs (“DOE FCTO”, 2016). In November of 2005, NRL

completed the first test flight of a small, fuel cell powered UAV named the Spider Lion. With a 95

W hydrogen fuel cell, the 1.71 kg aircraft flew for 3 hours and 19 minutes from 15 grams of

compressed hydrogen fuel (“DOE FCTO”, 2016).

Figure 2. Spider Lion UAV (www.google.com)

The hydrogen fuel cell used in the Spider Lion was manufactured by Protonex Technology Corp.

of Southborough, Massachusetts (“USAF contract for Protonex UAV power”, 2006). The next

NRL program to advance fuel cell technology in UAV applications was the Ion Tiger program. The

goal of the Ion Tiger program was to design and test a UAV capable of carrying a 2.26 kg payload

for a flight time of 24 hours. A custom airframe was necessary since no off the shelf airframes

15
had enough fuselage volume to accommodate the fuel cell and compressed hydrogen storage

tank while still having a high enough lift to drag ratio. Instead, the airframe was designed at NRL

using carbon fiber and Kevlar construction. The fuel cell, built by Protonex, was a 550 W, PEMFC

weighing only 1 kg. The hydrogen storage method used compressed hydrogen gas stored in a

carbon overwrapped aluminum pressure vessel (Swider-Lyons, 2011). Finished in the fall of

2009, the end result was an aircraft weighing 13.8 kg with a wingspan of 5.17 meters, capable of

carrying a 2.26 kg payload for a 24 hour flight. The first test flight of the Ion Tiger occurred in

October 2009 under windy conditions. The Ion Tiger flew for 23 hours and 19 minutes carrying a

1.81 kg payload. The total flight time fell slightly short of the 24 hour goal since the hydrogen burn

rate was higher than planned due to the windy conditions. In November 2009, under calmer wind

conditions, the Ion Tiger flew for 26 hours and 1 minute carrying a 2.26 kg payload (Swider-

Lyons, 2011). Post flight analysis revealed that the October flight under windy conditions had an

average propulsion power of 326 W while the November flight (mild conditions) had an average

propulsion power of 314 W. In both cases, 500 grams of hydrogen were consumed, but the total

energy delivered during the test flights were 7600 Wh for the October flight and 8160 Wh for the

November flight (Swider-Lyons, 2011).

Figure 3. Ion Tiger UAV (www.nrl.navy.mil)

In May 2013, the Ion Tiger reached a longer flight endurance of 48 hours and 1 minute using

liquid hydrogen storage instead of compressed gas storage (“Ion Tiger Fuel Cell Powered UAV”,

2016).

16
2.2 Hybrid Powered UAV Examples

A power system is said to be hybrid when two or more power sources are combined to

deliver power to the same load. For example, a lithium ion battery combined with a fuel cell

configured such that both sources deliver power to the same electric motor is considered a hybrid

power system. For UAV applications, a hybrid power system design can significantly reduce the

size of the fuel cell. For UAV cases where the aircraft takes off from the ground under its own

power (runway), the power required during the takeoff phase of flight is much higher than the

power required during airborne cruising conditions. A rough estimate of the ratio of required

takeoff power to minimum required cruising power is about four to one. The four to one ratio

depends upon the design of the aircraft and is a ratio estimate for a fixed wing UAV developed at

Arizona State University (ASU) (C. Nam, personal communication, September 17, 2015).

Although this ratio may vary for each UAV case, it is a helpful estimate to evaluate hybrid

powered UAV design. For example, a UAV that requires 100 W of power during cruising

conditions (steady state) would need 400 W to take off from a runway. A hybrid power system

can be employed in such cases by having the battery augment the fuel cell power during takeoff.

The fuel cell can then be used to power the UAV during the cruising phase of flight without help

from the battery. In effect, the hybrid approach reduces the required fuel cell size from 400 W to

100 W. During the takeoff phase of flight, the fuel cell delivers 100 W of power combined with 300

W of power from the battery. During the cruising phase of flight, the fuel cell continues to deliver

100 W of power and the battery delivers no power at all. In this manner, the synergistic

combination of fuel cell and battery maximizes flight time while minimizing the size and weight of

the power system. An example of a hybrid fuel cell and lithium ion powered UAV was developed

in 2010 by the National Cheng Kung University in Taiwan. Named the Grey-faced Buzzard, the

30 kg fixed wing aircraft was powered by a 1 kW fuel cell and 5.4 Ah lithium ion battery.

Combined, the fuel cell and battery deliver 2.5 kW of electric power to the motor at takeoff. With a

wingspan of 3.4 m and an electric motor rated at 4 kW, there is enough lift and propulsion to carry

an 8 kg payload. The test flight of the Grey-faced Buzzard lasted 15 minutes covering a distance

of 30 km. Although the test flight of the Grey-faced Buzzard was short, it was a successful
17
demonstration of the hybrid power system (“Fuel cell/battery hybrid UAV takes off in Taiwan”,

2010). In this case, the ratio of total power to fuel cell power is 2.5 to one. If the ratio for takeoff

power to cruising power is four to one, then a maximum power of 2.5 kW at takeoff yields a

required minimum cruising power of 625 W. The actual ratio may differ, but assuming four to one

is within 25% of the actual ratio, the Grey-faced Buzzard has an excess fuel cell power capacity

of 160 W or more during cruising flight. The excess fuel cell power may be used to power the

onboard electronics, power the fuel cell BOP equipment, or to increase propulsion power during

windy flight conditions.

Figure 4. Grey-faced Buzzard UAV (www.ubergizmo.com)

Two commercially built, hybrid powered UAVs are the FAUCON H2 from EnergyOr

Technologies and the Stalker XE from Lockheed Martin. Both UAVs use fuel cell and lithium ion

battery hybrid power systems, but the type of fuel cell is different in each. The FAUCON H2

utilizes a proton exchange membrane fuel cell (PEMFC) while the Stalker XE has a solid oxide

fuel cell (SOFC). The PEMFC is a low temperature fuel cell; operating at a relatively low

temperature range, from 30oC to 100oC. On the other hand, a SOFC is considered a high

temperature fuel cell; operating in a temperature range from 500oC to 1000oC (Larminie & Dicks,

2003). The PEMFC uses compressed hydrogen gas for fuel while the SOFC in the Stalker XE

consumes propane (LPG) fuel. Both aircraft employ lithium polymer batteries to augment the fuel

cell power when needed. These two UAVs are an interesting comparison since they are of similar

size but use different fuel cell technologies. In addition, neither UAV is considered an

experimental aircraft; both are operational, ruggedized sUAS; robust enough for commercial

18
missions. Both systems have a full complement of onboard avionics along with a ground control

station. The FAUCON H2 has a wingspan of 3 m, a weight of 9 kg, and a payload capacity of 1 kg

(“EnergyOr fuel cell powered UAV reaches 10 h flight endurance”, 2011). The Stalker XE has a

wingspan of 3.66 m, a weight of 7.5 kg, and a payload capacity of 2.5 kg (“Stalker UAS”, 2016). In

2011, the FAUCON H2 completed a demonstration flight lasting 10 hours 4 minutes, then

performed a successful autonomous landing (“EnergyOr fuel cell powered UAV reaches 10 h

flight endurance”, 2011). Also in 2011, Lockheed Martin reports the Stalker XE to have an 8 hour

flight endurance after numerous flight tests (“Lockheed Martin ruggedized UAS uses AMI fuel cell

power”, 2011). Both UAV manufacturers report the hybrid power system increases flight

endurance by a factor of four over a battery only power system.

Figure 5. FAUCON H2 (left) (www.energyor.com), Stalker XE (right) (www.lockheedmartin.com)

Two UAVs constructed at the university level are the Piper Cub UAV from the University

of Johannesburg and the Composite UAV from ASU. Constructed in 2009, the Piper Cub was

built to demonstrate that a 100 W fuel cell provides enough power for steady state flight. With a

wingspan of 2.31 m and a weight of 5.3 kg, the Piper Cub flew at cruising conditions for 9

minutes. During takeoff, the Piper Cub was powered by a lithium polymer battery (LiPo) and when

it reached steady, level flight (cruising conditions), power was manually switched over to the fuel

cell by remote operation on the radio controller (Furrutter, 2009). Internally, one of the radio

control channels determined the state of an electrical relay which connected either the battery or

the fuel cell to the electric motor. At no time was the fuel cell and battery both delivering power to

the electric motor simultaneously. The PEMFC used was the Horizon Fuel Cell Technologies H-

100 and the LiPo battery type is not reported (Furrutter, 2009). In 2012, ASU constructed a

Composite UAV based on the AIAA 2012 “Design Build Fly” competition specifications. Designed

19
and built by undergraduate students, the airframe was a custom design constructed out of carbon

fiber composite materials. The Composite UAV has a 2.03 m wingspan, a weight of 5.5 kg, and a

payload capacity of 2 kg (Krauch et al., 2012). The aircraft was powered by three electric motors;

one mounted on the front nose of the fuselage and two mounted on the trailing edges of the main

wing. The initial test flight in December 2012 was under LiPo battery power only. Unfortunately,

the flight was cut short by crashing soon after takeoff. Later, a second Composite UAV was

constructed and has become the centerpiece for the continued effort at ASU Polytechnic campus

to develop a hybrid powered UAV. In the fall semester of the following year, a hybrid power plant

study was completed at ASU Polytechnic campus (Monaco, 2013). The goal of this study was to

increase the flight time of the Composite UAV from an estimated 25 minutes using LiPo batteries

only, to 6 hours using a hybrid fuel cell and LiPo battery power system (Monaco, 2013). The work

presented in this thesis is a natural continuation to further develop the hybrid power system

architecture proposed by Monaco. To compare the Composite UAV and Piper Cub UAV, both

airframes are of roughly the same size and weight. Both aircraft are of similar design; that is both

airframes have a conventional high wing placement of the main wing and a conventional tail

design. Due to this similarity, it is reasonable to expect that the 100 W fuel cell size used in the

Piper Cub UAV will be somewhat close to the minimum fuel cell size required in the Composite

UAV hybrid power system.

Figure 6. Piper Cub UAV (left) (www.google.com), ASU Composite UAV (right)
(Nam, C., lecture notes, 2015)

20
Figure 7. ASU Composite UAV CAD Assembly (left) (Nam, C., lecture notes 2015),
ASU Composite UAV Disassembled (right) (Nam, C., lecture notes 2015)

Figure 8. ASU Composite UAV Test Flight, December 2012 (Nam, C., lecture notes 2015)

2.3 Lithium Polymer Battery Description

Among the different portable, secondary battery types available, lithium ion battery

technology offers the highest energy density. Compared to other rechargeable battery types such

as nickel-cadmium (Ni-Cd), nickel metal-hydride (Ni-MH), or lead acid batteries, lithium ion

batteries are the best choice for UAV power systems. Factors such as energy density, cell

voltage, weight, volume, and discharge rate are primary considerations for battery type selection.

Secondary factors such as cost and charge retention are attributes to consider as well but do not

have disparities large enough to disqualify lithium ion technology for typical sUAS applications.
21
Lithium is the preferred anode material due to its high standard reduction potential (-3.05 V) and

its low weight (0.534 g/cm3) resulting in higher energy densities of around 150 Wh/kg

(Madakannan, lecture notes 2014). With a nominal cell voltage of 3.7 V, lithium ion batteries

obtain higher battery voltages with fewer cells connected in series. Higher voltage batteries need

to supply less current to deliver the same amount of power. When a battery delivers less current,

then the amount of stored active chemicals within the battery lasts longer. The result is a relaxed

constraint on the amount of active chemical material stored within the battery to deliver a

specified amount of energy. Due to the higher cell voltage of lithium ion batteries, fewer cells are

required to obtain a desired battery voltage. Less active chemicals need to be stored to reach a

specified energy storage goal. The end result is lower battery weight and volume for lithium ion

batteries as compared to other options. The benefit of lithium ion batteries is clarified when

considering the weight and size required to deliver 100 Wh of energy to a load (e.g. an electric

motor). Discharge rate of early lithium ion batteries used to be a restricting factor for high

discharge rate applications, but the technology has improved over time to where this is no longer

the case. Zinc-Air batteries have an even higher energy density than lithium ion batteries, but

they cannot be used in UAV applications since they have very low discharge rates. Cost of lithium

ion batteries can be three to five times that of competing technologies. Although battery cost is a

factor, lithium ion battery cost is still low enough to be the popular choice of battery type in both

the model airplane community and the UAV industry.

Table 2

Battery Type Comparison (Madakannan, lecture notes 2014, Linden & Reddy, 2002)
Battery Specific Energy Volumetric Energy Nominal Cell 100 Wh 100 Wh
Type Density (Wh/kg) Density (Wh/L) Voltage (V) Weight (kg) Volume (L)
Lead Acid 35 70 2.00 2.86 1.43
Ni-Cd 35 100 1.20 2.86 1.00
Ni-MH 75 240 1.35 1.33 0.42
Li-Ion 150 400 3.70 0.67 0.25
Zinc-Air 220 442 1.65 0.45 0.23

22
Figure 9. Battery Type Volumetric Energy Density versus Specific Energy Density
(Madakannan, lecture notes, 2014)

Although the chemistry and construction of lithium ion batteries have many varieties, the

basic operating principles are the same. During discharge, lithium ions travel from the anode to

the cathode passing through the electrolyte layer. The process is completely reversible so that

the battery can be recharged. During charge, lithium ions travel from the cathode back to the

anode. Typical anode electrode material consists of lithiated carbon, LixC6. The layered carbon

anode material acts as a host matrix (crystal lattice) for lithium ions. In other words, lithium ions

can be freely inserted into and extracted from gaps between layers of the carbon material crystal

lattice; a process known as intercalciation. During discharge, lithium ions are extracted from the

anode crystal lattice. When a lithium ion is extracted, the anode material is oxidized resulting in a

free electron which is available to be delivered to the external electrical load. The lithium ion then

travels through the electrolyte to the cathode where it is inserted into the cathode crystal lattice.

When the lithium ion is inserted into the cathode crystal lattice, the cathode material is reduced

thus absorbing an electron from the external electrical load. Typical cathode electrode materials

vary; a common one being LiCoO2. Since lithium ions may be inserted into and extracted out of

the anode and cathode electrode materials, these materials are named lithium insertion

compounds. This process of intercalciation is the core principle behind every type of lithium ion

battery. The electrolyte is the layer separating the anode and cathode. An electrolyte material

must be an insulator to electrons and a good conductor of lithium ions. Electrolytes may be a
23
liquid, a gel, or a solid, such as a polymer or a ceramic. In many commercial lithium ion batteries,

the electrolyte is a liquid, consisting of an organic solvent with inorganic solutes. A typical liquid

electrolyte may consist of a Tetrahydro furan (THF) solvent with a LiPF6 solute dissolved at a

concentration of about 1 molar (Madakannan, lecture notes 2014).

Anode discharge reaction: Li xC  Li x 0.5C  0.5Li   0.5e  (1)

Cathode discharge reaction: Li 0.5CoO2  0.5Li   0.5e   LiCoO2 (2)

Overall discharge reaction: Li 0.5CoO2  Li xC  Li x 0.5C  LiCoO2 (3)

A Lithium Polymer (LiPo) battery is a type of lithium ion battery, but the name is used in

two different ways. Within the scientific and industrial research communities, the name lithium

polymer battery means a lithium ion battery where the electrolyte is a polymer. As shown in figure

9, lithium polymer batteries have even higher energy densities than lithium ion batteries with

liquid electrolytes. These types of LiPo batteries are widely researched, but are not commonly

available commercially. The second meaning of lithium polymer battery is used by the battery

industry. In this context, a lithium polymer battery is a lithium ion battery with a liquid electrolyte,

but it is packaged in a prismatic, flexible package. The polymer in this case refers to the casing of

the battery and not the electrolyte. LiPo batteries that are commonly available to consumers are

of the second type; that is a lithium ion battery with liquid electrolyte packaged in a flexible,

polymer casing (“Lithium polymer battery”, 2010).

2.4 Lithium Polymer Battery Ratings

For small model, electric aircraft applications, hobbyists typically use lithium polymer

batteries (LiPo). The same battery type is a good match for sUAS applications as well. The hobby

aircraft industry has well established that lithium polymer batteries are the best choice to minimize

weight while maximizing flight endurance. The rating system for these batteries has two key

parameters, the battery configuration and the maximum discharge rate. The battery configuration

consists of a four character designator xSyP. The x placeholder is a number typically from 1 to 5.

It is the prefix to the S designator which stands for the number of cells in series. For example, a

24
4S designator indicates that four lithium ion cells are in series within the battery pack. At 3.7 volts

per cell, a 4S battery pack will have a battery pack voltage of 14.8 volts. The y placeholder is a

prefix to the P designator, which indicates the number of cell strings in parallel. For example, a 2P

designator indicates that two lithium ion cell strings are in parallel. The more cell strings in

parallel, the greater the battery pack capacity. To clarify, a 4S2P designator indicates that there

are four lithium ion cells in series within a string, and that two strings are present in the battery

pack connected in parallel. Many times, the yP portion of the designator is left out, meaning that

there is only one cell string present in the battery pack. For example, a LiPo battery pack

designator of 4S indicates that there are 4 lithium ion cells connected in series and that there is

only one string present in the battery pack. The second key parameter for LiPo battery packs is

the maximum discharge rate. Measured in units of C (capacity), the discharge rate is listed as a

multiple of C. For example, if a battery pack has a capacity of one amp-hour, then a maximum

discharge rating of 20C indicates that the battery pack is capable of delivering an instantaneous

current of 20 amps.

2.5 Lithium Ion Battery Model

There are a variety of methods to model lithium ion battery packs for purposes of system

simulation. Reported in literature are empirical, electrochemical, electric circuit, and abstract

model types (Tsang, 2010). The goals of the system level simulation will determine which battery

model is appropriate. At the most basic level, common to all models, is the prediction of battery

voltage as a function of current. An electrical circuit for such a battery model consists of an ideal

voltage source and a series resistor. The ideal voltage source represents the open circuit battery

voltage while the series resistor represents the internal resistance of the battery. In the equation 4

below, Vbat is the terminal voltage, Eo is the ideal voltage source value, R is the internal resistance

of the battery, and the independent variable i represents the battery current.

Vbat  Eo  i  R (4)

With zero current draw, the battery voltage measured across the positive and negative terminals

of the battery (terminal voltage) is equal to the ideal voltage source value. As current drawn from

25
the battery increases, the terminal voltage drops linearly from the ideal voltage value. For this

simple battery model, only the internal resistance of the battery is accounted for; key parameters

such as total battery capacity and battery SOC are not included. The limitation of this model is

that the battery will run forever; power will continue to be delivered regardless of battery capacity,

the initial SOC, and the history of delivered power. A more realistic model will include the limited

electrical capacity of the battery. For example, a one amp-hour battery capacity states that a

battery, starting out at 100 percent SOC, is able to supply one amp of current to an electrical load

for one hour. During this discharge process, the battery voltage will decrease until it reaches a

final value close to zero. At this point, the battery is completely discharged and can no longer

deliver any electrical power. A more suitable model for the system level simulation purposes of

this study is presented by Olivier Tremblay et al. and is discussed further in section 3.2 (Tremblay

& Dessaint, 2009).

2.6 Proton Exchange Membrane Fuel Cell Description

Although the PEMFC operates on many of the same electrochemical principles as

batteries, it is fundamentally different than a battery. Both a battery and a PEMFC use chemical

oxidation and reduction to produce electrical energy. The main difference between them is the

source of chemical reactants. In a battery, the chemical reactants are stored within the battery,

thus limiting the electrical capacity. In a PEMFC, the chemical reactants are external; fed to the

fuel cell during operation. The PEMFC ideally has an infinite capacity; as long as hydrogen fuel

and air are continually fed to the fuel cell, it will continue to generate electrical power. Of course a

practical PEMFC has endurance limitations due to degradation during operation, but the

mechanisms that degrade PEMFC operation over time are different than the limited amount of

chemical reactants that limit battery capacity. The physical construction of a PEMFC is quite

different from a battery. The PEMFC accepts input chemical reactants in gaseous form while

most batteries store chemical reactants as a solid or liquid. The PEMFC must also have a

mechanism to purge water produced from the chemical reaction while batteries do not have this

issue. A PEMFC also has to regulate humidity of the input gaseous reactants to maintain

membrane hydration during operation while batteries do not. PEMFC must also manage its
26
temperature during operation. If the PEMFC runs too hot, then the membrane will become

dehydrated and the fuel cell will stop functioning. Batteries in general do not have to regulate their

temperature during operation. There may be some esoteric cases where battery temperature may

have to be regulated, but these cases are not the norm. Typically, a PEMFC has a

microprocessor to maintain proper operating conditions during operation, otherwise known as a

fuel cell controller. All of the equipment to maintain proper conditions of the PEMFC during

operation are grouped together and labeled balance of plant (BOP) equipment. PEMFC BOP

equipment include the fuel cell controller, temperature management equipment, water content

management equipment, and possibly compressors to boost gas intake pressures. Clearly, a

battery does not have all this BOP equipment.

A description of PEMFC operation can be started by describing the chemical reactions at

the electrodes. Hydrogen gas (fuel) is fed to the anode. At the anode, gaseous, diatomic

hydrogen molecules are oxidized resulting in two hydrogen ions and two electrons.

Anode oxidation: H2  2e   2H  (5)

The hydrogen ions pass through the electrolyte to the cathode. Same as battery electrolytes, the

PEMFC electrolyte must be an insulator to electrons while being a good ionic conductor. The free

electrons are available to the external circuit and their flow from the anode to the cathode must

pass through the external electrical load and not through the PEMFC electrolyte. At the cathode,

electrons return to the fuel cell from the external electrical load and combine with the hydrogen

ions and oxygen gas extracted from the air fed to the cathode.

1
Cathode reduction: 2e   2H   O2  H 2O (6)
2

The end result is that hydrogen gas (the fuel) and air (the oxidizer) are combined to yield water,

electrical energy, and heat.

1
Overall reaction: H2  O2  H2O (7)
2

27
 At the core of the physical construction of a PEMFC is the membrane electrode assembly

(MEA). The central component of the MEA consists of a thin sheet of polymer material which acts

as the fuel cell electrolyte; also called a membrane. Typically, the membrane is a special polymer

developed by the DuPont Corporation named NafionTM. Similar to Teflon, Nafion has the special

property of conducting positively charged hydrogen ions while acting as an insulator to electrons.

Typical thickness of the Nafion membrane ranges from 18 um to 25 um. Working outwards from

the center, on both sides of the membrane are the catalyst layers; one side being the anode

catalyst layer and the other being the cathode catalyst layer. The catalyst layers are essentially

the same on both sides, consisting of fine platinum particles dispersed over a supporting structure

of carbon powder immersed in a polymer binding agent. The details of the catalyst layers are

critical to efficient fuel cell operation. The thickness of the catalyst layers range from 5 um to 30

um. Moving outwards from the catalyst layers are the gas diffusion layers (GDL); one on the

anode side and the other on the cathode side. The GDLs are constructed from porous carbon

paper or cloth which allow the reactant gasses to diffuse through them. Typical thickness of the

GDLs range from 174 um to 450 um. Next, the outer layers of the MEA are the electrode plates;

one anode plate and one cathode plate. The electrode plates are made of either a graphite

composite material or metal. The electrode plates provide physical support for the entire MEA

and are the outer plates used to compress the sandwich structure of the MEA. Among the other

functions implemented by the electrode plates, they provide a conduction path for electrons in

and out of the MEA and form the gas flow channels to deliver reactant gas to the GDLs

(Madakannan, lecture notes 2016).

28
Figure 10. Membrane Electrode Assembly

A more detailed discussion of the MEA reveals many of the material requirements of

each layer which in turn illustrates the need for the multi-functional role that each layer provides.

At the center of the MEA is the Nafion membrane. Constructed from a sulphonated fluoro-

polymer material, the membrane requires a specific range of water content to conduct hydrogen

ions (Larminie & Dicks, 2003). Note that a positively charged hydrogen ion is a single proton, and

that the membrane is designed to be a proton conductor when properly hydrated. This is the

basis for the PEMFC name, proton exchange membrane fuel cell. One of the engineering

challenges of PEMFC design is to keep the membrane properly hydrated during operation. The

Nafion material has a fixed ratio of hydrophobic and hydrophilic properties such that it retains the

right amount of water if available. In other words, the Nafion membrane can dry out if not enough

water is present at its surfaces, but it will not become too hydrated if an excess amount of water

is available. To keep the membrane hydrated, the input hydrogen gas is humidified. Water vapor

carried by the hydrogen gas diffuses through the anode GDL and anode catalyst layer to the

membrane. On the cathode side, the cathodic reaction produces water at the cathode catalyst

layer. To prevent excess water from blocking the pours of the cathode catalyst layer and the
29
cathode GDL (flooding), it must be removed. Water diffuses outwards from the cathode catalyst

layer, through the cathode GDL, to the cathode plate, where it evaporates to the air flow. In

practice, the water management problem is more complicated, but this simplified explanation

keeps the description brief. Heat generated by the chemical reactions at the anode and cathode

catalyst layers must also be removed to keep the operating temperature of the fuel cell below

100oC. The reason for the upper temperature limit is to keep the membrane hydrated. An

operating temperature above 100oC will cause the membrane to dry out. Heat is conducted away

from the catalyst layers by the GDLs and ultimately dissipated by the electrode plates. Electrons

produced at the anode catalyst layer must be conducted through the anode GDL to the anode

plate to supply power to the external circuit. On the cathode side, electrons return to the cathode

plate from the external circuit and are conducted through the cathode GDL to the cathode catalyst

layer. Gas transport must occur on both the anode and cathode sides. Reactant gasses must

diffuse from the flow fields at the electrode plates, through the GDLs to reach the catalyst layers.

The catalyst layers must have a highly porous structure to maximize the surface area available

for efficient catalytic action. To summarize, the material properties required for each layer are

grouped together. For the membrane, the Nafion material must be uniformly hydrated and kept

below 100oC. When properly hydrated, the membrane is a good proton conductor and an

electronic insulator. The catalyst layers must be highly porous with an even distribution of

platinum catalyst. They must allow gas and water transport and be highly conductive both

thermally and electronically. The GDLs must also allow gas and water transport and be highly

conductive thermally and electronically. Last, the electrode plates must be mechanically strong to

support the MEA structure. They must be electronically conductive to deliver electrical power out

of the MEA. The electrode plates contain channels for reactant gas flow which not only delivers

the reactant gasses to the MEA, but also provides a location to humidify the input hydrogen gas

or evaporate water to the air flow. The electrode plates must be thermally conductive to dissipate

excess heat generated within the MEA.

To construct the fuel cell, the MEAs are stacked so that they are effectively connected in

series. The anode plate of one cell is connected to the cathode plate of the next cell to create a
30
series electrical connection. For example, to construct a fuel cell with a 12 V output, if each cell

produces a voltage of 0.8 V, then 15 MEAs must be used to produce 12 V across the stack. Since

the cell voltage (MEA voltage) is a low value (e.g. 0.8V) and the cell current is high (e.g. > 3 A),

even a small contact resistance between cells will cause a significant loss in fuel cell stack

voltage. To minimize voltage loss at the connections between MEAs, the electrode plates are

combined. The anode plate of one cell is combined with the cathode plate of the next cell into one

physical plate called a bipolar plate. The use of bipolar plates over separate anode and cathode

plates is common and significantly reduces voltage losses of the fuel cell. The complication of

designing bipolar plates is the balance between minimizing electrical resistance between cells

and maximizing gas flow for the cells. Each bipolar plate must deliver hydrogen gas on its anode

side and deliver air on its cathode side. The channels for gas flow are thus formed on both sides

of the bipolar plate. There is a tradeoff between channel size and minimizing electrical resistance.

Striking a balance between these two concerns is one challenge of bipolar plate design. A second

pair of concerns is the thickness of the bipolar plate and the mechanical strength of the plate.

Minimizing the thickness of the bipolar plate minimizes the electrical resistance and weight of the

fuel cell but it reduces the mechanical strength. Traditionally, Poco graphite has been the

preferred material for bipolar plate construction. Properties such as low electrical contact

resistance, good mechanical strength, low weight, and the ability to withstand a corrosive

environment, are advantageous for bipolar plate construction. The disadvantages of using Poco

graphite are the cost, the brittleness of the material, and the difficulty of machining. For very thin

bipolar plates made from Poco graphite, mechanical stress and shock can crack the material.

High material cost and expensive machining processes contribute to higher fuel cell prices. Using

metal as an alternative to Poco graphite has been investigated and reported in literature.

Although metals are less expensive, less prone to crack under stress and shock, and are easier

to machine, they tend to corrode which causes higher resistances at contact points (Tawfik,

2006).

A practical PEMFC system consists of more than the fuel cell stack alone; it also includes

the BOP equipment and the hydrogen storage system. Using a commercial example of the
31
Horizon Fuel Cell Technologies H-100 fuel cell and a compressed hydrogen storage tank, figure

11 shows the complete system.

Figure 11. Complete Fuel Cell System (Horizon, 2013)

The H-100 fuel cell stack consists of 20 MEA cells and has a nominal operating point of 12 V at

8.3 A. At the nominal operating point, the cell voltage is 0.6 V. The maximum stack temperature is

65oC and the fuel cell stack is rated to operate between 5oC and 30oC ambient temperature

(Horizon, 2013). The BOP equipment that ships with the H-100 fuel cell include the fuel cell

controller, a fan on the fuel cell stack, a hydrogen supply valve, and a hydrogen purge valve. For

the H-100 fuel cell, the water management is self-contained inside the fuel cell stack (Horizon,

2013). The fuel cell controller gets input data from the temperature sensor and controls the fan to

regulate fuel cell temperature; the thermal management system. The fuel cell controller also

controls the hydrogen supply and purge valves. For the hydrogen storage system, the storage

tank contains compressed hydrogen at pressures up to 310 bar. The hydrogen pressure is

reduced in two stages. The high pressure regulator is included with the hydrogen storage tank

and steps down the hydrogen pressure from 310 bar down to 27.6 bar. The second stage, or low

pressure regulator, steps the hydrogen pressure down from 27.6 bar to 0.5 bar. The low pressure

regulator is purchased separately. Note that the fuel cell controller needs an external power

supply.

32
2.7 Proton Exchange Membrane Fuel Cell Polarization Curve

To characterize the performance of a fuel cell, a graph of the stack voltage versus the

fuel cell current can be used. Known as a polarization curve, the graph shows the operating

voltage over the full range of fuel cell operating current. The polarization curve can be measured

empirically by measuring the stack voltage as a connected electrical load is varied. The first point

to measure is the open circuit voltage; that is the stack voltage present when the electrical load is

disconnected. The electrical load is then connected and decreased from 10 mega-ohms down to

the lowest value (maximum electrical load) where the stack voltage is close to zero. Fuel cell

theory describes the features of the polarization curve from first principles. The theoretical model

can produce the polarization curve of a PEMFC design by adjusting the parameters of an

analytical model. The theoretical model can be thought of as an equation with four contributing

terms, or it can be represented as an electrical circuit. The four terms in the model are: the open

circuit voltage, activation loss, ohmic loss, and the concentration loss. In thermodynamic lingo,

the open circuit voltage represents the reversible voltage and the three loss terms represent

irreversible losses. In electrochemical language, the losses are referred to as overpotentials.

The open circuit voltage is a result of the chemical reactions taking place inside the fuel

cell. In other words, the change in chemical potential energy due to the change in chemical

configurations (anode and cathode reactions) result in an electrical voltage at the fuel cell

terminals. To quantify the change of chemical potential energy, the thermodynamic quantity

called the Gibbs free energy is commonly used. Gibbs free energy can be defined for a chemical

reaction as the energy available to do external work. The key point for determining the energy

released in a chemical reaction is that it is the change in chemical potential energy that is

available to perform external work. This is similar to a mechanical situation where a ball is at the

top of a hill. As the ball rolls down the hill, its velocity increases. The ball gains kinetic energy

which is equal to the change of potential energy. Another way of thinking about the chemical

reactions inside the fuel cell is that they change chemical potential energy into electrical potential

energy. For the overall chemical reaction inside the fuel cell, a hydrogen diatomic molecule is

33
combined with half of an oxygen diatomic molecule to yield a water molecule plus heat and

electrical energy.

1
H 2  O2  H 2O (8)
2

Analyzing this chemical reaction in more detail, the hydrogen, oxygen, and water molecules all

have a Gibbs free energy of formation. The difference between the Gibbs free energy of

formation of the product (the water molecule) and the Gibbs free energy of formation of the

reactants (the hydrogen and oxygen molecules) yields the chemical potential energy released

from the reaction.

 
g f  gf
H2O


 
  gf
H2

1
2
 
gf
O2



(9)

Note that the lower case g with a bar above it denotes the Gibbs free energy per mole. Next,

consider the chemical reactions at the anode and cathode.

Anode reaction: H2  2e   2H  (10)

1
Cathode reaction: 2e   2H   O2  H 2O (11)
2

In these reactions, two electrons are released by the anode reaction and two electrons are

combined in the cathode reaction. Using conservation of energy, the chemical potential energy

released must equal the electrical energy produced.

gf  n  F  E o (12)

where: n = number of electrons released in the chemical reaction

F = 96485 coulombs per mole (Faraday’s constant)

Eo = open circuit voltage of fuel (V)

 g f
Eo  (13)
n F

34
The minus sign in equation 13 accounts for the negative charge of the electron. The change in

Gibbs free energy is negative since energy is released from the chemical reaction, so the minus

signs cancel and give a positive fuel cell voltage. That is, the fuel cell cathode is positive relative

to the anode. The number of electrons participating in the chemical reaction must be accounted

for. This is similar to the mechanical analogy of two balls being placed at the top of a hill. Last,

there is the heat produced by the reaction. The amount of heat produced is the difference

between the change in enthalpy and the change in Gibbs free energy. Simply put, the change in

enthalpy is the total change of chemical potential energy, including any internal losses. The

change in Gibbs free energy is the total change of chemical potential energy minus any internal

losses. The internal losses are inherent to the chemical reaction and are thermodynamically

irreversible. The mechanical analogy to these internal chemical losses would be friction. If a ball

rolls down a hill, the kinetic energy gained is equal to the difference in potential energy minus

energy lost to friction. The energy lost to friction manifests as heat. Similarly, the chemical energy

lost to internal chemical losses manifest as heat. Additionally, the internal chemical losses also

determine the maximum theoretical efficiency of the fuel cell. The maximum efficiency a fuel cell

can have is simply the ratio of the change of Gibbs free energy to the change of enthalpy.

gf
max   100%  (14)
 hf

For example, at 60oC for the PEMFC reaction, the Gibbs free energy is -231.5 kJ/mole and the

enthalpy is -284.7 kJ/mole (Revankar & Majumdar, 2014). The maximum theoretical efficiency of

the PEMFC at 60oC is then 81.3%. The open circuit voltage also depends upon input gas

pressures and fuel cell temperature. Including these parameters yields the Nernst equation for

the PEMFC reaction. The Nernst potential is close to the true open circuit voltage of the fuel cell.

The actual open circuit voltage has to account for fuel crossover which is discussed at the end of

this section.

o
g f
Nernst equation: Eo 
2F

RT
2F
 
 ln PH2 
RT
4F
 
 ln PO2 (15)

35
 where: E o = open circuit voltage of PEMFC without accounting for fuel crossover (V)

gf o = change of molar Gibbs free energy at standard conditions (kJ/mole)

PH2 = partial pressure of hydrogen (atmospheres)

PO2 = partial pressure of oxygen (atmospheres)

R = universal gas constant (8.314 J/(mole oK))

T = temperature (K)

F = Faraday’s constant (96485 Coulombs per mole)

At the electrodes, the chemical reactions have a finite rate of reaction. Due to the limited

reaction rate at the electrode surfaces, a portion of the cell voltage is used to drive the chemical

reactions. This energy loss is known as activation loss. It is the dominant energy loss mechanism

at low fuel cell currents. At zero fuel cell current, the activation loss is negligible. As fuel cell

current increases from zero, the activation loss increases logarithmically which give the

polarization curve a distinctive exponential decay for small fuel cell currents (Madakannan,

lecture notes 2016). Even when there is no fuel cell current, there are reactions taking place at

both electrodes. Transfer of electrons from the electrode to the electrolyte is balanced by a

transfer of electrons from the electrolyte to the electrode so that there is a net zero current across

the interface. At zero fuel cell current, the current density of just the forward reaction is known as

the exchange current density, io (mA/cm2). The magnitude of the exchange current varies,

depending upon the chemical activity at the electrode surface; typically ranging from 10-6 mA/cm2

to 102 mA/cm2. Factors such as electrode and electrolyte materials, electrode surface porosity,

catalyst activity, and temperature all contribute to the magnitude of the exchange current. Larger

values of the exchange current result in lower activation losses. For the PEMFC, the rate of

reaction is much faster at the anode than the cathode. The exchange current density is much

higher at the anode than at the cathode. The end result is that the activation loss at the cathode is

the dominant contributor the overall fuel cell activation loss. An equation for the activation

overpotential is given below in equation 16 (Madakannan, lecture notes 2016).

36
 RT i 
Vact   ln   (16)
2   F  io 

where: T = temperature (K)

R = universal gas constant (8.314 J/(mole oK))

i = fuel cell current density (mA/cm2)

 = charge transfer coefficient (typically 0.5 for PEMFC)

i o = exchange current density (mA/cm2)

The second loss mechanism in PEMFCs is ohmic losses. A combination of electronic

resistance of the electrodes and cell interconnects and the ionic resistance of the membrane,

ohmic losses drop an internal voltage equal to the product of total internal resistance and fuel cell

current. Ohmic loss is the dominant loss mechanism at midrange fuel cell currents.

Vohmic  i  R (17)

where: i = current density (A/cm2)

R = internal resistance of cell (   cm 2 )

The third loss mechanism in PEMFCs is concentration loss. When current is drawn from

the fuel cell, the reactant gas concentrations at the electrode-electrolyte interface is lowered. As

fuel cell current increases, the reactant concentration at the electrode continues to decrease. At

low current densities, the depletion of reactant gas concentration is small and the voltage loss

due to concentration loss is negligible. However, when the fuel cell current is large, the gas

concentration depletion at the electrode surfaces become significant and the concentration loss

becomes a limiting factor. In other words, the reactant gasses have a maximum diffusion rate to

the electrode surfaces. When reactant gasses can no longer be supplied fast enough, the

reactant concentration at the electrodes drops to zero, which in turn causes the fuel cell voltage

to drop to zero. Concentration losses are only significant near the maximum fuel cell current and

are the limiting mechanism for maximum fuel cell current (Madakannan, lecture notes 2016).

37
  J 
Vcon  B  ln  1   (18)
 Jmax 

where: B = concentration loss coefficient (V)

J = fuel cell current density at the membrane (A/cm2)

Jmax = fuel cell maximum current density (A/cm2)

Combining the Nernst voltage and the three loss mechanisms gives the working voltage

of one cell in the fuel cell stack.

Vcell  E o  Vact  Vohmic  Vcon (19)

The total output voltage of the fuel cell stack is then the sum of cell voltages present in the fuel

cell stack.

Vstack  n  Vcell (20)

where: Vstack = fuel cell stack voltage (V)

n = number of cells (MEAs) in fuel cell stack

Vcell = cell (MEA) voltage (V)

Figure 12. PEMFC Polarization Curve

38
 For the complete polarization curve shown in figure 12, there are three regions: the

activation region, the ohmic region, and the concentration region. In each region, the

corresponding loss mechanism is dominant. Note that the open circuit voltage is less than the

Nernst voltage. This is due to the last loss mechanism, fuel crossover. Due to the fact that the

membrane is not perfect, some fuel leaks over from the anode to the cathode and some oxygen

leaks over from the cathode to the anode. Also, to a small extent, some electrons leak through

the membrane as well. The result of fuel crossover is that recombination occurs which causes a

constant voltage loss across the entire polarization curve; usually around 0.2 volts (Larminie &

Dicks, 2003). The end result is that the actual open circuit voltage of a single cell is slightly less

than one volt and that the open circuit voltage for the entire fuel cell stack is the number of cells

times the cell voltage. For example, the Horizon H-100 PEMFC has 20 cells and an open circuit

voltage of 19 V (Horizon, 2013). The open circuit voltage of one cell is calculated to be 0.95 V.

2.8 The Electric Double Layer

At the electrode-electrolyte interface, there is a buildup of charge due to the difference in

material of the electrode and the electrolyte. On the electrode side, a thin layer of electrons

accumulates at the surface, while on the electrolyte side, a thin layer of cations accumulate near

the electrode surface. The net effect of the charge accumulation is that of a capacitor. This effect

is universal in solid electrodes immersed in liquid electrolyte systems. The first model of the

electric double layer was developed by H. Helmholtz in the 1850’s. In the Helmholtz model, the

negative and positive charge accumulation was modeled as a parallel plate capacitor. The

electron charge concentration at the electrode surface is analogous to the negative capacitor

plate. The cation concentration in the electrolyte near the electrode surface takes on the role of

the positive capacitor plate. The effective distance between the plates is the radius of one

solvated cation which results in a parallel plate capacitor model (Revankar & Majumdar, 2014).

 A
C (21)
d

where: C = effective capacitance of the electric double layer (Farads)

 = electric permittivity of electrolyte (Coul2/N cm2)

39
 A = electrode surface area immersed in electrolyte (cm2)

d = distance between negative and positive capacitor plates (cm)

Due to the charge accumulation of the electric double layer, a voltage difference is present across

the interface and is the source of the activation loss discussed earlier. The electrode surface is

very porous and has an effective surface area roughly103 times greater than the simple geometric

surface area of the electrode. The distance between the charge layers is typically about 20

angstroms which results in an effective capacitance of several farads for a common PEMFC

(Larminie & Dicks, 2003). Since the Helmholtz model, more refined models of the electric double

layer have been developed such as the Gouy-Chapman model in 1910, the Stern model in 1924,

and the Grahame model in 1951 (Revankar & Majumdar, 2014). The capacitance due to the

electric double layer affects the dynamics of the fuel cell response to sudden changes in current.

The voltage has a delayed response to a step increase or decrease in current load. As the

current load changes abruptly, the voltage response comes to its equilibrium value in a smooth

curve-like fashion, similar to the voltage response of an RC circuit. An electric circuit model of the

double layer phenomena places a capacitor in parallel with a resistance that represents the

activation and concentration losses.

2.9 Hydrogen Storage

An important and challenging topic for hydrogen powered and hybrid powered UAVs is

hydrogen storage. Currently, there are four methods of hydrogen storage: compressed gas, liquid

hydrogen, solid state chemical, and physical absorption (Zhang, 2014). The compressed gas and

liquid hydrogen methods are well developed technologies and are used in many hydrogen

powered UAV applications. For example, the NRL Ion Tiger has used compressed hydrogen for

its 2009 demonstration flight and liquid hydrogen in its 2013 flight (Swider-Lyons, 2011, “Ion Tiger

Fuel Cell Powered UAV”, 2016). Although compressed gas and liquid hydrogen are considered

mature hydrogen storage technologies, there is continuing innovation to increase the volumetric

and gravimetric density. To use target metrics from the automobile industry, the Department of

Energy (DOE) has published design goals for hydrogen storage in light duty vehicles. In the

Freedom Car Fuel Partnership report published in 2009, the DOE lists 2015 target values for the
40
volumetric density and gravimetric density as 40 g/L and 5.5% respectively (“Targets for Onboard

Hydrogen Storage Systems for Light-Duty Vehicles”, 2009). The gravimetric density is another

way of stating the mass density, but includes the weight of the storage tank. Gravimetric density

is useful in that it measures the effectiveness of the storage system. For example, if a

compressed gas tank can store hydrogen at a very high mass density, but has an enormous

weight, then it will not be very useful for vehicle applications. Gravimetric density takes the

storage tank weight into account by stating the ratio of hydrogen weight to the storage tank

weight plus hydrogen weight.

mH2
Gravimetric density: g   100% (22)
mtan k  mH2

There is a wide variety of compressed gas tanks used in experimental UAV applications.

Typically, low end UAV compressed gas storage tanks are compressed air tanks from the

paintball industry. For example, the University of Johannesburg Piper Cub used a 0.255 kg CO2

paintball cylinder to store hydrogen. Although the cylinder was capable of holding 220 bar of

pressure, the Piper Cub team only filled it to 30 bar (Furrutter, 2009). This may have been done

for safety reasons. The aluminum cylinder was designed to store CO2 and not hydrogen. If the

hydrogen gas diffuses into the aluminum cylinder walls, it can weaken the aluminum cylinder

through hydrogen embrittlement. Considered as a type of corrosion, hydrogen embrittlement

causes a loss of metal ductility similar to stress corrosion cracking. The result is that the energy

required to form stress cracks is reduced and the strength of the aluminum cylinder is derated

(Madakannan, lecture notes, 2015). In some commercial storage tanks, a polymer liner is used to

combat hydrogen embrittlement (Zhang, 2014). In contrast to the paintball storage tank, high end

compressed gas storage tanks have been developed for UAV applications. For the Ion Tiger

project, the NRL team used a custom aluminum tank with a carbon overwrap. The carbon

composite overwrap reinforces the aluminum tank. The carbon overwrapped aluminum hydrogen

tank weighed 3.6 kg with a water volume of 22 L and had the capacity to store 500 g of hydrogen

(Swider-Lyons, 2016). Using the ideal gas equation, the calculated pressure of the tank is found

41
to be 280 bar (about 4000 psi). There is a discrepancy though, the reported pressure of the tank

is 5000 psi (Swider-Lyons, 2016). When highly compressed, hydrogen can no longer be treated

as an ideal gas. At high compressions, intermolecular forces can no longer be ignored and a

compressibility factor must be introduced (Zhang, 2014).

P V  n  Z  R  T (23)

where: P = pressure of tank (N/m2)

V = water volume of tank (m3)

n = number of moles

Z = compressibility factor (unitless)

R = 8.314 J/(mol oK), the ideal gas constant

T = temperature (oK)

Taking the ratio of 5000 psi over 4000 psi, the compressibility factor must be 1.2. Referring to the

compressibility factor chart, a compressibility factor of 1.2 is a valid figure for 4000 psi at 300oK

(Zhang, 2014). The performance of the carbon overwrapped aluminum storage tank used by the

NRL team is excellent, coming in at a volumetric density of 22.7 g/L and a gravimetric density of

12.2%. Compared to the DOE 2015 targets, the NRL storage tank falls short of the 40 g/L

volumetric density goal but exceeds the 5.5% gravimetric density goal. Next, a commercially

available compressed air paintball storage tank, the Ninja ProTM 90/4500, is evaluated. Made out

of carbon fiber, the 90 cubic inch compressed air tank can hold up to 4500 psi (310 bar) of

pressure. If used for hydrogen storage, and including a 20% safety factor (pressurizing to only

248 bar), the Ninja Pro 90/4500 tank can hold 24.6 g of hydrogen. With a weight of 1.45 kg and a

water volume of 3.56 L, the volumetric density works out to be 6.9 g/L and the gravimetric density

is 1.7% (“Ninja Pro v2 90/4500 specs”, 2016). Considering the output flow of hydrogen at

standard temperature and pressure (STP), the Ninja Pro v2 90/4500 tank can deliver 298.9 liters

of hydrogen. If the Ninja Pro v2 90/4500 tank is delivering hydrogen to a Horizon H-100 PEMFC

(feed rate of 1.3 SLPM), it can keep the fuel cell running at full power (100 W) for 3 hours and 50

42
minutes. In other terms, the Ninja Pro v2 90/4500 supplying a Horizon H-100 fuel cell, can store

383.2 Wh of energy.

Figure 13. Ninja Pro v2 90/4500 Compressed Air Tank (www.ninjapaintball.com, 2016)

2.10 Hybrid Power Systems Management

Typically, APM system design for UAV applications use DC-to-DC converters. One or

more power sources are connected to the electrical bus through DC-to-DC converters. The

function of the DC-to-DC converter is essentially that of a voltage regulator. The input voltage to

the DC-to-DC converter may vary over a wide range while the converter maintains the output

voltage at a fixed setpoint regardless of the load current. With many off-the-shelf DC-to-DC

converters, the output voltage is fixed and cannot be changed. For most of the DC-to-DC

converters used in APM systems, the output voltage is not fixed and can be controlled by a

microprocessor. A large majority of the APM systems discussed in UAV APM research literature

use one or more DC-to-DC converters in the main power path to control power flow. The voltage

sent to the power bus from one or more sources is controlled by a DC-to-DC converter which

determines the power split. Figure 14 shown below illustrates three circuit concepts that use DC-

to-DC converters.

43
Figure 14. APM Circuit Concepts using DC-to-DC Converters

In part (a) of figure 14, the fuel cell is connected to the bus through a DC-to-DC converter while

the battery is connected directly. In this case, the battery voltage determines the bus voltage. The

fuel cell voltage is varied just below the bus voltage to control the power split. If the fuel cell

voltage is adjusted above the battery voltage, then the fuel cell supplies 100% of the ESC load

and the battery may recharge. Depending upon the battery SOC, the bus voltage may be less

than the fully charged battery voltage. The microprocessor must monitor the bus voltage to

accurately control the power split. In part (b), the fuel cell is connected directly to the bus and the

battery is connected through a DC-to-DC converter. In this case, the fuel cell voltage determines

the bus voltage. Here, the battery voltage is adjusted to control the power split to the ESC load.

The battery voltage is adjusted just above the bus voltage to determine the power split. If the

battery voltage is set below the bus voltage, then the fuel cell delivers 100% of the load power

and may charge the battery. Note that the DC-to-DC converter is two-way; that is power can

propagate in both directions. A bi-directional DC-to-DC converter is needed if battery recharge is

desired. As in the circuit presented in figure 14, part (a), a supervisory microprocessor must

monitor the bus voltage in order to accurately control the power split. In part (c), both fuel cell and

battery are connected to the bus through DC-to-DC converters. This is the most flexible scenario

44
and has the advantage that the bus voltage can be maintained at an optimum value for the ESC

load. In this scenario, the relative values of the fuel cell and battery voltage will determine the

power split to the ESC load. This third scenario is unique in that two DC-to-DC converters are

connected directly to each other through the bus. There are several methods to control the power

split between the two converters, namely the droop-share method or the driver booster method

(“Current Sharing in Power Arrays”, n.d.). In the droop share method, the output impedance of

one converter is varied to control the power split. An error signal is used in the DC-to-DC

converter control loop to cause the output voltage to be a function of both the setpoint and the

converter output current. In the driver booster method, one converter acts as a driver and the

second converter acts as a booster. The driver sets the bus voltage while the booster augments

bus current when needed. In the driver booster method, microprocessor control coordinates the

action of both converters (“Current Sharing in Power Arrays”, n.d.). In general, there are several

disadvantages to using DC-to-DC converters in UAV APM systems. First and most important is

the weight of the DC-to-DC converter. As the power rating of a DC-to-DC converter increases, so

does its weight. For example, a commercially available 3 W DC-to-DC converter weighs around 2

g while a 400 W converter weighs around 220 g. When used for a low power load, such as

supplying a microprocessor circuit, the weight of the DC-to-DC converter has little impact on UAV

weight. However, when used to convert the output voltage of the fuel cell or battery (main power

path), then the DC-to-DC converter weight has a detrimental impact on overall UAV weight. Note

that using a DC-to-DC converter on the battery output is more costly in weight than using one on

the fuel cell output. The DC-to-DC converter power rating must match the maximum power

expected. With the battery supplying higher power than the fuel cell, the converter weight will

naturally be higher. The secondary disadvantages of DC-to-DC converters are cost, size,

reliability, and complexity. Third, DC-to-DC converters radiate electromagnetic interference (EMI)

which may interfere with the radio control receiver. Last, DC-to-DC converters are efficient in

converting power; typical power conversion efficiencies range from 82% to 95%, but they are not

perfect and still consume power.

45
 Fuel cells operate best under constant load conditions. Testing fuel cell performance

under pulsed load conditions reveals limitations. Consider a pulsed current load where the low

and high load currents are 10% and 90% of the fuel capacity respectively. At the low load current

level, excessive air flow across the cathode can cause the membrane to dry out. As a result, the

membrane ionic resistance increases. When the high current pulse occurs, the stack voltage now

dips lower due to increased internal resistance. Due to increased power dissipation across the

membrane, the membrane temperature rises (Jones, Mepsted, Moore, 1998). The result is a

repeating cycle of membrane dehydration and excessive membrane temperatures. Also, during

the high load current pulses, there is a limit to the change in rate for which the air can diffuse to

the cathode. In effect, the abrupt increase in current load causes oxidant starvation at the

cathode (Suh, 2006). Lower oxidant partial pressure due to starvation then drives the stack

voltage dip further down. Pulsed current operation tends to cause excess water build up at the

cathode, which in turn, tends to block oxidant diffusion (Suh, 2006). The result is a cycle of

oxidant starvation and excess water build up at the cathode. Purging of excess water is required

more often during pulsed current operation.

To protect the fuel cell against frequent transient loads, a battery can be combined with

the fuel cell to create a hybrid power system. The battery can be connected directly across the

fuel cell as in the PPM case, or it can be connected to the fuel cell through a DC-to-DC converter

as in the APM case. A study conducted in 1998 compared the pulsed load performance of a

PEMFC, a lead acid battery, and the two sources combined (Jones, Mepsted, Moore, 1998). The

fuel cell alone was able to drive the load, but excessive voltage drop of the fuel cell stack was

problematic and frequent purging of the fuel cell was necessary. The lead acid battery was able

to power the load, but it could only drive the load for 3 hours and 20 minutes. When the two

sources were connected in a PPM configuration, the load was driven for six days continuously

and no fuel cell purging was necessary. The key difference of the hybrid performance over the

fuel cell alone was that the battery charged during low load current intervals. The battery charge

current kept the fuel cell current above a minimal value which was enough to keep the membrane

hydrated and to prevent water buildup at the cathode (Jones, Mepsted, Moore, 1998).
46
 One difficulty with the PPM configuration is that the VI-curves of the battery and fuel cell

must be closely matched. If the fuel cell VI-curve is of a much higher voltage than the battery VI-

curve, then the fuel cell current capacity may limit the hybrid maximum current and the battery will

tend to overcharge. If the fuel cell VI-curve is of a much lower voltage than the battery VI-curve,

then the fuel cell will consume battery power and the PPM system will be ineffective. Using an

APM configuration allows the fuel cell VI-curve to differ from the battery VI-curve. Using a DC-to-

DC converter between the fuel cell and battery lifts the restriction of matching VI-curves and

delivers higher peak currents during pulsed load operation. A study performed in 2003, examined

APM hybrid power system performance under pulsed load conditions (Jiang, Blackwelder,

Dougal, 2003). Three control strategies were used to drive the DC-to-DC converter. First, the

maximum power strategy, operated the fuel cell at the maximum power point. Under the

maximum power strategy, a battery starting at a SOC around 70% would recharge and remain at

100% SOC. Second, the maximum efficiency strategy; the fuel cell operating point was set to

around 70% of maximum current. The battery SOC would slowly decrease until completely

discharged. Third, the adaptive strategy, was a combination of the first two strategies so that the

battery could be maintained within a SOC target range. When the battery SOC was below the

upper threshold, maximum power strategy was active which slowly recharged the battery. When

the battery SOC reached the upper threshold, maximum efficiency mode became active which

allowed the battery to slowly discharge. When the battery SOC reached the lower threshold

value, the DC-to-DC controller then switched back to maximum power mode. The maximum

power, maximum efficiency cycle would then repeat to keep the battery SOC between the upper

and lower threshold values (Jiang, Blackwelder, Dougal, 2003).

A variety of APM control algorithms are presented in literature such as the study

completed in 2007 by Jiang et al. (Jiang, Gao, Dougal, 2007) and the study completed in 2014 by

Lee and Park (Lee & Park, 2014). In the first paper by Jiang et al., the amount of fuel cell current

delivered to the bus is a piece-wise linear relationship that depends upon the battery voltage

which is used as a measure of the battery SOC. Similar to the 2003 study by Jiang et al. (Jiang,

Blackwelder, Dougal, 2003), the fuel cell current setpoint is set at either the maximum power
47
point or the maximum efficiency point. However, in the Jiang et al. 2007 study, instead of

switching between maximum power mode and maximum efficiency mode only, a new hybrid

mode for the DC-to-DC converter combines the two modes. Between the maximum and minimum

battery voltages, a linear combination of the two modes are used which connects the two

operating points by a line. In the 2014 study by Lee and Park, solar cells are added as a third

power source and are also connected to the electrical bus through a DC-to-DC converter. The

control algorithm uses battery SOC and load power to create 6 sectors of operation where each

sector defines a predetermined power split between the three power sources.

2.11 ASU APM Circuit

An APM electrical architecture that does not use DC-to-DC converters was developed at

ASU Polytechnic campus for the Composite UAV project (Monaco, 2013). Although the APM

system was only tested in the lab and not used in actual flight, it serves as an example of an APM

approach that is further developed in this research. In the ASU APM circuit, the fuel cell is directly

connected to the bus while the battery is connected to the bus through a switch. When the switch

is off, then only the fuel cell powers the load. When the switch is closed, then both battery and

fuel cell deliver power to the load. During the takeoff and initial climb phases of the flight profile,

the switch is closed; maximum power is delivered to the load. When the cruise flight phase is

reached, the switch is then opened and only the fuel cell powers the load. The salient feature of

this approach is that the high energy losses of the PPM method at low load currents is avoided

while both power sources are used at high load currents. Since the energy losses of a PPM

circuit are low when the load currents are high, the ASU APM circuit reaps the benefits of a PPM

architecture while avoiding the power contention issue at low load currents. The ASU APM circuit

complexity is low; only the switch and a method to control the switch is needed. The cost and

weight are low since only a few electronic components are needed. The benefits of an APM

approach are combined with the benefits of the PPM approach while avoiding the excessive

energy losses of PPM architecture at low load currents. The actual electronic components used

to construct the ASU APM circuit did not use a switch, but instead used a metal oxide field effect

transistor (MOSFET) to perform the function of the switch. The MOSFET was then controlled by
48
an Arduino microprocessor (Monaco, 2013). During takeoff and initial climb phases of the flight

profile, the microprocessor controls a bias network (a circuit consisting of resistors to operate the

MOSFET) so that the MOSFET is in its saturation region of operation with the controlling gate-to-

source voltage (Vgs) at its saturation value. This minimizes the drain-to-source resistance (rds) of

the MOSFET and is analogous to the switch being closed. Typical values of rds for a power

MOSFET in the on-state are in the 2 mΩ to 100 mΩ range (rds = rds(on)). When the cruise phase of

the flight profile is reached, the microprocessor then turns off the MOSFET by changing the logic

state of its controlling pin to the bias network. The MOSFET is then in an off-state where rds > 10

MΩ (rds = rds(off)); this is analogous to the switch being open. This example demonstrates that an

effective APM system can be constructed from a simple, low weight, low cost, electronic circuit.

During lab testing of the ASU APM circuit, the MOSFET transistor burned up and the

circuit was abandoned in favor of another approach using a relays. The ASU Relay circuit

performed as expected and was used for further testing. The failure of the MOSFET was

attributed to the current rating. Under test conditions, the battery current was measured at 50 A

while the MOSFET was rated for only 30 A (Monaco, 2013). Even though the MOSFET had a

current rating that was too small for the actual current conditions of the circuit, a MOSFET with a

current rating of 50 A or larger may have failed as well. The reason is that proper heat sinking is

also required to operate a power MOSFET under high current conditions. In other words, if a 60 A

power MOSFET was used instead, but had no heat sink, a 50 A current load would have caused

the MOSFET to fail anyway. In typical power MOSFET data sheets, maximum junction

temperature is listed (usually around 150oC) along with thermal resistance ratings such as

junction-to-ambient and junction-to-case. Circuit design must ensure that the MOSFET junction

temperature is kept below the maximum allowable value. Although the ASU APM circuit did not

succeed for its intended purpose, there are several features of the circuit that suggest a different

approach to an APM design that is discussed further in section 4.1. Shown below in figure 15 is a

conceptual diagram of the original ASU APM circuit.

49
Figure 15. ASU APM Circuit (Monaco, 2013)

In figure 15, the ASU APM circuit shows the microprocessor controlling the state of Q1

which connects or disconnects the battery from the bus. The points A and B represent the

microprocessor monitoring the fuel cell and battery voltage levels. Diodes D1 and D2 prevent any

reverse current flows between the power sources. The original design intent was to either fully

connect or fully disconnect the battery from the bus. In this scenario, the diodes are necessary

since the battery is effectively connected directly across the fuel cell. When Q1 is in the on-state,

the circuit is similar to the PPM case where power contention may occur between the sources.

The diodes prevent power exchange between the sources. There is a drawback to the diodes

though, they drop a forward bias voltage and thus dissipate power during normal operation. The

power dissipation of the diodes are significant since they are in the main power path.

50
 CHAPTER 3

SIMULATION COMPONENTS

3.1 Simscape and Simulink

In this study, system simulations are performed in the MatlabTM environment from

Mathworks, Inc. Within the Matlab SimulinkTM environment, SimscapeTM physical modeling is

used to develop power system components which are then assembled into a hybrid power

system model. Simscape is designed to model physical systems using a Physical Network

Approach (PNA) (“Simscape User’s Guide”, 2014). Simscape blocks are placed in a Simulink

model to represent a system component, however Simscape blocks are fundamentally different

from Simulink blocks. Simulink blocks represent mathematical operations while Simscape blocks

represent physical components. Assembling Simulink blocks into a model results in a graphical

representation of a mathematical model. Connections between Simulink blocks carry numerical

information called signals and define an order of mathematical operations defined within Simulink

blocks. Simulink blocks are considered causal since for a given input, an appropriate output is

developed. Simulink signals always have a direction associated with them and do not carry units.

On the other hand, assembling Simscape blocks results in a graphical model of a physical

system. Connections between Simscape blocks represent physical connections between physical

components and do not define any order of mathematical operations. Simscape blocks have two

types of ports, physical ports and physical signal ports. Physical ports carry a flow of energy in or

out of a block and do not have a direction. When physical ports on two Simscape blocks are

connected, energy can flow between the two components in either direction. Simscape models

are considered non-causal since the blocks have no input or output; models consist of energy

flow between blocks. Connections between physical ports always carry units. Physical signal

ports carry information about a physical value (with units) and do have a direction assigned.

Physical signal ports are not part of the energy flow network of a physical model; instead, they

can be used to monitor internal block variables or carry information between blocks.

Within a Simulink model, the canvas is the window working and graphical area where

blocks are placed, connections are made, and the graphical presentation of the simulation model
51
is displayed. On the canvas, both Simulink blocks and Simscape blocks are placed, but they are

still functionally separated. Simulink blocks inherently work inside the Simulink canvas without

any auxiliary helper blocks. Simscape blocks require a solver block to be connected to the

reference node of the Simscape model. One great feature of the Simulink environment is that

Simulink and Simscape models may be connected together on the same canvas. Observer and

translation blocks are used to read and convert a Simscape physical value to the Simulink model

domain. A second great feature of the Simulink environment is the ability to use Matlab scripting

files to automate simulations. Since Simscape models are contained within the Simulink canvas,

both Simulink and Simscape models, or combinations of the two, can be automated through

Matlab scripts. The result is that a physical model may be simulated for a range of parameter

values and the results from each simulation run can be extracted, grouped, and graphed to reveal

how a parameter affects system performance.

The Simscape physical modeling method allows for the creation of custom blocks

through the use of the Simscape text-based language. A custom system component can be

described by a text file and then compiled into a Simscape library. Once compiled, the custom

component has a graphical block inside the library that may be dragged onto the Simulink canvas

like any other block. Although the Simscape language is similar to the Matlab language, it is

different and has a required file structure to create a Simscape component. A Simscape

component file has three main sections: a declaration section, a setup section, and an equation

section. In the declaration section, through and across variables, physical ports, physical signal

ports, and parameters are defined. In the setup section, the relationships between the through

and across variables are defined. The setup section also defines initial conditions and provides a

location for parameter checking functions. The third section, the equation section, defines the

mathematical equations of the component (“Simscape Language Guide”, 2014).

3.2 Lithium Ion Battery Component Model

For a hybrid power system simulation, the battery model employed must strike a balance

between fidelity and computational efficiency. A good lithium ion battery model for the system

level simulation purposes of this study is presented by Olivier Tremblay and Louis-A. Dessaint
52
(Tremblay & Dessaint, 2009). The model equations take battery capacity, SOC, and current

delivery history into account. In this model, discharge and charge equations are separate. The

equation for the discharge case (i > 0) is listed below in equation 24.

 
Vbat  Eo  i  R  A  exp B   i (t )dt  Kc 
Q
Q   i (t )dt

  i (t )dt  i *  (24)

The equation for the charge case (i < 0) is listed below in equation 25.

 
Vbat  Eo  i  R  A  exp B   i (t )dt  Kc 
Q
Q   i (t )dt
  i (t )dt  KR 
Q
i* (25)
 i (t )dt  0.1 Q
where: Vbat = battery voltage (V)

Eo = open circuit voltage of battery (V)

Q = battery capacity (Ah)

i = battery current (A)

i(t)dt = charge transfer to or from battery (Ah)

i* = filtered battery current (A)

R = internal resistance of battery (Ω)

A = exponential zone amplitude (volts)

B = exponential zone time constant inverse (amp-hours-1)

Kc = polarization constant (volts/(amp-hour))

KR = polarization resistance (Ω)

The Tremblay battery model is based from earlier work by Shepherd (Shepherd, 1965). The key

difference is the use of filtered current i*. The filtered current is the instantaneous battery current

after it has been passed through a low pass filter. When implementing the Tremblay model in

Simulink, the filtered current prevents an algebraic loop (Tremblay & Dessaint, 2009). In this

study, the Tremblay model is implemented using the Simscape language and the algebraic loop

is no longer a problem. In this Simscape model, battery current can be used in place of the

53
filtered current. The Simscape code to implement the lithium ion battery model is listed in

appendix A.

As a first test of the battery model, a commercially available LiPo 4S 8 Ah, battery is used

as an example. To plot battery discharge curves for a range of discharge rates, a Simscape

model is developed which loads the battery at a regulated current level. The load circuit used is a

constant current regulator circuit with a programmable current setpoint (figure 16).

Figure 16. LiPo 4S 8 Ah Battery Discharge Circuit

Transistor Q1 (p-channel MOSFET) is the regulating element which determines the load current

drawn from the battery under test. Resistor R2 functions as a current sense resistor and is fixed

at a low resistance value of 0.1 Ω. Operational Amplifier (OpAmp) U1 measures the differential

voltage across R2 and amplifies this voltage by a factor of 20. OpAmp U2 then takes the

difference between the current setpoint voltage and U1 output to drive the gate voltage of Q2.

Transistor Q2 (n-channel MOSFET) in conjunction with resistor R1 then control the gate voltage

of Q1. As the battery voltage changes, the current drawn from the battery is held at a fixed level

set by a current setpoint voltage (Vcsp).

Vcsp  1.7  2  i Battery (26)

The current setpoint voltage is a linear relation to the resulting load current on the battery. The

first term is a 1.7 V offset which sets the quiescent operating point of the gate voltage of Q2. The

second term relates the desired load current on the battery to the Q2 gate voltage rise above the

54
quiescent value. The end result is a programmable constant current load circuit that can be used

to test the battery.

The battery model parameters are set to represent a typical LiPo battery pack used by

model airplane enthusiasts. It represents a state-of-the-art LiPo battery that is commercially

available (“MaxAmps LiPo 4S 8Ah”, 2016). This particular battery is in the midrange of the

product family and is a good compromise between capacity and weight. The size, weight, and

electrical capacity are well matched to the ASU Composite UAV and will be used as the reference

battery in the power system simulations of chapter 4. With a nominal voltage of 14.8 volts and

battery pack weight of 0.764 kg, the specific energy density of this 8 Ah battery pack works out to

be 155 Wh/kg. A nice feature of this battery pack is the high discharge rating. Using two 4S, 4 Ah

lithium ion batteries in parallel, the maximum discharge rating is reported to be 150 C. Operating

at half the maximum discharge rate (75 C), the battery pack can deliver 600 A of current for 48

seconds yielding a specific power density of 11,623 W/kg. The Tremblay battery model

parameters used to represent the MaxAmps 4S 8Ah battery pack are given in table 3.

Table 3

LiPo 4S 8Ah Battery Model Parameter Values

Parameter Name Symbol Value Units
o
Open Circuit Voltage E 15.6 V
Battery Capacity Q 8 Ah
Internal Series Resistance R 0.005 Ω
Initial State of Charge 100 %
Polarization Constant KC 0.0075 V/(Ah)
Polarization Resistance KR 0.0064 Ω
Exponential Zone Amplitude A 1.20 V
Exponential Zone time Constant B 2.00 1/(Ah)

The discharge curves for the LiPo 4S battery are obtained by Simscape simulations of the

discharge circuit (figure 16) over a range of load currents. For each simulation run, battery

voltage and delivered charge data is extracted. The data is then combined and graphed together

to yield the plot in figure 17. The Matlab automation script is listed in appendix B.

55
Figure 17. LiPo 4S 8 Ah Battery Discharge Curves

To compare the LiPo 4S battery model against a Simscape battery model from the Sim Power

Systems library, a second discharge simulation is created. It compares the battery voltages over

a 2 hour period when both are supplying a 4 A load (0.5 C discharge rate). The Simulink model is

listed in appendix C and the graph of battery voltages versus time is shown below in figure 18.

Figure 18. Battery Discharge Comparison at 0.5C Discharge Rate

56
3.3 Battery Charger Model

To recharge the LiPo 4S battery during flight, an onboard battery charger circuit is

included with the APM system. A standard lithium ion battery recharge cycle consists of a

constant current phase followed by a constant voltage phase (Tsang, Sun, Chan, 2010). The

constant current phase is used during most of the charge cycle, and then the constant voltage

phase is used at the end of the cycle to top off the battery charge. To simplify the charger circuit,

only the constant current phase is implemented. The battery charger circuit (figure 19) uses a

small DC-to-DC converter to generate a 24 V source and then uses a programmable constant

current source to control the battery recharge rate.

Figure 19. Battery Recharge Circuit

Input power from the electrical bus is fed to the DC-to-DC converter which produces a constant

24 V source at resistor R2. Transistor Q1 (p-channel MOSFET) acts as a series regulator to

control the current flow to the battery. Similar to the constant current load circuit described in

section 3.2, the difference between the setpoint voltage and the sense resistor voltage is sent to

the gate of Q2 (n-channel MOSFET) which in turn, controls the gate voltage of Q1. The key

difference here is the resistor R3, which drops the proper amount of voltage for a specified

recharge rate. The value of R3 will depend on the range of recharge rates desired. For this case,

the recharge circuit is designed to operate near a 0.1 C recharge rate which sets the value of R3

57
to 8.25 Ω. A Simscape model is created to test the recharge circuit on the LiPo 4S battery (figure

20). Note that the recharge rate will determine the size of the DC-to-DC converter necessary. For

this circuit, the 24V output of the converter delivers 0.8 A which allows for a 20 W converter to be

used. Minimizing the DC-to-DC converter power rating keeps the recharge circuit weight to a

minimum.

Figure 20. Battery Recharge Simulation

The simulation starts out with the battery at a 40% battery SOC and is run for 5.5 hours at a

charge rate of 0.1 C (0.8 A). At the end of the charge cycle, the battery is at 96% SOC (figure 21).

Figure 21. Battery SOC during Recharge

The corresponding battery voltage and recharge current are shown in figures 22 and 23

respectively.

58
Figure 22. Battery Voltage during Recharge

Figure 23. Battery Current during Recharge

3.4 PEMFC Component Model

The PEMFC has a variety of mathematical models to choose from. The type of simulation

will determine which model is appropriate. For purposes of this study, an electrical circuit model

has a good balance between model fidelity and computational cost. The electric circuit equivalent

model for a PEMFC consists of four sections; an ideal voltage source and three series resistors.

In thermodynamic lingo, the ideal voltage source represents the reversible Nernst voltage while

the three resistors represent irreversible losses. In electrochemical language, the three sources of

energy loss are called polarizations and consist of the activation loss, the ohmic loss and the

concentration loss. To improve the fuel cell model, the effect of the electrochemical double layer

can be included. A capacitor added across the activation and concentration resistors causes a
59
delay between fuel cell voltage and current. Shown below in figure 24, part (a) shows the electric

circuit equivalent model of a PEMFC without the electric double layer capacitor. The fuel cell

voltage response to a step increase in current has no delay. An improved model includes the

double layer capacitor as shown in figure 24 part (b).

Figure 24. Electrical Circuit Equivalent Model of PEMFC

Note that the electric circuit may represent a single cell or the entire fuel cell stack. The values

assigned to the electrical components are either for a single cell or multiplied by the number of

cells in the stack. In standard electrical circuit analysis, resistors are assumed to be fixed in value.

In the PEMFC equivalent circuit, the resistor values are current dependent which makes the

circuit model non-linear.

The Simscape component for the PEMFC electrical circuit model has several ways it can

be implemented. The entire circuit can be implemented at a single level (flat design) using

standard Simscape blocks connected graphically, or it can be implemented in a single text file.

The difficulty with this approach is that a flat design results in a very complicated file, graphically

or text-based. To break the problem down into simpler pieces, each electrical component within

the circuit can have its own file. The sub-component files can then be assembled into the

electrical circuit (hierarchical design). The top level file connects the sub-components and can be

60
graphical or text-based. In this study, the PEMFC Simscape component implements the electrical

circuit model by a hierarchical approach using text-based files for the sub-components and the

top level.

In recent literature, alternative representations of the electrical circuit model of a PEMFC

have been reported. Buasri and Salameh suggest an electrical model similar to figure 24 (b), but

add a series diode between Ract and Rcon (Buasri & Salameh, 2006). The exponential region of

the diode forward conduction mode is similar to the exponential decay of the fuel cell polarization

curve in the activation region (see figure 12). In the Buasri model, expressions for Vact and

Vohmic are unchanged from the standard expressions given in equations 16 and 17 respectively.

However, the Buasri model suggests a new expression for  V con which is quite different from the

standard equation given by equation 18. In equation 27 shown below, the Buasri concentration

loss expression has two parameters, K1 and K2 which are determined empirically (Buasri &

Salameh, 2006).

 V con  K 1  i FC  e K 2  i FC (27)

A second approach to the PEMFC electrical circuit model uses the same circuit as shown

in figure 24 (b), except the expressions for Eo, Vact , and Vohmic have been modified from the

standard expressions given by equations 15, 16, and 17 respectively (Lee & Wang, 2007,

Qingshan et al., 2008). In this model, referred to as the Lee model, the expression for  V con

remains unchanged from equation 18. In this study, the Lee model is compared to the standard

expressions given in section 2.7 and is used as the PEMFC electrical model for the hybrid power

system simulations of chapter 4. The Simscape PEMFC component developed in this paper is an

implementation of the Lee PEMFC model (Lee & Wang, 2007).

To calculate the Nernst potential using the standard method, the Nernst equation is used

in conjunction with a Gibb’s free energy table.

o
 g f
Nernst equation: E o
s tan dard 
2F

RT
2F

 ln PH 2 
RT
4F

 ln PO2   (28)

61
Table 4

Gibb’s Free Energy for Hydrogen Reaction (Madakannan lecture notes, 2016)
Form of Water Temperature
Product (C) g (kJ/mol)
liquid 25 -237.2
liquid 80 -228.2
gas 80 -226.1
gas 100 -225.2
gas 200 -220.4

In the Lee model, the Nernst potential is approximated by equation 29 (Lee & Wang, 2007). The

Simscape sub-component code for the Nernst potential is listed in appendix D.

o
ELee
    
 1.229  0.85  103  T  298.15   4.3085  105  T  ln PH2  0.5  ln PO2 
 (29)

where: T = fuel cell temperature (K)

P H 2  partial pressure of hydrogen (atm)

PO 2  partial pressure of oxygen (atm)

In the Lee model, the approximation for the Nernst potential does not require a value for the

Gibb’s free energy, only temperature and gas pressures are needed. The percentage error using

the Lee Nernst potential expression over temperature and hydrogen pressure can be illustrated

by a surface plot of the percentage difference (Matlab code listed in appendix E).

 E0  ELee
o

Eo   s tandard
o   100% (30)
 Es tandard 

62
Figure 25. Percent Error of Lee Model Approximation of Nernst Potential

From figure 25, the Lee model is within +4% and - 3% of the standard Nernst potential

calculations. The percent error is predominantly dependent upon hydrogen pressure and not

temperature. The green grid is the zero-error plane which shows minimum error when the

hydrogen pressure is near 2 atmospheres. In figure 26, a different view of the same data is

helpful to determine the percent error as a function of hydrogen pressure. Each line is the percent

error at a different hydrogen pressure.

63
Figure 26. Percent Error of Lee Model for Different Hydrogen Pressures

To calculate the activation polarization using standard methods, a value for the charge

transfer coefficient   and a value for the exchange current density  i o  are required.

RT i 
 Vact s tandard   ln FC  (31)
2    F  io 

The expression for the activation loss used in the Lee model is shown below in equation 32 and

only requires values for temperature and air pressure (Lee & Wang, 2007). Note that PO is the 2

partial pressure of oxygen in air, which is 20.95%.

 Vact Lee   
  1   2  T  3  T  ln CO2   4  T  ln  i FC  
 (32)

where: 1  1.81 (empirical constant in units of V)

 2  0.00286 (empirical constant in units of V/K)

 3  7.6  10 5 (empirical constant in units of V/K)

 4   0.78  10  4 (empirical constant in units of V/K)

64
 PO2
CO2  (concentration of oxygen at catalytic interface)
 498 
 
6
5.08  10 e  T 

It is difficult to compare the activation loss approximation used in the Lee model to the standard

method since  and io are not known for a given set of empirical parameters 1   4 . In

addition, the CO2 parameter in the Lee model represents the concentration of oxygen at the

cathode catalytic layer. To account for cathode GDL properties, only parameter  3 can be

adjusted which may not yield a good representation over all temperatures and air pressures. The

Simscape sub-component code for the activation resistor is listed in appendix F.

The ohmic loss expression used in the Lee model is essentially the same as the standard

method described by equation 17. The Lee model goes into more detail in how to calculate the

internal resistance of the cell. First, the cell resistance is divided into electronic resistance and

ionic resistance components as shown below in equation 33 (Lee & Wang, 2007).

 V ohm ic  i   R m  R c  (33)

where: R m  ionic resistance of membrane (  )

Rc  electronic resistance of cell (  )

i = fuel cell current (A)

Second, the ionic resistance of the Nafion membrane (Rm) is expressed in terms of membrane

specific resistivity,  m .

m  tm
Rm  (34)
A

where  m  specific resistivity (  cm )

t m = membrane thickness (cm)

A = cell active area (cm2)

Third, the specific resistivity of the Nafion membrane is given by an empirical expression

(equation 35).

65
  i  T   i FC  
2 2.5

181.6  1  0.03  FC  0.062       

 A  303   A  
m  (35)
  i FC     T  303  
  0.634  3   A    exp  4.18   T 
     

where:  adjustable parameter (between 14 and 23)

A = cell active area (cm2)

i FC = fuel cell current (A)

Fourth, in the Lee model, there is no value given for Rc (Lee & Wang, 2007). The electronic

resistance of the cell is most likely measured empirically. To compare the ohmic loss between the

Lee model and the standard method is ambiguous. Since the standard method only states a

constant value for the cell resistance, it is only an approximation to the true cell resistance over

the fuel cell operating conditions.

Figure 27. MEA Resistance versus Current Density for Different Relative Humidity Values

(Madakannan lecture notes, 2016)

In figure 27, the cell resistance is plotted against current density for a range of relative

humidity values of the anode gas. Since relative humidity does not affect the electronic resistance

of the cell, it is the ionic cell resistance that is affected. In other words, specific resistance of the
66
Nafion membrane is affected by the stoichiometry and relative humidity of the anode gas.

Although the hydration state of the Nafion membrane depends upon fuel cell design, figure 27

illustrates that the cell resistance is not always constant over a range over current densities. For a

relative humidity of 25%, cell resistance is inversely proportional to current densities under 0.7

A/cm2. In this case, a constant resistance is not a good approximation to cell resistance over the

full range of current densities. On the other hand, at 100% relative humidity, cell resistance is

indeed constant across the full range of current densities. Figure 27 shows a worst case increase

in cell resistance of 60% at 0.3 A/cm2 depending upon relative humidity. The point is that Nafion

membrane specific resistance is affected by stoichiometry and relative humidity of the anode gas.

The Lee model for cell resistance uses the  parameter in the specific resistivity expression

(equation 35) to account for stoichiometry and relative humidity of the anode gas. The parameter

 can vary between 14 and 23 and is determined by membrane preparation, relative humidity,

and stoichiometry of the anode gas (Lee & Wang, 2007). To visualize the cell resistance as a

function of temperature and current, the following scenario is calculated in Matlab. Assume a cell

has an electronic resistance of 20 m Ω, an active area of 20 cm2, and a membrane thickness of

51 um. The cell operates over a temperature range of 40oC to 90oC and a range of current

densities from 0.1 A/cm2 to 1 A/cm2. For each integer value of  between 14 and23, create a

surface plot. The Matlab code for this cell resistance example is listed in appendix G and the

resulting plot is shown below in figure 28.

67
Figure 28. Cell Resistance versus Current Density and Temperature (14 <  < 23)

In figure 28, the top surface plot is for  = 14 and the surface plots proceed downwards as 

increases. The bottom surface plot corresponds to  = 23. The surface plots show a cell

resistance dependence upon both current density and temperature (Lee model). The ratio of

highest to lowest cell resistance is about 1.16 to 1. In other words, depending upon the value of

 , temperature, and the current density, the cell resistance can increase by 16% in the Lee

model. The Simscape sub-component code for the ohmic resistor is listed in appendix H.

To complete the PEMFC Simscape model, the subcomponents are assembled in a top

level file. However, before the final PEMFC model can be assembled, there are three more sub-

components. First, the concentration loss of the fuel cell is represented in the electrical circuit as

the concentration resistor sub-component. This Simscape sub-component is simply a current

dependent resistor that implements the concentration loss expressed by equation 18. The

Simscape sub-component code for the concentration resistor is listed in appendix I. Second, the

double layer capacitor represents the internal capacitance of the fuel cell due to the

electrochemical double layer. This Simscape sub-component is an implementation of a fixed

capacitor and the code is listed in appendix J. Third, the flow rate sub-component calculates

68
hydrogen and air flow from fuel cell current. The flow rate sub-component code is listed in

appendix K. The PEMFC model is assembled in the top level file listed in appendix L. The

assembled fuel cell model is then tested in the following Simscape simulation (figure 29).

Figure 29. PEMFC Test Simulation

In the test simulation shown above in figure 29, the PEMFC has a load that increases over time.

The stack voltage and fuel cell current data are extracted which results in the following

polarization curve and power curve plots shown below in figure 30.

Figure 30. PEMFC Model Polarization and Power Curves

The parameters for the PEMFC model are set to mimic the Horizon H-100 fuel cell. On the

polarization curve, two points to check are the open circuit voltage and the nominal operating

point. From the Horizon H-100 fuel cell stack user’s manual, the open circuit voltage should be 19

V and the nominal operating point should be 12 V at 8.3 A (Horizon, 2013). This is in agreement

69
with the PEMFC model polarization curve. On the PEMFC power curve, the maximum power

should be 100 W at the nominal operating point at 8.3 A (Horizon, 2013). In accordance with H-

100 operation, the hydrogen pressure is set to 0.5 atm. and the air pressure is left at 1 atm. The

PEMFC model parameters are shown below in figure 31.

Figure 31. PEMFC Model Parameters

Note that the number of cells is the same as the Horizon H-100 fuel cell stack (20 cells). Last, the

dynamic behavior of the PEMFC model is tested for a range of double layer capacitance values.

The results for the double layer capacitor ranging from 1 uF to 1 F are shown below in figure 32.

Figure 32. PEMFC Dynamic Response for Range of Double Layer Capacitance
70
3.5 Balance of Plant Model

To properly model the Horizon H-100 fuel cell, the BOP electrical model must be

included. Within the BOP equipment that ships with the fuel cell, only the fuel cell controller, fan,

and temperature sensor consume electrical power. The fuel cell controller is powered by an

external 13 V power supply and it then supplies power to the fan and temperature sensor. The

electrical model of the BOP equipment is shown below in figure 33.

Figure 33. BOP Circuit

Input power for the BOP circuit comes from the PMS electrical bus. The DC-to-DC converter then

produces a regulated 13 V to run the fuel cell controller. The variable resistor labeled as

BOP_Load acts as the fuel cell controller electrical load. A control signal adjusts the BOP_Load

as a function of time and the value of the control signal is the BOP_Load resistance.

Figure 34. BOP_Load Control Signal

During operation, the fuel cell controller consumes a somewhat constant amount of power. The

change in BOP_Load over time is due to the fan being on or off. In figure 34, the control signal

simulates the BOP_Load when the fan is normally off and then turns on for a short duration. Note

that the BOP_Load control signal is set to 18 Ω when the fan is off and is set to 11 Ω when the

71
fan is on. This will correspond to a BOP electrical load of roughly 0.76 A when the fan is off and

1.24 A when the fan is on.

3.6 Motor Propeller Model

Most of the electrical load on the power system consists of the ESC. The ESC delivers

power to the BDCM which in turn, spins the propeller. The propeller then converts the rotational

motion of the BDCM shaft into linear thrust. To size the electrical load, the physical properties of

the aircraft and the design goals are required. In this study, the ASU Composite UAV is the

aircraft used for the simulations. One design goal for the power system is to power the aircraft

using only the fuel cell during cruise. It is given that the power required during cruise flight

conditions is 25% of maximum and that the thrust to weight ratio of the aircraft is between 0.2 and

0.4 (Raymer, 1989). Converting the aircraft weight of 5.5 kg to an equivalent gravitational

downward force gives a value of 53.9 Newtons (N) which then leads to the following thrust

constraint formula (equation 36).

10.8  N  T  21.6  N (36)

The ASU Composite UAV is designed to have a cruising speed of 20 m/s (Krauch et al., 2012).

Using the definition of power (power = force x velocity), the thrust constraint equation can be

expressed in terms of power. The required power during cruising conditions (cruise power) is

25% of maximum power (Pmax).

54  W  PThrust ,cruise  108  W (37)

216  W  PThrust,m ax  432  W

(38)

Equations 37 and 38 express the amount of thrust power necessary to yield a thrust to weight ratio

between 0.2 and 0.4. To convert from thrust power to the rotational mechanical power at the BDCM

shaft, the propeller efficiency can be used. During this study, different propellers for model aircraft

applications have been examined, however propeller efficiency is not a specification that is reported

in the sources reviewed by this author. Instead, a reasonable estimate for a propeller efficiency

value is 0.85 (Sadraey, 2013).

72
 63.5  W  PMech ,cruise  127  W
(39)

254  W  PMech ,m ax  508  W

(40)

To then convert from mechanical power to electrical power, the BDCM efficiency can be used. At

this point, the BDCM must be selected in order to obtain an accurate efficiency value. In the original

ASU Composite UAV design, three BDCMs were used. All three motors used were from the

manufacturer AXI (Czech Republic); AXI product number 2217/16. In this study, the three BDCM

are replaced by one larger BDCM, the AXI 4120/14. From the AXI application notes, the AXI

4120/14 combined with a 13 x 8 propeller has an efficiency of 84% (“AXI Motor Applications

4120/14”, 2015).

75.6  W  PElec ,cruise  151  W

(41)

302  W  PElec ,m ax  605  W

(42)

As a result of equation 41, a fuel cell power of 100 W should be sufficient to meet the design goal

of powering the UAV during cruise flight using only the fuel cell. During takeoff and climb portions

of flight, the maximum battery power required will then be in the following range (equation 43).

202  W  PBattery ,m ax  505  W

(43)

The motor-propeller subsystem (MPSS) is modeled in Simscape as a graphical block using

both Simscape library components and custom Simscape text-based components. The inputs to

the MPSS are power in from the electrical bus and a throttle command that ranges from 0% to

100%. The outputs from the MPSS are the propeller speed setpoint, actual propeller speed,

electrical power consumed, and mechanical power produced at the BDCM shaft. Central to the

operation of the MPSS is the Simscape servomotor component from the SimElectronicsTM add-on

library. The servomotor component represents a BDCM and its controller (the ESC) operating in a

closed loop torque control mode. An outer control loop based on propeller speed is then placed

around the servomotor. A torque load is then applied to the shaft of the servomotor to represent

the aerodynamic load of the propeller. The propeller speed and torque load are adjusted to model
73
the AXI 4120/14 BDCM with a 13 x 8 APC propeller. The MPSS block diagram is shown below in

figure 35.

Figure 35. MPSS Block Diagram

Using the figures given in the AXI application notes, the AXI 4120/14 motor combined with a 13 x

8 APC propeller has the following specifications listed below in table 5.

Table 5

AXI 4120/14 BDCM with 13 x 8 APC Propeller Specifications

Specification Value
Input Voltage (V) 12.9
Maximum Propeller Speed (RPM) 7300
Maximum Electrical Power Input (W) 381
Maximum Mechanical Power Output (W) 323
Efficiency (%) 84

In the MPSS block diagram (figure 35), the throttle input is converted to a power request value

which is then converted to an equivalent propeller speed value. The propeller speed value is then

used as the setpoint to the outer control loop. The outer control loop generates an error signal

from the difference of the setpoint and the servomotor shaft speed and then feeds that error

signal into a PI controller. The output of the PI controller is then converted to a torque value and

fed to the servomotor torque request input. The output of the servomotor is the shaft speed which

is then used as feedback for the outer control loop and also as an input for a load-torque lookup

74
table. The output of the load-torque lookup table is fed to the input of a custom Simscape torque

load component which applies the torque load to the shaft of the servomotor. The Simscape code

for the revolutions per minute (RPM) versus power lookup table, the torque versus RPM lookup

table, and the torque load component are listed in appendix N.

Figure 36. Propeller Speed and Motor Power versus Throttle

In figure 36, the propeller actual speed matches the setpoint as the throttle ranges from

0% to 100% which demonstrates that the outer control loop is functioning as expected. The curve

of the propeller speed versus throttle graph is determined by the RPM versus power lookup table.

Data for the lookup table was derived from a combination of measured data and the motor

manufacturer’s application note (“AXI Motor Applications 4120/14”, 2015). First, the maximum

point on the propeller speed versus throttle graph is established from the AXI application note,

(100%, 7300 RPM). Second, a test fixture was constructed to measure the propeller speed and

power versus throttle curves of the AXI 1420/14 motor. Images of the thrust test bench (TTB) are

shown below in figure 37. Experimental data measuring ESC waveforms, thrust, motor power,

ESC current, ESC voltage, and propeller RPM were collected during the fall 2015 semester. A

portion of this data, considered as the highlights, is listed in appendix X.

75
Figure 37. Thrust Test Bench

In the TTB setup, the ESC driving the AXI 4120/14 motor was the JETI Advance 70 Pro electronic

speed controller. Two limitations were present during testing. First, the propeller used in the TTB

was different than the 13 x 8 APC propeller modeled in the MPSS block. Instead, the TTB used a

10 x 7 APC prop. Second, the power sources used to drive the ESC/motor combination were

limited in power capacity and could not drive the motor to its full potential. The end result of motor

testing with the TTB was a characterization of the propeller speed versus throttle curve and the

power versus throttle curve over a limited range. Shown below in figure 38 are two graphs of

measured data collected with the TTB.

Figure 38. Thrust Test Bench Data

76
The curve of the propeller speed versus throttle graph in figure 38 is brought over to the propeller

speed versus throttle graph in figure 36 with the curve scaled to fit the maximum point, (100%,

7300 RPM). Similarly, the power versus throttle graph shows a linear relationship which is

brought over to the mechanical power curve of figure 36. The mechanical power curve of figure

36 is determined by the lookup table inside the TorquevsRPM block of the MPSS model. Note

that in the power versus throttle curve of figure 36, the electrical and mechanical curves are

nearly identical for values under 110 W. Similar to the propeller speed versus throttle graph of

figure 36, the maximum electrical and mechanical power points in the power versus throttle graph

are set to the values given in the AXI application note (“AXI Motor Applications 4120/14”, 2015).

The curves shown in figure 36 result from a simulation using the MPSS block while the curves

presented in figure 38 are measured data from the TTB. Although the characteristics between

figures are closely matched, they are not perfectly matched. The interplay between

subcomponents within the MPSS block make a perfect match difficult and therefore MPSS

parameters were tuned for a best fit.

77
 CHAPTER 4

SYSTEM INTEGRATION

4.1 The Current Assist and Observer Circuit

In this research, a unique approach to APM system design for UAV applications is

presented. Building upon the concept of the ASU APM circuit, a new APM system design is put

forth that uses a MOSFET approach to control battery current delivered to the electrical bus. The

proposed APM system design has the fuel cell connected directly to the bus while the battery is

connected to the bus through a p-channel enhancement MOSFET. This is similar to the ASU

APM circuit, but in this new circuit design, the MOSFET is used in a linear fashion to control

battery current. Additionally, the fuel cell voltage and current and the battery voltage and current

are monitored. With this information, a supervisory microprocessor can run a digital observer

algorithm to track hydrogen consumption, fuel cell efficiency, battery SOC, and power delivered

from both sources. Due to the circuit features of battery augmented current delivery to the bus

and power source monitoring, the proposed APM circuit is named Current Assist and Observer

Circuit (CAOC). In this APM system design, no DC-to-DC converter is used in the main power

path, thus reducing the APM weight to a minimum. Also, no series diodes are used at the power

sources as in the ASU APM circuit, so no energy is lost to the forward voltage drop of the series

diodes. During the cruise phase of flight, in steady-state operation, the power loss of the CAOC

drops to near zero yielding a better power efficiency than some DC-to-DC converter approaches

(compared to figure 14). Figure 39 shown below displays a conceptual diagram of the CAOC.

78
Figure 39. CAOC Concept

Resistors R1 and R2 are current sense resistors with extremely low resistance values of 5 mΩ.

Due their extremely low resistance, the voltage drop and power dissipated by R1 and R2 is small,

almost negligible. Operational amplifiers (OpAmps) U1 and U2 measure the current from the fuel

cell and battery respectively. They take the differential voltage drop across R1 and R2 and

amplify the voltage (e.g. voltage gain factor around 20). The OpAmps convert the measured

differential voltage to a single-ended voltage which can then be sent to the observer. The output

voltage of U1 is also sent to the inverting input of U3. OpAmp U3 takes the difference between a

setpoint voltage and the output of U1, amplifies the difference, and controls the gate voltage of

Q1. MOSFET Q1, normally in the off-state due to resistor R3, begins to conduct only when the

gate voltage is driven lower than the source voltage minus the MOSFET threshold voltage

(Vthreshold). The setpoint voltage is generated by the supervisory microprocessor and is a voltage

command for the fuel cell current setpoint. When the actual fuel cell current is below the setpoint,

Q1 remains in the off-state. When the fuel cell current increases past the setpoint value, U3

drives the gate voltage of Q1 to a lower voltage, thus forcing Q1 to conduct slightly. The amount

that Q1 conducts is then proportional to the difference between the actual fuel cell current and the

setpoint value. The end result is that the fuel cell current is held at a constant value fixed by the

79
setpoint voltage. If the bus current goes above the fuel cell current setpoint, extra current is

supplied by the battery. If the bus current is below the setpoint, then the fuel cell current will

decrease to the bus current and the battery remains disconnected from the bus. Last, the fuel cell

voltage and battery voltage are monitored and sent to the observer. Note that the MOSFET Q1

has to handle the maximum current delivered by the battery. Device selection for Q1 must have a

continuous current specification larger than the maximum battery current plus a 50% safety

factor. In addition, special consideration must be given to the heat sinking of Q1 so that the

junction temperature stays below the maximum allowed value. A side benefit of the CAOC

method is that the setpoint may be lower than the bus current during cruise conditions. In this

case, the battery will contribute part of the cruise mode current. In this manner, the battery can

contribute to maximum flight endurance and is a new option for mission planners.

4.2 Hybrid Power Management Controller

To implement the hybrid power management controller (HPMC) as a Simscape model,

the CAOC is combined with a microprocessor. The circuit components used for the CAOC are

pulled from the SimElectronics library while the microprocessor is represented by a Stateflow

block. Stateflow is a separate Matlab product that functions inside the Simulink environment to

create flow charts and state machines. The Stateflow block acts as the supervisory

microprocessor by processing output data from the CAOC and controlling the setpoint voltage

input to the CAOC.

80
Figure 40. HPMC Circuit

In figure 40, the tan colored block labeled PM_Controller is the Stateflow block that models the

supervisory microprocessor. Inside the Stateflow block, the APM control algorithm can be

implemented as a state machine or flow chart. From this point forward, the Stateflow block will be

referred to as the microprocessor block (MB). Components R1, R2, R3, U1, U2, U3, and Q1

make up the CAOC and perform the same function in figure 40 as they do in figure 39. Note that

the inputs to the MB are the fuel cell and battery currents and their integrated values. The MB

does not monitor the fuel cell voltage or the battery voltage; only the current values are necessary

for power management control. During flight, the throttle sets the motor power level as a

percentage. During takeoff and the initial climb, the throttle is set to 100% and full power

delivered to the motor. When the desired cruising altitude is reached, the aircraft levels out and

the throttle is reduced to 25%. Note that the MB does not use the throttle as an input. In this

scheme, power management does not require information about the throttle setting.

Inside the MB, there are three flow charts that define a simple power management

scheme used during flight. The first two flow charts act as coarse fuel gages for the battery SOC

and the hydrogen fuel level. They divide the battery SOC and hydrogen fuel level into three

regions: full, mid, or low.

81
Figure 41. Battery and Hydrogen Coarse Fuel Gages

For the two fuel gage flow charts shown in figure 41, the integrated battery and fuel cell currents

track the amount of electrical charge delivered from the two sources. For the battery state, charge

delivered from the battery is divided by 3600 before it is fed into the MB which changes the units

from coulombs to amp-hours. The amp-hours delivered from the battery is divided by the battery

capacity and then the ratio is subtracted from the initial SOC. When the battery energy reserve

drops below 90% of battery capacity, the battery state changes from full to mid. Likewise, when

the battery energy reserve drops below 40%, the battery state changes from mid to low. Note that

the battery state can also increase from low to mid to full. Since the battery can be recharged

during flight, the battery state must be able to change in both directions. For the hydrogen fuel

state, the charge delivered from the fuel cell is fed into the MB. The corresponding amount

hydrogen consumed is calculated in terms of liters at STP.

 n q 
Liters     24.46 (44)
 2F 

In equation 44, n is the number of cells in the fuel cell stack, q is the amount of charge delivered

from the fuel cell in units of coulombs, and F is Faraday’s constant. The factor 24.46 is the

number of liters per mole for any ideal gas at STP. The amount of hydrogen consumed is

subtracted from the total amount of hydrogen stored. When the amount of hydrogen fuel drops

below 80% of the total hydrogen initially stored, then the fuel state drops from full to mid. When

the hydrogen fuel level drops below 30%, the fuel state drops from mid to low. Note that the fuel

82
state can only decrease since the aircraft cannot refuel during flight. The third flow chart controls

the setpoint voltage to the CAOC. The setpoint is a voltage command to the CAOC which

controls the current delivered from the fuel cell. If the load demands more current than the fuel

cell setpoint level, then the battery automatically supplies the extra current required. When the

setpoint is at a maximum value of 9 V, then the fuel cell delivers current up to its nominal

operating point value before the battery delivers any current. For this model, the fuel cell

parameters are set to model the Horizon H-100 fuel cell stack and the nominal operating point is

8.3 A at 12 V. In other words, with a setpoint of 9 V, the fuel cell will supply load currents up to 8.3

A without any help from the battery. When the load requires current levels above 8.3 A, the

battery will augment the fuel cell current. As the setpoint is lowered, the battery will begin to

deliver current at lower load currents. For example, with the setpoint at 7 V, the fuel cell will

deliver up to 6.3 A; the battery will supply any load current above 6.3 A. Shown below in figure 42

is the setpoint flow chart (block name SP_Block).

Figure 42. Setpoint Flow Chart

In the flow chart displayed in figure 42, the setpoint starts out at a maximum value of 9 V.

At the beginning of a flight profile, maximum power is needed from both fuel cell and battery to

satisfy the power demands of 100% throttle. The exit condition from the leftmost block labeled

Normal is when the battery current drops below 0.1 A. This keeps the setpoint at 9 V during the

takeoff and climb phases of the flight profile. When the cruising altitude is reached, the throttle is

reduced from 100% to 25% at which point the fuel cell can supply the total load current and thus

the battery current drops to zero. Transition from the Normal block to the Cruise block then lowers

the setpoint to a level of 7.5 - h. The variable h is set by a parameter called hybrid which can vary

between 0 and 2. The parameter hybrid controls the ratio of fuel cell power to battery power

during cruise flight. In this manner, the battery can augment fuel cell power during cruising
83
conditions and extend flight endurance by trading battery power for hydrogen fuel. If during cruise

flight, the battery enters the low state, then the setpoint is raised to 7.5 by leaving the Cruise

block and entering the FConly block. This sets the fuel cell current level to the expected load

during cruise flight. Note that the FConly block works in conjunction with the Delta_1 block to

raise the setpoint to 7.5 V at a controlled rate. Once the setpoint reaches 7.5 V, the battery

current will drop to zero and the battery SOC will decrease no further under normal cruise load.

The state machine will also leave the cruise block when the cruise phase of the flight profile is

complete. At this point, the battery still has enough energy reserve to assist the fuel cell with

temporary loads greater than the expected cruise load for events such as wind gusts, aerial

maneuvers, or taxiing.

4.3 Hybrid Power System Model

To model the proposed UAV hybrid power system, the Simscape components are

assembled in a top level file. The fuel cell and the battery are connected to the electrical bus

through the HPMC. Two electrical loads, the motor-propeller model and the BOP model, are

connected directly to the bus. The battery charger receives power from the bus and its output is

connected to the battery. In addition, two observer blocks are used to monitor voltages, currents,

power, energy consumed, system state, and fuel cell model internal values. The diagrams for the

Observer_1 and Observer_2 blocks are listed in appendices O and P respectively. The model

configuration parameters choose the ode15s solver with a reduced relative tolerance of 10-6. A

screen capture of the model configuration parameters is displayed in appendix Q. The top level

diagram of the hybrid power system is shown below in figure 43.

84
Figure 43. Hybrid Power System Model

The throttle input to the hybrid power system is realized using a Simulink signal builder block at

the top level named Flight_Profile. The signal stored is the flight profile; that is, the throttle

settings during the takeoff, climb, cruise, and landing phases of flight. A second signal builder

input to the model is contained within the Balance_of_Plant block named the Fuel_Cell_BOP.

This signal represents the BOP load to the electric bus and is expressed directly in units of ohms.

The final input to the model is a signal builder block contained within the Controller named

BAS_signal. It activates the current assist feature of the HPMC. When the BAS_signal is zero,

current assist is not active. In this case, the fuel cell will supply power up to its nominal operating

point before the battery assists with excess load current demands. When the BAS_signal is one,

current assist is active and the hybrid parameter may lower the current assist threshold, allowing

the battery to augment the fuel cell at a lower power point than the nominal.

4.4 System Level Simulation

In 2012, the ASU Composite UAV was originally designed for electric powered flight

using a 5 Ah LiPo-3S battery pack (Krauch et al., 2012). The following year, a hybrid power plant

study investigated the use of a hybrid fuel cell and battery power system to potentially increase

the ASU Composite UAV flight endurance from 25 minutes (battery only), to 6 hours (Monaco,

2013). In this study, a hybrid power system for the ASU Composite UAV is now modeled in
85
Simscape. The power system simulation demonstrates both the functionality of the HPMC and

determines the amount of hydrogen fuel required to realize a 6 hour flight time.

In this section, the flight time is set to 6 hours with a typical flight profile and BOP load.

The flight profile starts out with a 3 minute interval at 100% power to simulate the UAV starting on

the runway and taking off. The next 4 minutes are set to a 60% power level which represents the

initial climb to cruising altitude. The UAV then cruises for 58 minutes at 25% power. At 65 minutes

into the flight, the UAV climbs again for 4 minutes at a shallow climb rate requiring a power level

of 40%. The UAV then cruises again for 45 minutes at 25% power. Wind gusts appear at the 114

minute time mark and persist for 7 minutes. During the wind gusts, the throttle alternates between

25% and 35% power to simulate brief intervals of increased power for course corrections. The

UAV then finds clear skies and smooth flight conditions to cruise for another 3.8 hours at 25%

power. At the 5.9 hour mark, the UAV is near the landing strip and temporarily increases power to

30% to line up for final descent to the runway. Power then decreases to 20% for 4 minutes during

the final descent and landing phases of the flight profile. Once on the ground, the total powered

flight time has been 357 minutes; almost 6 hours. The UAV then increases power to 30% to taxi

off the runway and return to its designated maintenance area. Finally, the UAV idles at 10%

power for a minute before shutting down. In total, the flight profile simulates 6 hours of power

delivered to the electric motor from the hybrid fuel cell/battery power system.

Figure 44. Simulation Input Signals

86
 There are three input signals to the simulation as shown above in figure 44. The first

signal is the throttle percentage which defines the flight profile. The second input signal is the

BOP load. The actual BOP load signal is defined in terms of ohms (section 3.5), but is estimated

in Watts in figure 44 for a hybrid value of zero. The fuel cell BOP signal sets the nominal BOP

load near 10.4 W. At four points during the flight, a 3 minute interval of increased BOP load to

roughly 16.9 W occurs to simulate the fuel cell fan activating. The third input signal is the BAS

signal which is either off (value = 0) or on (value = 1). In figure 44, the BAS signal is multiplied by

8 to make it easier to see the logic level on the graph. With the BAS_signal, current assist is

activated just before the 9 minute mark, maintained in the on state during the flight, and then

deactivated during final descent and landing.

In this first group of two simulations, the hybrid parameter is first set to zero and then to a

value of 1.5. When the hybrid parameter is zero, battery usage is minimized and hydrogen fuel

consumption is maximized. The fuel cell current threshold for battery assist is 6.8 A. In other

words, for electrical loads less than 6.8 A, the fuel cell will supply the total current delivered to the

load. If the electrical load goes above 6.8 A, then the battery contributes the extra load current.

When the hybrid parameter is set to a value of 1.5, the fuel cell current threshold for battery assist

is lowered to 5.3 A. That is, the fuel cell will deliver load currents up to 5.3 A while the battery will

deliver load currents above 5.3 A. The results of the two simulations are plotted in figures 45

through 47 for comparison. Measured runtime for each simulation is 25 seconds. The simulations

show that 366.5 liters of hydrogen are consumed when the hybrid parameter is set to zero and

327 liters are consumed when the hybrid parameter is 1.5. The battery final SOC is 72.1% when

the hybrid parameter is zero and 35.2% when the hybrid parameter is 1.5. Adjusting the hybrid

parameter from zero to 1.5 has effectively traded 36.9% battery SOC for 39.5 liters of hydrogen.

In the graph titles of figure 46, total energy is listed as 583.5 Wh for the hybrid parameter set to

zero, and 585.2 Wh for the hybrid parameter equal to 1.5. These energy values are the total

energy delivered from the fuel cell and battery during the flight profile. Note that there is a 1.7 Wh

increase in energy delivered from the sources when the hybrid parameter is increased to 1.5. The

87
energy increase is expected to be attributed to extra power dissipated by the HPMC transistor

Q1.

Figure 45. Fuel Cell and Battery Voltage and Current versus Time (hybrid = 0 and 1.5)

Figure 46. Electrical Power versus Time (hybrid = 0 and 1.5)

88
Figure 47. Battery SOC and Hydrogen Consumption versus Time

To automate the setting of model parameters and input signals, a Matlab script is used.

Within the flight profile script, the hydrogen pressure is set to 0.5 atm and the air pressure is set

to 1 atm. Other parameters set by the script are the total hydrogen fuel stored onboard, battery

capacity, battery initial SOC, number of cells in the fuel cell stack, and the hybrid parameter. Input

signals to the simulation are set using the Matlab signal builder function. The flight profile script is

listed in appendix R. Matlab scripts to generate figures 45, 46, and 47 are listed in appendices S,

T, and U respectively.

Inside the HPMC, transistor Q1 acts as a series regulator to control the battery current

delivered to the bus. When the battery current is zero, no power is dissipated by Q1; however, as

the battery current increases, so does the power dissipated by Q1. Shown below in figure 48 is a

closer look at the power consumed by transistor Q1. The Matlab code to generate figure 48 is

listed in appendix V.

89
Figure 48. Q1 Power Dissipation versus Time

From figure 48, the peak power dissipated by Q1 is 107 W which occurs at the 140 second time

mark. From the flight profile, the throttle is held at 100% from 140 seconds to 320 seconds. Q1

power decays during this time interval while the propeller spins at its top rotational speed of 7300

RPM. As the throttle is reduce from 100%, the battery current decreases and so does Q1 power.

With the hybrid parameter at zero, Q1 power drops to zero when no current assist is needed from

the battery. In this case, the HPMC draws minimal power; only the quiescent power necessary for

the electronics. When the hybrid parameter is 1.5, Q1 cruise power is 1.29 W and the battery

delivers a cruise assist current of 956 mA. During this time, the voltage dropped across Q1

(source relative to drain) is Vsd = 1.35 V and the effective source-to-drain resistance is 1.41 Ω.

If the power supplied by the sources, the power dissipated by Q1 and the load are

integrated over the entire flight profile, the total energy of the sources and loads can be found.

Taking the difference of the total energies between the hybrid = 1.5 and hybrid = 0 simulation

runs provides further insight into the operation of the system. Subtracting the total energies of the

hybrid = 0 run from the hybrid = 1.5 run gives the following results: the sources produce 1.7462

Wh more energy, Q1 dissipates 3.4157 Wh more energy, and the motor-propeller load consumes

1.4339 Wh less energy. The expected result is that the change in Q1 load energy plus the

change in motor-propeller load energy should equal the change in source energy. Note that the

90
motor-propeller load consumes less power, so the change in motor-propeller load energy is

negative. The change in BOP load energy is negligible.

3.4157  Wh  1.4339  Wh  1.9818  Wh  1.7462  Wh (45)

3600  J 
1.9818  Wh  1.7462  Wh      848.16  J (46)
 Wh 

Equation 46 shows that there is an 848 Joule discrepancy in the change of total energy between

simulation runs. When compared to the total energy consumed during the hybrid = 1.5 simulation

run, this amounts to a 0.04% error. The Matlab script which calculates the change in total

energies between runs is listed in appendix W. Table 6 shown below lists the total energies of

both simulation runs.

Table 6

Simulation Total Energies for Hybrid = 0 and 1.5

Simulation Run hybrid = 0 hybrid = 1.5 change (h1.5 - h0)
Sources Total Energy (Wh) 583.4839 585.2301 1.7462
Motor Total Energy (Wh) 510.9646 509.5307 -1.4339
Q1 Total Energy (Wh) 7.8764 11.2921 3.4157
BOP Total Energy (Wh) 61.2790 61.2790 0.0000
Sources minus Loads (Wh) 3.3639 3.1283 -0.2356

In table 6, the loads consist of the motor-propeller load, Q1 load, and BOP load; the sources are

the fuel cell and the battery. Note that for the two simulations, the sources produce about 3 Wh of

extra energy that is not accounted for in the listed loads. The extra energy is consumed by the

two DC-to-DC converters in the BOP block and the battery charger block. Note in figure 45, the

bus voltage is 14.50 V when the current assist is active and the bus voltage drops to 13.65 V

when there is no current assist. During the h = 1.5 simulation run, the bus voltage is at the higher

voltage level of 14.5 V for about 3 hours. Note that the excess energy for the h = 1.5 run is lower

than that of the h = 0 run and the difference between the two values accounts for the 848 J

discrepancy. Also, having calculated the total energy dissipated by Q1 and the total energy

produced by the sources, power system efficiency can be calculated as the ratio. For the h = 0

91
and h = 1.5 runs, power system efficiency excluding the BOP and battery charger loads is

calculated at 98.6% and 98.1% respectively.

To reach a flight time of 6 hours for the ASU Composite UAV, the two previous

simulations calculated the amount of hydrogen required to be 366.5 L without current assist and

327 L with current assist. If the hydrogen storage tank onboard the UAV is the Ninja Pro v2

90/4500 filled to a pressure of 248 bar as mentioned in section 2.9, then the total amount of

hydrogen available is only 298.9 L. The LiPo battery used during the first two simulations was

modeled after the MaxAmps LiPo 4S 8Ah battery. Two possibilities to increase the flight time to 6

hours are to increase the hydrogen pressure of the storage tank or to use a larger battery. For the

first option of increasing the hydrogen storage tank pressure, the 20% safety factor mentioned in

section 2.9 can be decreased. The hydrogen storage tank is approved for air pressures up to 310

bar. The problem is that the storage tank is designed to hold compressed air and not hydrogen.

Decreasing the safety factor from 20% down to 12% may not be safe. However, if the storage

tank is filled to a pressure of 272.8 bar, then 328.8 L of hydrogen can be stored. Using the current

assist feature and filling the storage tank to 272.8 bar would result in a 6 hour flight time. For the

second option of a larger battery, a second group of simulations varies the LiPo battery capacity

from 8 Ah to 12 Ah and determines the amount of hydrogen consumed during each simulation.

Results for this group of five simulations are shown below in figure 49 and the Matlab code is

listed in appendix Y. The amount of hydrogen fuel consumed during each simulation run is listed

in the graph legends. For example, a line in the graph legend lists the text “Battery: 8Ah,

H2:327.2”. This translates to 327.2 liters of hydrogen consumed over the six hour flight profile for

the case where the battery capacity equals 8 Ah. The hybrid parameter is set to 1.5 for all

simulation runs of figure 49 and table 7.

92
Figure 49. Battery SOC and Hydrogen Consumed for Different Battery Capacities

From figure 49, increasing the battery capacity to 12 Ah is necessary to keep the hydrogen

consumption under 298.9 L. Table 7 listed below shows the total hydrogen consumed and final

battery SOC for each of the battery capacities simulated over the six hour flight profile.

Table 7

Hydrogen Consumed and Battery Final SOC for Different Battery Capacities
Battery Capacity Hydrogen Consumed Hydrogen Consumed Battery Final SOC
(Ah) (L) (g) (%)
8 327.2 27.0 35.2
9 319.0 26.3 35.4
10 311.6 25.7 36.3
11 304.1 25.1 36.9
12 296.8 24.5 37.5

Table 8

LiPo 4S Battery Weights (“MaxAmps LiPo 4S Family”)

MaxAmps Battery Size Weight Discharge Rating
(Ah) (kg) (C)
8 0.764 150
9 0.862 100
11 0.850 40
12 1.110 100

93
From table 8, changing from the LiPo 4S 8 Ah battery to the 12 Ah battery increases the power

system weight by 346 g. Also note that the 11 Ah battery has a significantly lower battery weight

of only 850 g. If the 11 Ah battery could be used, then the power system weight would only

increase by 86 g. The weight savings of the 11 Ah battery over the 12 Ah battery is probably due

to the lower discharge rating. From figure 45, the maximum battery current is under 25 A; well

below the 40 C discharge rating of the 11 Ah battery. If the hydrogen tank pressure is increased

to 254 bar (18% safety factor), then 306 L of hydrogen is available and the 11 Ah battery is

sufficient to complete the six hour flight. Due to the reduced weight of the 11 Ah battery, and only

a 2% compromise to the hydrogen storage tank safety factor, this third option is the most

attractive. Figure 50 shown below displays the power system performance during the six hour

flight profile for the 11 Ah battery case (hybrid = 1.5).

Figure 50. Hybrid Power System Performance for 11 Ah Battery

In figure 50, the controller deactivates the current assist at the 300 minute time mark due to the

battery SOC entering the Battery_Low state (40% SOC). At this point, the CAOC setpoint is

increased from 6 to 7.5 at a controlled rate (see figure 42). Note that the bus voltage (fuel cell

voltage) decreases from 14.5 V to 13.65 V during this transition. Since the fuel cell voltage is

higher when the current assist is active, an added benefit of increased fuel cell efficiency is

realized. Using the higher heating value (HHV) of hydrogen, the fuel cell efficiency is 49% with

current assist and 46% without. If referencing the lower heating value (LHV) of hydrogen, then the

fuel cell efficiency is 58% with current assist and 54.6% without. To examine the current and

94
power graphs from figure 50 more closely, figure 51 shows the same data with the y-axis ranges

changed.

Figure 51. Hybrid Power System Performance for 11 Ah Battery Expanded View

To summarize, system level simulations have shown that with 306 L of hydrogen and an 11 Ah

LiPo battery, a six hour flight of the ASU Composite UAV is possible using the current assist

feature of the HPMC. Without current assist, 366.5 L of hydrogen would be necessary to

complete the six hour flight. Using battery power during cruise conditions has in effect, reduced

the amount of hydrogen required by 60.5 L and has transferred about 74 Wh of energy load to the

battery. The battery SOC at the end of the flight is 79.7% without current assist and 36.9% with

current assist. The difference between these battery SOC values is due to the battery delivering

the extra 74 Wh of energy required to complete the six hour flight profile.

4.5 Inflight Battery Recharge

When excess hydrogen is available, the battery charger circuit can be used to recharge

the battery in flight. Mission planners may want to recharge the battery in flight to reach higher

altitudes or to reduce maintenance time on the ground. Although the battery charger is already

integrated into the hybrid power system model (see figure 43), the controller SP_Block (see figure

42) must be altered. With the battery charger, the controller SP_Block is replaced by the

CTRL_Block shown below in figure 52.

95
Figure 52. CTRL_Block Flow Chart

The Stateflow default transition is the starting point for the CTRL_Block flow chart and enters the

Normal block on the left side. The CAOC setpoint voltage is assigned the value stored in the

flowchart variable Setpoint. Starting at the Normal block, a working variable named csp (current

setpoint) is initialized to a value of 9 which is then assigned to Setpoint. When the variable

Setpoint is updated, the CAOC setpoint voltage is updated to a physical value of 9 V. When the

battery current drops below 0.1 A and the BAS signal has a value of one, then the state machine

transitions from the Normal block to the Ctrl_Mode block. The Ctrl_Mode block simply determines

if the battery charger is active during the simulation run. If the battery charger is active, then

variable ca (charger armed) is one, otherwise it is zero. Note that the flow chart logic of the

SP_Block is contained within the CTRL_Block. At the Ctrl_Mode block, the transition branch

where ca = 0 goes to the same flow chart as the SP_Block. In other words, when the battery

charger is not active (ca = 0), the same SP_Block flow chart logic is used. The value of ca is set

in the Matlab environment by the flight profile script. If the value of ca is set to one, then the

battery charger is active and the state machine will transition from the Ctrl_Mode block to the

Charger_Gate block. The Charger_Gate block then waits for the BAS signal to equal a value of

two before it activates the battery charger. Here, the BAS signal is used for two purposes. When

the BAS signal is zero, neither current assist or battery charger are active. With a BAS signal

value of one, the current assist is active but the battery charger is not. When the BAS signal has

a value of two, both current assist and battery charger are active. The BAS signal controls at

which time intervals during the flight current assist and the battery charger are active. Shown

96
below in figure 53 is a graph of the input signals to the hybrid power system model for the six

hour flight profile. When the BAS signal changes value from one to two, the state machine

transitions from the Charger_Gate block to the Charger_On block. Within the Charger_On block,

the BatCharge variable controls the rate of battery charging and is set to variable cv which is

assigned in the Matlab environment by the flight profile script. A BatCharge value of 0.8

corresponds to a battery charge current of 302 mA. The battery charger will remain active until an

appropriate exit condition is reached. Exit conditions include the BAS signal changing value to

one or zero, or the fuel cell current exceeding 8 A, or the battery charge state entering the

Battery_Full state, or the hydrogen fuel state entering the Fuel_Low state. Upon any valid exit

condition, the state machine transitions from the Charger_On block to the Charger_Off block.

Near the end of the flight profile, the BAS signal will change to zero and the state machine will

transition from the Charger_Off block to the FConly block. The CAOC setpoint is then raised to

7.5 as previously done with the SP_Block.

Figure 53. Input Signals for Inflight Battery Recharge Simulation

In figure 53, the BAS signal value is multiplied by 4 to make it easier to see the logic level on the

graph.

Transitions between the Charger_On block and the Charger_Off block are two way.

Assuming that the fuel state is above the Fuel_Low state and battery charge state is below the

Battery_Full state, the fuel cell current must be under 8 A for the battery charger to remain active.

If the fuel cell current goes above 8 A, then the battery charger is deactivated and may reactivate

when the fuel cell current drops below 8 A. This limit on the fuel cell current prevents overloading

97
of the fuel cell and guarantees the bus voltage will remain above 12 V. If this safeguard was not

in place, then current assist events that occur during battery recharge may overload the fuel cell

which would cause the bus voltage to drop below its minimum operating voltage. A second

condition that may cause the state machine to toggle between the Charger_On block and the

Charger_Off block is battery SOC. To place an upper limit on the battery SOC, the Battery_Full

state signals that the battery charge is complete. If during flight, the battery charge state drops

below 90% due to current assist or throttle commands, then the charger may reactivate and

continue to charge the battery up to its 90% charge limit.

Figure 54. Fuel Cell and Battery Voltage and Current during Inflight Battery Recharge

Figure 55. Battery SOC Current during Inflight Battery Recharge

4.6 Flight Time and Hybrid Power System Weight

To evaluate possible flight times for various hybrid power system configurations, the six

hour flight profile can be used. The flight profile is partitioned into three sections, takeoff, cruise,

and landing. During cruising conditions, the power consumed can be considered constant if wind

gusts and BOP variations are ignored. During takeoff and landing portions of the flight profile,

power is not constant, but the total energy consumed is. Integrating load power during takeoff and
98
landing sections of the flight profile determines the total energy consumed. Power system

configurations can then be evaluated in terms of total energy required. The total energy used

during takeoff and landing is considered constant while the power used during cruise is in terms

of energy per unit time. The total energy a hybrid power system configuration can deliver (EHPS) is

then used to determine the maximum flight time possible. The takeoff and landing energy (ETOL)

is subtracted from EHPS and the amount of energy left over determines how long the UAV can

remain at cruise conditions. First, the electrical power consumed from the motor-propeller block is

added to the power used by the BOP block. The total electrical load energy is then plotted in units

of Watt-hours per minute (Wh/min). Figure 56 below displays the power consumed during the six

hour flight profile in units of Wh/min.

Figure 56. Power Consumed During Six Hour Flight (Wh/min)

From figure 56, the cruise power is found to be 1.51 Wh/min. From the flight profile, the takeoff

segment occurs from time zero to 480 seconds and the landing segment occurs between 21200

seconds and 21800 seconds. Integrating the power during takeoff and landing gives the total

takeoff energy as 31.9 Wh and the total landing energy equal to 13 Wh (ETOL = 44.9 Wh). The

Matlab script that calculates these energies can be found in appendix Y. Flight time is then

calculated in terms of total energy available from the hybrid power system configuration.

99
 t  [ EHPS  44.9  / 1.51]  18 (47)

where: t = total flight time (minutes)

EHPS = total energy available from hybrid power system (Wh)

Note that takeoff accounts for 8 minutes and landing takes 10 minutes. In equation 47, since the

energy for takeoff and landing is subtracted from the energy available (EHPS – ETOL), the 18

minutes for takeoff and landing must be added to the total flight time. To calculate EHPS, the fuel

cell and battery energies are added together. The amount of energy the fuel cell produces is

equal to the amount of hydrogen energy available times the fuel cell efficiency. For one liter of

hydrogen at STP, the equivalent Wh of energy is calculated in equation 48 shown below.

1 mol   2.02  g   120  kJ   1 Wh 

1 L   24.46
H2
 L   mol 
    2.7528  Wh
 3.6  kJ 
(48)
 g 

To approximate the fuel cell efficiency during the flight, the average between the two efficiencies

obtained with current assist and without is 56.3% (LHV). Therefore, to convert from liters of

hydrogen available (LHA) to total watt hours delivered by the fuel cell (EFC), equation 49 shown

below can be used as an approximation.

 2.7528  Wh 
EFC   LHA    0.563  1.55  LHA (49)
 L 

For the total energy delivered from the battery (EBAT), the battery capacity (CBAT) in Ah is

multiplied by the battery voltage. From figure 50, the average battery voltage during flight is

approximately 15.7 V. Assuming that the battery starts out at 100% SOC and ends at 36% SOC

(see table 7), there is 64% of the battery capacity available. Then, to convert from battery

capacity to total energy available from the battery, equation 50 shown below can be used as an

approximation.

EBAT  CBAT  15.7   0.64  10  CBAT (50)

Combining equations 49 and 50, a useful approximation for EHPS is obtained.

EHPS  1.55  LHA  10  CBAT (51)

100
Substituting equation 51 into equation 47, the flight time is calculated as a function of hydrogen

available and battery capacity.

 1.55  LHA  10  CBAT  44.9 

t    18 (52)
 1.51 

To check the validity of equation 52, simulation data from figure 49 can be used. Actual flight time

for the six hour flight profile is 21800 seconds or 363.3 minutes. Figure 49 lists the simulation

results for the amount of hydrogen consumed over a range of LiPo 4S battery capacities. Using

this data, the calculated flight time using equation 52 is compared against the actual flight time.

Table 9

Comparison of Simulation Data and Flight Time Approximation Formula

LiPo Battery Capacity Hydrogen Available Calculated Flight Error from Actual
(Ah) (L) Time (min) Flight Time (%)
8 327.2 377.1 3.8
9 319.0 375.3 3.3
11 304.1 373.3 2.7
12 296.8 372.4 2.5

From table 9, the error between calculated flight time and actual flight time is under 4%. Note that

the calculated flight time is larger than the actual flight time of 363.3 minutes. Cruise power is

considered constant and variations in cruise power due to wind gusts and BOP load changes are

ignored. The approximation of equation 52 assumes a fuel cell operating at 56.3% efficiency, a

LiPo 4S battery, current assist is active during cruise, and 64% of battery capacity consumed. For

the case where no fuel cell is present and only a battery is used, then LHA can be set to zero.

To evaluate hybrid power system performance for different configurations, power system

weight and flight time are two useful metrics. Using only a battery, the power system weight is

lower than a hybrid configuration, but flight endurance is limited. Using only a 100 W fuel cell will

not work for the ASU Composite UAV since the takeoff power required is 400 W (see figures 46

and 48). Instead, if a 500 W fuel cell were used, then only the fuel cell is required onboard. The

500 W fuel cell startup could be powered externally before takeoff. Using commercially available

LiPo batteries and compressed air tanks, a comparison of weight versus flight time is graphed for

101
combinations of power system components. Total power system weight includes contributions

from the HPMC, fuel cell stack, fuel cell controller, hydrogen storage tank, pressure regulator, and

the battery. A commercial brand of LiPo 4S batteries is MaxAmps (www.maxamps.com).

Table 10

MaxAmps LiPo 4S Battery Family Specifications (“MaxAmps LiPo 4S Family”)

Battery Capacity Weight Discharge Rating
(Ah) (kg) (C)
4 0.388 150
6 0.570 100
8 0.764 150
9 0.862 100
11 0.850 40
12 1.110 100
16 1.244 20
17 1.294 15
23 1.664 25

Three commercially available fuel cells are from Horizon Fuel Cell Technologies

(www.horizonfuelcell.com). The H-100 and H-500 fuel cells are for general use and are not

weight optimized. The AST01-01 is a 200 W fuel cell that is optimized for low weight aviation

applications.

Table 11

Horizon Fuel Cell Weights

Stack Weight Controller Weight Total Weight
Model # (kg) (kg) (kg)
H-100 1.29 0.40 1.69
H-500 2.52 0.40 2.92
AST01-01 0.47 0.08 0.55

State of the art carbon fiber compressed air tanks are available from www.ansgear.com. All three

Ninja Pro v2 tanks are rated for air pressures up to 310 bar. For hydrogen storage, a 20%

reduction in maximum pressure is assumed as a safety factor. The listed hydrogen storage

capacities in table 12 shown below are calculated at a storage pressure of 248 bar.

102
Table 12

Ninja Pro v2 Tank Weights and Storage Capacities (www.ansgear.com)

Weight Storage Capacity
Model # (kg) (L)
50/4500 0.936 166.0
68/4500 1.190 225.8
90/4500 1.450 298.9

On all Ninja Pro v2 storage tanks is a high pressure regulator which steps down the tank pressure

from 248 bar down to 27.6 bar. The weight of this first stage regulator is included in the tank

weights listed in table 12. A second stage pressure regulator is necessary to further reduce

hydrogen pressure down to 0.52 bar required by the fuel cell. A commercially available pressure

regulator is the Beswick miniature pressure regulator which has a weight of 0.0062 kg (“Beswick

Pressure Regulator”, 2016). The weight of the HPMC is assumed to be 0.032 kg.

There are 9 power system configurations considered, two conventional (non-hybrid) and

seven hybrid. The two non-hybrid configurations are battery only and fuel cell only. The first

configuration is with the battery only and three battery capacities are chosen: 6 Ah, 11 Ah, and 17

Ah. Equation 52 is used to calculate the expected flight time for each battery capacity (LHA = 0).

A line representing this power system configuration is then constructed on a weight versus flight

time graph as follows. For the lowest battery capacity, the weight is the smallest of the three

batteries. This is the y-coordinate of the first point. Minimum flight time is assumed to be 18

minutes since this is the amount of time for takeoff plus landing. This is the x-coordinate of the

first point. For the second point, the calculated flight time of the 6Ah battery is used as the x-

coordinate while the y-coordinate remains the same. For the third point, the next larger weight of

the 11 Ah battery is used as the y-coordinate while the x-coordinate remains the same as the

second point. The fourth point uses the calculated flight time of the 11 Ah battery as the x-

coordinate while the weight or y-coordinate is the same as the third point. The fifth point keeps

the x-coordinate of the fourth point and uses the 17 Ah battery weight as the y-coordinate. Last,

the sixth point uses the calculated flight time of the 17 Ah battery as the x-coordinate and the y-

coordinate is the same as the fifth point. The end result is a stair-step line with three steps on the

103
graph. Each step represents the minimum weight for the power system configuration for a range

of possible flight times. Next, consider the other non-hybrid configuration which uses the H-500

fuel cell without a battery. The construction of this line is similar to the battery only line except the

weight increase is not due to larger battery sizes; instead the weight increase between steps is

due to larger hydrogen storage tank sizes. The three hydrogen storage tanks considered are the

Ninja Pro 50, 68, and 90. The resulting H-500 only line is again a stair-step line with three steps.

Next, the seven hybrid power system configuration lines are constructed in similar manner to the

H-500 line; but in these cases both fuel cell and battery are present. Six of the hybrid

configurations are combinations of the H-100 or AST01-01 fuel cells combined with the 6 Ah, 11

Ah, or 17 Ah batteries. In all six of these hybrid configuration cases, the current assist feature of

the HPMC is active. Last, the seventh hybrid configuration is labeled the conventional hybrid

configuration where a DC-to-DC converter is present and no current assist is used. For this

conventional hybrid case, a minimum battery size necessary for takeoff and landing is chosen;

the 6 Ah battery. The weight of a 400 W DC-to-DC converter (0.22 kg) is added to the power

system weight which represents a conventional approach for hybrid power systems management.

This conventional hybrid case is special in that it provides a baseline of comparison for the

transistor-based, current assist power management configurations. Since the conventional hybrid

configuration does not use current assist and it includes the weight of a DC-to-DC converter, the

benefits of the transistor-based HPMC using current assist can be noted by comparison. The end

result is a graph of the power system weight versus flight time as shown below in figure 57. The

Matlab script to calculate and plot flight times in figure 57 is listed in appendix AA. The flight time

calculations using equation 52 are approximations only. They assume that the aerodynamic

characteristics of the ASU Composite UAV do not change when the weight changes. In actuality,

the energy used during takeoff and landing as well as cruise power will vary as weight changes.

Additionally, the distribution of weight within the ASU Composite UAV fuselage will affect the

aerodynamic performance of the airframe. Therefore, the flight time data presented in figure 57

should be considered as a rough estimate only, not accounting for aerodynamic performance

changes. All flight time data in figure 57 is applicable to the ASU Composite UAV only.
104
Figure 57. Flight Time versus Weight for Power System Configurations

In figure 57, there are three thin black lines that extend diagonally across the graph.

These are the battery extrapolation lines. These lines represent the weights and flight times

expected when using multiple batteries. The left-most battery extrapolation line represents the

case where multiple 6 Ah batteries are used without a fuel cell. Note that there are 8 circle

markers on this line; each marker represents an additional 6 Ah battery. A total of eight 6 Ah

batteries (0.57 kg each) can fit on the graph where the upper weight limit is 5 kg. Similarly, the

right-most battery extrapolation line represents the case where multiple 17 Ah batteries are used

without a fuel cell. There are 3 circle markers on this line which are the number of 17 Ah batteries

(1.294 kg each) that are under 5 kg. Between the left and right battery extrapolation lines is a

sector on the graph where battery only performance can be expected (battery only sector).

The six non-conventional hybrid configurations are cases that use the transistor-based

HPMC presented in this study. In all six cases, the current assist feature of the HPMC is used.

This group can be divided into two sub-groups; one sub-group using the H-100 fuel cell (solid

lines) and the second using the AST01-01 fuel cell (dashed lines). Each sub-group uses the

same color code for the battery capacity present in the configuration (blue = 6 Ah, red = 11 Ah,

105
green = 17 Ah). The two sub-groups have the same flight times, but the AST01-01 sub-group is

shifted downwards due to the lower weight of the AST01-01 fuel cell.

From figure 57, a comparison between the conventional hybrid configuration (bold purple

line) and the H-100 using the 6 Ah battery hybrid configuration (blue line) can be made. Both

power system configurations have the same fuel cell and battery; only the power management

controllers are different. The conventional hybrid system uses a DC-to-DC converter to control

the power split between fuel cell and battery. The non-conventional configuration (blue line) uses

the transistor-based HPMC approach discussed in sections 4.1 and 4.2. A weight savings of 0.22

kg shifts the blue line down from the purple line on the graph. Due to the current assist feature of

the HPMC, the blue line is also shifted to the right by 40 minutes compared to the purple line. The

gain in flight time is due to the remaining battery energy available after takeoff and landing

energies are subtracted out. In other words, the conventional hybrid configuration does not have

current assist and cannot take advantage of excess battery energy available.

The graph presented in figure 57 can be divided into three regions. First, the region on

the graph between the battery extrapolation lines is labeled the battery only sector. The region to

the left of the battery only sector is where power system performance is worse than using battery

power only. In this context, the term power system performance can be defined as the ratio of

flight time over power system weight (minutes per kg). This first region to the left of the battery

only sector will be labeled region one. Within the battery only sector, power system performance

is roughly equivalent to using battery power only. This second region defined by the battery only

sector will be labeled region two. The third region is the area on the graph to the right of the

battery only sector. Within region three, power system performance is better than a battery only

power system. For a hybrid system to perform better than a battery only system, points on a

configuration line should lie within region three. Note that the boundaries of region 2 are not a

fixed ratio of flight time over weight, instead there is a range of ratios along the boundary. Along

the boundary between region 1 and 2, the flight time to weight ratios range from 46 min/kg to 67

min/kg. Along the boundary between regions 2 and 3, the flight time to weight ratio ranges from

63 min/kg to 84 min/kg.
106
 Interpreting the results of the power system weight versus flight time graph of figure 57

develops several discussion points. For the battery only configuration, power system weight is the

lightest but flight time is restricted to under 2 hours when using only one battery. The flight time to

weight ratio for the three batteries (6 Ah, 11 Ah, 17 Ah) are 49.1, 71.9, and 77.9 min/kg

respectively. The trend shows that larger battery capacities yields better ratios. With the H-500

fuel cell only configuration, for the three storage tanks (Ninja 50 = 166L, Ninja 68 = 225.8 L, Ninja

90 = 298.9L), the flight time to weight ratios are 41.1, 53.5, and 67.4 min/kg respectively. Better

ratios are obtained with larger hydrogen storage tanks. As seen on the graph of figure 57, the H-

500 fuel cell only configuration line lies almost completely in region 1. In other words, using the H-

500 fuel cell only configuration with the Ninja 68 tank yields worse performance than loading the

UAV with six 6 Ah batteries. Using the H-500 fuel cell with the Ninja 90 tanks yields worse

performance than loading the UAV with three 17 Ah batteries but slightly better performance than

eight 6 Ah batteries. Next, for the three H-100 hybrid configurations, much of the configuration

lines lie within region 2. That is, using the H-100 fuel cell with Ninja 50 or 68 tanks yields flight

time to weight ratios similar to battery only configurations. Only when using the Ninja 90 tanks, do

the H-100 configuration lines extend into region 3. The H-100 hybrid configurations using the

Ninja 90 tank give flight time to weight ratios of 90.1, 92.0, and 91.8 min/kg for the 6 Ah, 11 Ah,

and 17 Ah batteries respectively. One would like to see better performance from hybrid power

configurations, but the H-100 fuel cell is a general purpose fuel cell, not weight optimized for

aviation applications. Next, for the AST01-01 hybrid configurations, all maximum flight time points

lie in region 3. Any hybrid power system configuration using the AST01-01 fuel cell performs

better than battery only systems. The flight time to weight ratios of the AST01-01 hybrid

configurations using the Ninja 90 tank are 129.9, 128.8, and 123.5 min/kg for the 6 Ah, 11 Ah,

and 17 Ah respectively. The AST01-01 fuel cell is weight optimized for aviation applications and

the weight savings significantly improve performance to be superior to battery only solutions. The

AST01-01 hybrid configurations demonstrate the importance of minimizing weight when

designing hybrid power systems for UAV applications.

107
 CHAPTER 5

DISCUSSION AND CONCLUSION

5.1 Research Results

Using the Simscape language to model the battery, fuel cell, and electrical loads proved

to be helpful at both the component level and the system level. The main advantage of the

Simscape physical modeling approach over the conventional Simulink approach was the ability to

partition the model into independent physical component blocks. The battery, fuel cell, motor-

propeller, and BOP models were developed separately from the top level power systems model.

Each component was tested and verified in isolation which simplified the modeling task. When

the top level hybrid power system model was assembled, there were no concerns about changing

the placement or connections between components.

The battery model is based on a commercially available LiPo 4S 8Ah battery and uses

the Tremblay battery model reported in literature (Tremblay & Dessaint, 2009). Implementation of

the model is realized as a Simscape text file which is listed in appendix A. To verify the battery

model and to adjust the model parameters, an adjustable constant current load is connected to

the battery. Parameters for the battery model are then adjusted to match the discharge curves of

the actual LiPo 4S battery over a range of discharge rates. The battery model verification result is

presented as a group of battery discharge curves plotted in figure 17. To check the validity of the

battery model in a second way, the battery discharge curve for a 0.5 C discharge rate is

compared against the Sim Power Systems lithium ion battery model. Although the two curves are

not a perfect match, they are in close agreement. The combined plot of the Tremblay battery

model and the Sim Power Systems lithium ion battery model 0.5 C discharge curves are shown in

figure 18.

The fuel cell model is based upon the Horizon H-100 fuel cell and uses the electrical

circuit equivalent model shown in figure 24 (b). Alternative mathematical expressions for the

Nernst potential, activation polarization, and ohmic polarization are presented by Lee in literature

and are used in this fuel cell model (Lee & Wang, 2007, Qingshan et al., 2008). The

implementation of the fuel cell model is a hierarchical Simscape model where each
108
subcomponent consists of a text file and the top level fuel cell component is another text file

which assembles the subcomponents. Listings of the subcomponent models and the fuel cell

component model are listed in appendices D, F, H, I, J, K, and L. To test the fuel cell model, an

adjustable constant current load is connected to the fuel cell. During a test simulation, the load

current is varied and the fuel cell voltage and current values are recorded (figure 29). The fuel cell

voltage is then plotted against the current to obtain the polarization curve. In addition, fuel cell

power is also plotted against current yielding the fuel cell power curve. Both plots are presented

in figure 30. The fuel cell model parameters are adjusted so that the model polarization curve

matches the polarization curve reported for the H-100 fuel cell. Two key points on the polarization

curves are the nominal operating point at (8.3 A, 12 V) and the open circuit point at (0A, 19V).

The power curve peaks at (8.3 A, 100 W) in agreement with the H-100 fuel cell power curve

(Horizon, 2013). The dynamic response of the fuel cell is also simulated for a range of double

layer capacitances which is presented in figure 32.

The electrical load consists of two components, the motor-propeller model and the BOP

model. The BOP model is implemented as a Simscape graphical file and consists of a small DC-

to-DC converter and a programmable resistor load (figure 33). A signal builder block controls the

load resistance and simulates a 10 W BOP load with short intervals of 17 W load to represent the

fuel cell fan load. The motor-propeller model is also a Simscape graphical file (figure 35). The

motor-propeller electrical load is based upon data obtained with the TTB and motor manufacturer

data. A test simulation varies the throttle input and records the output electrical and mechanical

power. The end result of the test simulation are two graphs, propeller speed and power versus

throttle percent (figure 36).

The HPMC concept is developed with the CAOC described in section 4.1. Instead of

using a DC-to-DC converter in the main power path, a MOSFET is used to regulate both battery

and fuel cell current. A setpoint voltage determines a fuel cell threshold current for which any load

current demands above the threshold are supplied by the battery. The setpoint may also be set

below the cruise load current value so that the battery contributes to cruise power. The HPMC is

implemented as a Simscape graphical file shown in figure 40. Controlling the CAOC setpoint
109
voltage is a Stateflow block which simulates the function of a microprocessor. Functions of the

Stateflow block are discussed in section 4.2. Simulation tests of the HPMC show the CAOC

functioning as expected (figures 45 and 46). To verify that the fuel cell and battery are operating

within their safe zones, figure 45 shows the fuel cell current is maintained under its maximum

operating current and figure 47 shows the lowest battery SOC is kept above 35%.

The complete hybrid power system is assembled as a graphical Simscape file and is

displayed in figure 43. The flight profile is the input to the power system model which simulates a

typical six hour flight. The flight profile consists of three signals, the throttle command, the BOP

load, and a logic control signal (BAS signal). A graph of the six hour flight profile is shown in

figure 44. The simulation results presented in section 4.4 show that 366.5 L of hydrogen are

consumed without current assist and 327 L of hydrogen are used when current assist is active.

Battery SOC and fuel cell current are maintained in their safe zone of operation as shown in

figures 45 and 47.

The battery charger model is developed in section 3.3 and the Simscape graphical file is

displayed in figure 19. The battery charger model uses a small DC-to-DC converter to create a 24

volt supply that is combined with a programmable constant current circuit. To test the battery

charger, it is connected to the battery model and a simulation is run where the battery starts out

at 40% SOC and is charged at a 0.1 C rate up to 96% SOC. Results of this test simulation show

the battery SOC increasing linearly in figure 21, the battery voltage during recharge in figure 22,

and the constant charge current in figure 23. Inflight battery recharge simulations are then carried

out in section 4.5. Additional logic is added to the Stateflow block within the HPMC block to

control the charger (figure 52). Inflight battery recharge is simulated and the results displayed in

figure 55 show the battery SOC increasing during the flight. The Stateflow logic that controls the

battery charger sets a battery SOC upper limit at 90%. During simulation, this upper limit is

reached and figure 55 shows the charge cycle ending at the 90% upper limit.

To derive an algebraic expression to calculate flight time for different batteries and

hydrogen storage tanks, the input flight profile is divided into three sections, takeoff, cruise, and

landing. Noting that the throttle command is roughly constant during cruise, this attribute is
110
leveraged to determine flight time. The energy consumed during takeoff and landing are

subtracted out and the flight time is determined by the constant energy consumption during cruise

plus the time elapsed during takeoff and landing. Next, the usable energy from the stored

hydrogen and battery are combined and used to determine potential flight time. The resulting

flight time equation is listed as equation 52. Next, a group of commercially available batteries and

compressed air tanks are selected and their weights and capacities are listed in tables 10 and 12

respectively. The flight time equation is then used to approximate potential flight times for nine

power system configurations. The first configuration consists of a battery only configuration and

flight times are calculated for 3 battery sizes. A second configuration consists of a Horizon H-500

fuel cell only and flight times are calculated for three hydrogen storage tank sizes. Six hybrid

power configurations consist of the Horizon H-100 or AST01-01 fuel cell and a 6 Ah, 11 Ah, or 17

Ah battery. Flight times are calculated for all six hybrid configurations. The last configuration is a

conventional hybrid power configuration using the H-100 fuel cell, 6 Ah battery and a DC-to-DC

converter. This last configuration serves as a baseline for comparison of the other hybrid

configurations. The results are displayed as a graph of power system weight versus flight time

shown in figure 57. The increased ratio of flight time to weight for the AST01-01 hybrid

configurations over the H-100 hybrid configurations demonstrate the importance of minimizing

weight in UAV power systems. Results from figure 57 also demonstrate the importance of

hydrogen storage in UAV applications. The more hydrogen that can be stored onboard, the

greater the benefit of hybrid power systems.

111
5.2 Future Research

The series regulator component inside the CAOC is the MOSFET Q1. During periods of

high battery current, the power dissipated by Q1 can reach 100 W. Referring to figure 48, the

takeoff portion of the flight profile is where Q1 dissipates the most power. In order to keep Q1

internal junction temperature below its rated maximum (usually 150oC), a heat sink must be used.

A future research opportunity is to design a heat sink for Q1 and evaluate Q1 junction

temperature during flight with extra attention focused on the takeoff period. A possible location for

the heat sink is to place it on the upper surface of the UAV fuselage; the outer surface of the heat

sink being flush with the outer surface of the fuselage and the inner surface of the heat sink

mated to transistor Q1. UAV motion through the air creates an airflow over the heat sink which

enhances the cooling capacity of the heat sink. During takeoff, the UAV starts at zero velocity and

gains speed as it proceeds down the runway. This period of time is where the heat sink must

dissipate the most power but the airflow over the heat sink is minimum. A finite element analysis

(FEA) thermal analysis of the power dissipation profile (figure 48) for Q1 combined with a minimal

or increasing airflow over the heatsink can simulate Q1 junction temperature during the takeoff

period. Another scenario where heat sink analysis may be necessary is during cruise when

current assist is active. Although the power dissipated by Q1 is much lower, around 1.3 W, it is

constant for a prolonged period of time (hours).

To improve hydrogen storage capacity onboard the UAV, liquid hydrogen storage can be

used in place of compressed hydrogen storage tanks. Liquid hydrogen (LH2) is an attractive

alternative to compressed hydrogen storage due to its high volumetric density of 70.8 g/L

(Khandelwal et al., 2013). When compared to the volumetric energy density of hydrogen gas at

STP (9.829 MJ/L), LH2 has 864.3 times the volumetric energy density (8495 MJ/L). The clear

advantage of liquid hydrogen is balanced by the technical difficulties of storage and the highly

energy intensive process of liquefying hydrogen. The source of complication to store liquid

hydrogen is the inherent temperature and pressure range at which liquid hydrogen exists. Typical

LH2 storage tanks are kept at a temperature of 20.5oK and at a pressure of 1 to 3 bars

112
(Khandelwal et al., 2013). To maintain its liquid state, the LH2 must be kept below the critical point

of 33oK and 12.9 bar and above its triple point of 13.8oK and 0.0719 bar (“Properties of

Hydrogen”, 2016).

Figure 58. Hydrogen Phase Diagram (https://hydropole.ch)

A cryogenic pressure vessel named a cryostat is a temperature controlled storage tank used to

contain LH2. Designed as a dual shell system, the inner and outer metallic shells are separated

by a sandwiched layer of insulation. Careful thermal design must account for thermal conduction,

convection, and radiation in order to maintain the extremely low storage temperature. The inner

shell contains the cryogenic hydrogen and uses a multi-layered material as an insulator wrap

(Tzimas, 2003). The insulation wrap consists of a reflective foil on the inner surface which reflects

radiative thermal transfer, and then an alternating stack of thin insulating material such as a glass

fiber and thin layers of aluminum (Khandelwal et al., 2013). Mechanical spacers are placed

between the inner and outer shells to maintain a controlled gap between the two vessels. The

gap is then evacuated to a near perfect vacuum to minimize thermal convection, thus maximizing

the thermal insulation. Despite the extreme measures to create a perfectly insulating pressure

vessel, there is still heat leakage. Additionally, the orthohydrogen content is never zero in

cryogenic hydrogen, and when it changes to parahydrogen, heat is generated internally. Due to

heat effects, there is always a certain amount of LH2 that “boils off”. The boil off rate will

determine the effective storage time of a cryostat. An acceptable boil off rate for aviation

113
applications is 0.1% of LH2 by weight per hour (Khandelwal et al., 2013). An example LH2 storage

system for an automotive application has a gravimetric density of 14.2% (Larminie & Dicks,

2003). Despite the complicated implementation of a liquid hydrogen storage system, the NRL

team demonstrated its use in a UAV application with the 2013 flight of the Ion Tiger (“Ion Tiger

Fuel Cell Powered UAV”, 2016). A second notable example of a UAV powered by LH2 fuel is the

Global Observer built by AeroVironment. Designed as an ISR platform, the Global Observer flew

for a 24 hour period in 2005 as a demonstration of its capabilities (Moffit, 2010).

5.3 Conclusion

Interest in sUAS has been on the rise as the costs decrease and civilian applications and

opportunities expand. Within the aerospace community, the sUAS sector has been recognized as

the most dynamic new market with sales expected to exceed 8 billion US dollars by 2018

(Manufacturing Close-Up, 2015). Continued research in the area of electric powered UAVs has

made steady progress, increasing UAV capabilities such as flight endurance. To bring UAV flight

endurance to the next level, a key and disruptive technology is the use of fuel cells as an electric

power source. Integrating fuel cell technology into sUAS continues to challenge engineering

practice as fuel cell weight and limited power delivery are barriers to overcome. Combining

conventional lithium ion batteries with PEMFCs into a hybrid power system has the advantage of

minimizing fuel cell weight while still providing high power when needed. Many APM schemes

reported in literature optimize hybrid power system performance through the use of DC-to-DC

converters between power sources. The issue with this approach is that the DC-to-DC converters

are in the main power path and therefore require high power ratings. High power DC-to-DC

converters add significant weight to the power system. This study adds to APM literature with a

new design approach that uses a transistor in place of the DC-to-DC converter. The transistor

weight is much lower than the weight of a high power DC-to-DC converter thus reducing weight of

the power system. In addition, significant power is dissipated by the transistor only during high

power delivery from the battery. During the majority of the flight, power dissipated by the

transistor can drop to zero. Over the duration of the flight, power system efficiency of the
114
transistor APM approach exceeds 98% which is greater than typical efficiencies achieved with

DC-to-DC converters.

The Simscape model of the UAV power system developed in this study demonstrates

several points. First, it verifies the functionality of the transistor-based APM approach. Second, it

implements the Tremblay battery model as a Simscape model and verifies model behavior in a

larger system. Similarly, the Lee fuel cell model is implemented as a Simscape model and verified

in system level simulations. Third, the flight simulations determine the amount of hydrogen

required for the ASU Composite UAV to achieve a six hour flight time. Fourth, the flight

simulations with current assist activated demonstrate that battery power can be used during

cruise flight to enhance flight endurance or reduce the amount of hydrogen necessary to achieve

a flight time goal. Fifth, the UAV power system model demonstrates inflight battery recharge

which can be useful for certain scenarios. Sixth, implementing the CAOC in the HPMC

demonstrates that monitoring fuel cell and battery voltage and current allows a microprocessor to

implement an observer algorithm that can closely track battery SOC, fuel cell efficiency, and

hydrogen consumption. The final point demonstrated by the UAV power system simulations are

the potential flight times that can be expected for various hybrid power system configurations.

Overall, the UAV hybrid power system model proved useful in many areas.

115
 REFERENCES

AE3007 (n.d.). Retrieved June 18, 2016, from http://www.rolls-royce.com:

http://www.rolls-royce.com/products-and-services/defence-aerospace/products/uav/ae-
3007.aspx#engine-specifications.

Altair-UAV (n.d.). Retrieved Jun 17, 2016, from http://www.theuav.com:

http://www.theuav.com/altair_uav.html.

AXI Motor Applications 4120/14 (2015), Retrieved November 24, 2015, from
http://www.modelmotors.cz:
http://www.modelmotors.cz/index.php?page=61&product=4120&serie=14&line=GOLD.

Bard, A., Faulkner, L. (2001). Electrochemical methods fundamentals and applications. Wiley.

Beswick Pressure Regulator (2016). Retrieved November 5, 2016, from

http://catalog.beswick.com:
http://catalog.beswick.com/#WORKFLOW;PRODUCT,PR-MLS,FILTER_REGULATORS

Bradley, T. (2008). Modeling, design and energy management of fuel cell systems for aircraft.
Georgia Institute of Technology Dissertation.

Buasri, P., & Salameh, Z. (June, 2006). An electrical circuit model for a proton exchange
membrane fuel cell (PEMFC). Power Engineering Society General Meeting, IEEE 2006.

Current Sharing in Power Arrays (n.d.). Maxi, Mini, Micro Design Guide, rev. 4.9, pp. 20 – 24,
Retrieved September 17, 2016, from www.vicorpower.com
http://www.vicorpower.com/documents/applications_manual/05current_sharing.pdf.

DOD dictionary (2007). Dictionary of Military and Associated Terms. JP 1-02, 569.

DOD Unmanned Systems Roadmap, 2007-2032.

DOE FCTO (2016, May). Hydrogen fuel cells for small unmanned air vehicles. Retrieved
September 11, 2016 from http://energy.gov:
http://energy.gov/sites/prod/files/2016/05/f32/fcto_webinarslides_h2_fc_small_unmanned_air_veh
icles_052616.pdf.

EnergyOr fuel cell powered UAV reaches 10 h flight endurance (2011, September). Fuel Cell
Bulletin, 2011, pp. 4-5.

FAUCON H2 UAV image (2016). Retrieved September 16, 2016, from http://energyor.com:
http://energyor.com/static/template/eo/images/products/h2-endurance/MB-99-l.jpg.

Fuel cell/battery hybrid UAV takes off in Taiwan (2010, June). Fuel Cell Bulletin, 2010, pp. 4-5.

Fuentes, G. (2005, April 4) DARPA Test-Flies 7-Ounce UAV. Defense News. p. 26.

Furrutter, M., Meyer, J. (2009, September). Small fuel cell powering an unmanned aerial vehicle.
AFRICON 2009, pp. 1-6.

General Atomics MQ-1 Predator (June 9, 2016). Retrieved June 18, 2016 from
https://en.wikipedia.org:
https://en.wikipedia.org/wiki/General_Atomics_MQ-1_Predator.

116
Gibbs, Y., (July 30, 2015). NASA Armstrong fact sheet: Altair. Retrieved from
http://www.nasa.gov:
http://www.nasa.gov/centers/armstrong/news/FactSheets/FS-073-DFRC.html.

Grey-faced Buzzard UAV image (2016), Retrieved June 16, 2016, from http://cdn.ubergizmo.com:
http://cdn.ubergizmo.com/photos/2010/5/asia-hybrid-uav.jpg.

Gundlach, J. (2014). Designing unmanned aircraft systems, a comprehensive approach, (pp. 14-
22). AIAA.

Hornet UAV image. Retrieved September 10, 2016, from https://www.avinc.com:

https://www.avinc.com/images/uploads/general/7/hornet2-1_bg.jpg.

Horizon (August, 2013). H-100 fuel cell stack user manual. Retrieved February 10, 2016, from
http://media.wix.com:
http://media.wix.com/ugd/047f54_bc938ab51966e498e9919dfebb88e09d.pdf
Hydrogen phase diagram image (2016). Retrieved September 23, 2016, from
https://hydropole.ch:
https://hydropole.ch/wp-content/uploads/2014/12/1372142909_phasediag.gif.

Interstate TDR (May 18, 2016). Retrieved June 17, 2016, from https://en.wikipedia.org:
https://en.wikipedia.org/wiki/Interstate_TDR.

Ion Tiger Fuel Cell Powered UAV (2016). Retrieved September 10, 2016, from
http://www.nrl.navy.mil:
http://www.nrl.navy.mil/lasr/content/ion-tiger-fuel-cell-powered-uav.

Ion Tiger UAV image. Retrieved September 11, 2016, from http://www.nrl.navy.mil:
http://www.nrl.navy.mil/PressReleases/2010/image3_3-10r.jpg.

Jefferson, M. (2002, December 17) DARPA to seek long endurance with ‘Hornet’ micro air
vehicle. Aerospace Daily.

Jiang, Z., Gao, L., Blackwelder, M., & Dougal, R. (December, 2003). Design and experimental
tests of control strategies for active hybrid fuel cell/battery power sources. Journal of Power
Sources, 2003 (130), pp. 173 – 171.

Jiang, Z. Gao, L., & Dougal, R. (June, 2007). Adaptive control strategy for active power sharing in
hybrid fuel cell/battery power sources. IEEE Transactions on Energy Conversion, 2007 (22), pp.
507 – 515.

Jones, P., Lakeman, J., Mepsted, G., & Moore, J. (December, 1998). A hybrid power source for
pulse power applications. Journal of Power Sources, 1998 (80), pp. 242 – 247.

Khandelwal, B., Karakurt, A., Sekaran, P., Sethi, V., & Singh, R. (January, 2013). Hydrogen
powered aircraft : The future of air transport. Progress in Aerospace Sciences, (60) 2013.

Keane, J., & Carr, S. (2013). A brief history of early unmanned aircraft. Johns Hopkins APL
Technical Digest, 32(1), 558-571. Retrieved June 17, 2016, from http://www.jhuapl.edu/:
http://www.jhuapl.edu/techdigest/TD/td3203/32_03-Keane.pdf.

Krauch, T., Berg, S., Beckstead, A., Wilkinson, D., Coronel, M., Kastelic, G., Padron, N., & Vo, D.
(December, 2012). Design, construction, and testing of a composite multi-role UAV load bearing
aircraft. Arizona State University Project Report.

117
Langley Aerodrome Number 6 (n.d.). Retrieved June 17, 2016, from http://airandspace.si.edu:
http://airandspace.si.edu/collections/artifact.cfm?object=nasm_A19050002000.

Larminie, J., & Dicks, A. (2003). Fuel cell systems explained, second edition. West Sussex: J.
Wiley & Sons.

Lee, B. (2013). TDR-1 First Operational US Navy Drone Successful in Combat in 1944. Retrieved
June 18, 2016, from http://www.nnapprentice.com/
http://www.nnapprentice.com/alumni/letter/TDR_1.pdf.

Lee, B., & Park, P. (October, 2014). Active power management system for an unmanned aerial
vehicle powered by solar cells, fuel cell, and batteries. IEEE Transactions on Aerospace and
Electronic Systems, 2014 (50), pp. 3167 – 3177.

Lee, D., & Wang, L. (June, 2007). Dynamic and steady-state performance of PEM fuel cells under
various loading conditions. IEEE Power Engineering Society General Meeting.

Lockheed Martin ruggedized UAS uses AMI fuel cell power (September, 2011). Fuel Cell Bulletin,
2011 (9), p. 4.

Linden, D., & Reddy, T. (2002). Handbook of batteries, third cdition. McGraw-Hill.

Lithium polymer battery (December, 2010). Retrieved September 8, 2016, from

https://en.wikipedia.org:
https://en.wikipedia.org/wiki/Lithium_polymer_battery.

Lycoming O-435 (September 25, 2015). Retrieved June 17, 2016, from https://en.wikipedia.org:
https://en.wikipedia.org/wiki/Lycoming_O-435.

Madakannan, A. (2014). Lecture Notes (fall, 2014).

Madakannan, A. (2015). Lecture Notes (fall, 2015).

Madakannan, A. (2016). Lecture Notes (spring, 2016).

Manufacturing Close-Up (Jan. 8, 2015). ABI research: Commercial sector is sweet spot for sUAS
market. Close-Up Media, Inc.

MaxAmps LiPo 4S 8Ah (September, 2016). Retrieved September 10, 2016, from
www.maxamps.com:
https://www.maxamps.com/lipo-8000-4s-14-8v-dual-core-battery-pack.

MaxAmps LiPo 4S Family (October, 2016). Retrieved October 9, 2016, from www.maxamps.com:
https://www.maxamps.com/products/4s-14-8v-lipos.

McConnell, V. (2007, December) Military UAVs claiming the skies with fuel cell power. Fuel Cell
Bulletin. pp. 12-15.

Moffit (2010). A methodology for the validated design space exploration of fuel cell powered
unmanned aerial vehicles. Georgia Institute of Technology Dissertation.

Monaco, E. (2013). Optimization of flight time using fuel cell on unmanned aerial vehicles.
Arizona State University Project Report.

MQ-1B Predator (September 23, 2015). Retrieved June 18, 2016, from http://www.af.mil/:
118
http://www.af.mil/AboutUs/FactSheets/Display/tabid/224/Article/104469/mq-1b-predator.aspx.

MQ-1 Predator/MQ-9 Reaper (2016). Retrieved June 17, 2016 from http://www.bga-
aeroweb.com:
http://www.bga-aeroweb.com/Defense/MQ-1-Predator-MQ-9-Reaper.html.

Nam, C., (2015). Lecture Notes (spring, 2015).

NASA (August, 2006). Earth observations and the role of UAVs: A capabilities assessment,
version 1.1, (p. 12). Retrieved June 16, 2016, from http://www.nasa.gov:
http://www.nasa.gov/centers/dryden/research/civuav/index.html.

Ninja Pro v2 90/4500 tank image (2016). Retrieved September 22, 2016, from
www.ninjapaintball.com:
http://www.ninjapaintball.com/tanks?lightbox=image_1dk3.

Ninja Pro v2 90/4500 specs (2016). Retrieved September 22, 2015, from www.evike.com:
http://www.evike.com/products/50046/.

Northrop Grumman Facts (May, 2008). Retrieved June 17, 2016 from
http://www.northropgrumman.com:
http://www.northropgrumman.com/capabilities/rq4block20globalhawk/documents/hale_factsheet.
pdf.

Northrop Grumman RQ-4 Global Hawk (June 11, 2016). Retrieved June 17, 2016, from
https://en.wikipedia.org:
https://en.wikipedia.org/wiki/Northrop_Grumman_RQ-4_Global_Hawk.

NRL History (2016). Retrieved September 11, 2016 from http://www.nrl.navy.mil:

http://www.nrl.navy.mil/about-nrl/history/edison/.

NRL Fact Book (2014). Retrieved September 11, 2016, from http://www.nrl.navy.mil:
http://www.nrl.navy.mil/content_images/2014_FactBook_cleansed.pdf.

Piper Cub UAV image (2016). Retrieved September 16, 2016, from www.google.com:
https://www.google.com/search?q=University+of+Johannesburg+Fuel+Cell+UAV+image&biw=11
73&bih=629&tbm=isch&imgil=CrGj8aTkfFlw_M%253A%253BjpZaRc4C3jXF5M%253Bhttp%252
53A%25252F%25252Fwww.unmannedsystemstechnology.com%25252F2015%25252F08%2525
2Famerican-university-of-sharjah-conducts-first-hydrogen-fuel-cell-powered-uav-flight-in-
uae%25252F&source=iu&pf=m&fir=CrGj8aTkfFlw_M%253A%252CjpZaRc4C3jXF5M%252C_&u
sg=__JWg-
_PRowNXbAPMgIbBGdR5D59g%3D&ved=0ahUKEwiojv2a5pTPAhVD7GMKHYMCBSQQyjcINQ
&ei=tF_cV6jkO8PYjwODhZSgAg#imgrc=CrGj8aTkfFlw_M%3A.

Properties of Hydrogen (2016), Brookhaven National Laboratory, Retrieved on September 23,

2016, from www.bnl.gov:
https://www.bnl.gov/magnets/staff/gupta/cryogenic-data-handbook/Section3.pdf.

Qingshan, X., Nianchun, W., Ichiyanagi, K. & Yukita, K. (April, 2008). PEM fuel cell modeling and
parameter influences of performance evaluation. Electric Utility Deregulation and Restructuring
and Power Technologies, Third International Conference.

Raymer, D. (1989). Aircraft design: A conceptual approach. AIAA, p. 79.

Revankar, S., & Majumdar, P. (2014). Fuel cells principles, design, and analysis. CRC Press.
119
Sadraey, M. (2013). Aircraft design, a systems engineering approach. Wiley, p.127.

Shepherd, C. (1965, July). Design of primary and secondary cells II. An equation describing
battery discharge. Journal of Electrochemical Society, 1965 (112), pp.657-664.

Simscape User’s Guide, R2014a (Mathworks, 2014). Mathworks Inc., Natick.

Simscape Language Guide, R2014b (Mathworks, 2014). Mathworks Inc., Natick.

Spider Lion UAV image (2016). Retrieved September 10, 2016, from https://www.google.com:
https://www.google.com/search?q=Spider+Lion+UAV+image&biw=1261&bih=686&tbm=isch&img
il=Zmf3bIJ5rti8qM%253A%253B45l3yXl_LjbCGM%253Bhttp%25253A%25252F%25252Fwww.d
esignation-systems.net%25252Fdusrm%25252Fapp4%25252Fspider-
lion.html&source=iu&pf=m&fir=Zmf3bIJ5rti8qM%253A%252C45l3yXl_LjbCGM%252C_&usg=__2
F0scuC-t9v5fQsNBKu2edawbgQ%3D&dpr=1.2&ved=0ahUKEwiwqe2w1ofPAhVN-
2MKHeUpANEQyjcIOg&ei=R37VV_DpEc32jwPl04CIDQ#imgrc=Zmf3bIJ5rti8qM%3A.

Stalker UAS (2016). Retrieved September 15, 2016, from http://www.lockheedmartin.com:

http://www.lockheedmartin.com/us/products/stalker-uas.html.

Stalker UAV image (2016). Retrieved September 16, 2016, from http://www.lockheedmartin.com:
http://www.lockheedmartin.com/us/products/stalker-uas.html.

Suh, K. (2006). Modeling, analysis, and control of fuel cell hybrid power systems. University of
Michigan Dissertation.

Swider-Lyons, K., MacKrell, J., Rodgers, J., Page, G., Schuette, M., & Stroman, R. (September,
2011). Hydrogen fuel cell propulsion for long endurance small UAVs. AIAA centennial of Naval
Aviation Forum “100 Years of Achievement and Progress”, AIAA 2011-6975.

Swider-Lyons, K. (May, 2016). Hydrogen fuel cells for small unmanned air vehicles. DOE
presentation May 26, 2016, Retrieved September 23, 2016, from http://energy.gov:
http://energy.gov/sites/prod/files/2016/05/f32/fcto_webinarslides_h2_fc_small_unmanned_air_veh
icles_052616.pdf.

Targets for Onboard Hydrogen Storage Systems for Light-Duty Vehicles (September, 2009).
Freedom Car Fuel Partnership, DOE.

Tawfik, H., Hung, Y., & Mahajan, D. (September, 2006). Metal bipolar plates for PEM fuel cell – A
review. Journal of Power Sources, 2006 (doi:10.1016/j.jpowsour.2006.09.088).

TPE331-10 Turboprop Engine (April, 2006). Retrieved June 17, 2016, from
http://www51.honeywell.com:
http://www51.honeywell.com/aero/common/documents/myaerospacecatalog-
documents/BA_brochures-documents/TPE331.10.pdf.

Tremblay, O., & Dessaint, L. (May, 2009). Experimental validation of a battery dynamic model for
EV applications. World Electric Vehicle Journal, 2009 (3), pp. 289-299.

Tsang, K., Sun, L., & Chan, W. (June, 2010). Identification and modelling of lithium ion battery.
Energy Conversion and Management, 2010 (51), pp. 2857-2862.

120
Tzimas, E., Filiou, C., Peteves, S., & Veyret, J. (2003). Hydrogen storage: State-of-the-art and
future perspective. Institute for Energy, Petten, the Netherlands. Retrieved September 23, 2016,
from www.Linde.com:
http://publications.jrc.ec.europa.eu/repository/bitstream/111111111/6013/1/EUR%2020995%20E
N.pdf.

USAF contract for Protonex UAV power (March, 2006). Fuel Cell Bulletin, p. 3.

Wright Brothers (n.d.). Retrieved June 17, 2016, from http://airandspace.si.edu:

http://airandspace.si.edu/explore-and-learn/topics/wright-brothers.cfm.

Zhang, J., Li J., Li, Y., Zhao, Y. (2014). Hydrogen generation, storage, and utilization. Wiley, pp.
75-80.

121
 APPENDIX A

SIMSCAPE CODE FOR LITHIUM ION BATTERY COMPONENT

122
component LiPo_4S
% Battery model of LiPo 4S, 8 Ah

nodes
p = foundation.electrical.electrical; % +:right
n = foundation.electrical.electrical; % -:right
end

outputs
SOC = { 0, '1'}; % SOC:left
BatteryVoltage = { 0, 'V' }; %V:left
BatteryCurrent = { 0, 'A' }; %I:left
end

variables(Access = protected)
v = {0, 'V'}; % Across variable, voltage
i = {0, 'A'}; % Through variable, current
CT = {0, 'c'}; % Charge Transfer, coulombs
end

parameters
OCV = { 15.6, 'V' }; % Battery Open Circuit Voltage (volts)
Q = { 8, 'A*hr' }; % Battery Capacity (Amp-hours)
IntRes = { 0.005, 'Ohm' }; % Internal Series Resistance (Ohms)
InitialSOC = { 100.0, '1' }; % Initial State of Charge (percentage)
K = { 0.0075, 'V/(A*hr)'}; % Polarization Constant (V/(Amp-hours))
Kr = {0.0064, 'Ohm'}; % Polarization Resistance (Ohm)
A = { 1.2, 'V' }; % Exponential Zone amplitude (V)
B = { 2.0, '1/(A*hr)' }; % Exponential Zone time constant inverse (1/(Amp-hours))
end

function setup
if OCV <= 0
error('OCV must be greater than zero');
end
if Q <= 0
error('Battery capacity must be greater than zero');
end
% From the Initial SOC and the Capacity (Q), determine
% the starting value for CT (charge transferred)
CT = Q*{3600,'c/(A*hr)'}*(1 - (InitialSOC/100));
end

branches
i: n.i -> p.i; % define current flow (source: neg to pos)
end

equations
v == p.v - n.v; % Voltage across battery (pos to neg)
let
ohmic = i * IntRes;
ExpZone = A*exp(-B*(CT));
PolVolt = K*(Q/(Q - CT)) * CT;
PolResDischarge = Kr*(Q/(Q - CT)) * i;
PolResCharge = Kr*(Q/(CT - Q)) * i;
123
 in % Battery equation
if i >= 0
v == OCV + ExpZone - PolVolt - ohmic - PolResDischarge;
else
v == OCV + ExpZone - PolVolt - ohmic - PolResCharge;
end
end
i == CT.der; % Current is derivative of Charge Transferred
SOC == 100.0*(1 - (CT/Q)); % State of Charge
BatteryVoltage == v; % export the battery voltage
BatteryCurrent == i; % export the battery current
end
end

124
 APPENDIX B

LIPO BATTERY DISCHARGE CURVES AUTOMATION SCRIPT

125
% Discharge_4S_curves.m
% LiPo_4S battery discharge curve graph
%
% Plot the discharge curves for the LiPo_4S, 8 Ah battery for
% 0.5C, 1.0C, 2.0C, 4.0C, 8.0C 16.0 C
DR = [0.5 1.0 2.0 4.0 8.0 16.0]; % battery discharge rate vector
tmax = zeros(1,length(DR)); % array of maximum simulation times
dlc = [0.0 0.1 0.9; 0.7 0.4 0.2; 0.1 0.9 0.1; 0.4 0.1 0.1; 0.9 0.0 0.0; ...
0.0 0.0 0.0];
BatCap = 8.0; % battery capacity in Amp-hours
DC = BatCap * DR; % discharge currents
for j = 1:length(DR)
csp = 1.7 + 2*DC(j); % current setpoint
tmax(j) = (BatCap / DR(j))*3600; % stop time for simulation
sim('LiPo_Discharge_01', 'StopTime', num2str(tmax(j))); % run the model
xvals = simlog.C2Ah.O.series.values;
yvals = simlog.Battery_model_of_LiPo_4S_8_Ah.BatteryVoltage.series.values;
plot(xvals, yvals, 'LineWidth', 2, 'Color', [dlc(j,1) dlc(j,2) dlc(j,3)]);
if j == 1
hold on;
end
end
hold off;
title('LiPo 4S, 8 Ah, Battery Discharge Curves');
ylabel('Battery Voltage (V)');
xlabel('Charge (Ah)');
legend('0.5 C', '1.0 C', '2.0 C', '4.0 C', '8.0 C', '16.0 C', ...
'Location', 'southwest');
text(0.1, 6.3, 'Legend: Discharge Rate');
set(gca, 'FontSize', 14);
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

126
 APPENDIX C

BATTERY DISCHARGE COMPARISON MODEL

127
Figure 59. Battery Discharge Comparison Model

128
 APPENDIX D

SIMSCAPE CODE FOR NERNST POTENTIAL SUBCOMPONENT

129
component Enernst
% Nernst Voltage

% revesible cell voltage

nodes
p = foundation.electrical.electrical; % +:top
n = foundation.electrical.electrical; % -:bottom
end

inputs
H2_pressure = { 1, 'atm' }; % H2p:bottom
Air_pressure = { 1, 'atm' }; % Airp:bottom
end

outputs
Vrev = { 0, 'V' }; % Vr:top
end

parameters
T = { 65 + 273.15, 'K' }; % cell temperature (K)
noc = { 12, '1' }; % number of cells
end

parameters(Hidden = true)
Tref = { 25 + 273.15, 'K' }; % reference temperature
end

variables(Access = protected)
v = {0, 'V'}; % Across variable, voltage
i = {0, 'A'}; % Through variable, current
end

function setup
if T < {(60 + 273.15), 'K'} || T > {(100 + 273.15), 'K'}
error('Temperature must be between 60 and 100 Celsius');
end
if noc < 1
error('number of cells must be greater than one');
end
Tref = { 25 + 273.15, 'K' };
end

branches
i: p.i -> n.i; % define current flow (source: neg to pos)
end

equations
let
Eo = {1.229,'V'} - {0.85e-3,'V/K'}*(T - Tref) + ...
{4.31e-5,'V/K'}*T*(log(H2_pressure * {1,'1/atm'}) + ...
0.5*log(Air_pressure * {0.2095,'1/atm'}));
in
v == 0.8 * noc * Eo;
end
v == p.v - n.v; % Voltage across battery (pos to neg)
130
 Vrev == v;
end
end

131
 APPENDIX E

MATLAB CALCULATION OF LEE MODEL ERROR OF NERNST POTENTIAL

132
% pe3D_Lee_Nernst.m
% Percentage error of Lee approximation of Nernst potential
% x-axis: temperature (degree Celsius), range: 25 - 200
% y-axis: hydrogen partial pressure (atm.), range: 0.5 - 3
% z-axis: ((Estd - Elee)/Estd)*100%

% constants
F = 98485; % Faraday's constant (Coulombs / mole)
R = 8.314; % Universal gas constant (J/(mole*K))
% color matrix
cm = [5 65 0; 255 69 0; 255 0 0; 218 165 32; 189 183 140;
154 205 50; 127 255 0; 0 100 0; 102 205 160; 32 178 170;
95 158 160; 0 0 205; 0 255 250; 128 0 0; 128 0 128;
0 128 128; 180 180 192; 178 34 34; 255 215 3; 250 140 5];
% normalize color matrix
cm = (1/255)*cm;
% define a function handle for interpolated data from Gibb's table
fh_gfe = @(g)interp1([25 80 100 200],[-237.2 -226.1 -225.2 -220.4], g);

% create the x-y grid (20 x 20)

xlin = linspace(25, 200, 20); % temperature values
ylin = linspace(0.5, 3.0, 20); % pressure values
[x,y] = meshgrid(xlin, ylin); % independent variable grid
ulin = xlin;
vlin = linspace(1.5, 2.5, 7);
[u,v] = meshgrid(ulin,vlin);

% pre-allocate memory
Elee = zeros(20,20); % Lee approximation of Nernst potential
Estd = zeros(20,20); % Standard calculation of Nernst potential
term_1 = zeros(20,20); % first term in standard calculation
term_2 = zeros(20,20); % second term in standard calculation
zero_plane = zeros(7,20); % zero reference plane
z = zeros(20,20); % pre-allocate percent error matrix

for i = 1:20
for j = 1:20
Elee(i,j) = 1.229 - 0.85*10^-3 * (xlin(i) - 25) + ...
4.3085*10^-5 * (xlin(i) + 273.15)*(log(ylin(j)));
term_1(i,j) = ( (-1*fh_gfe(xlin(i)))/(2*F) )*1000;
term_2(i,j) = ((R*(xlin(i)+273.15))/(2*F))*(log(ylin(j)));
Estd(i,j) = term_1(i,j) + term_2(i,j);
z(i,j) = ((Estd(i,j) - Elee(i,j))/Estd(i,j))*100;
end
end

% plot the results

figure
subplot(1,2,1)
colormap hsv
surf(x,y,z,'FaceColor','interp','EdgeColor', 'none', ...
'FaceLighting', 'gouraud');
axis tight; view(-58,32);
hold on;
mesh(u,v,zero_plane,'FaceLighting','gouraud','LineWidth',0.3);
133
 hold off;
title('Lee approximation of Nernst potential percent error');
xlabel('Temperature (C)'); ylabel('Hydrogen Pressure (atm)');
zlabel('Error (%)');
set(gca, 'FontSize', 14);
subplot(1,2,2);
for k = 1:20
plot(xlin, z(k,:), 'LineWidth', 2, 'Color', cm(k,:));
if k == 1
hold on;
end
if k < 9
tx = text(20+k*18, z(k,5)+0.1, num2str(ylin(k)));
else
if mod(k,2)
text(100, z(k,5)+0.1, num2str(ylin(k)));
else
text(60, z(k,5)+0.1, num2str(ylin(k)));
end
end
end
hold off; title('Lee approximation of Nernst potential percent error');
xlabel('Temperature (C)'); ylabel('Error (%)');
hl = legend('H2 Pressure = 0.500', 'H2 Pressure = 0.632', ...
'H2 Pressure = 0.763', 'H2 Pressure = 0.895', 'H2 Pressure = 1.026', ...
'H2 Pressure = 1.158', 'H2 Pressure = 1.290', 'H2 Pressure = 1.421', ...
'H2 Pressure = 1.553', 'H2 Pressure = 1.684', 'H2 Pressure = 1.812', ...
'H2 Pressure = 1.947', 'H2 Pressure = 2.079', 'H2 Pressure = 2.211', ...
'H2 Pressure = 2.342', 'H2 Pressure = 2.474', 'H2 Pressure = 2.605', ...
'H2 Pressure = 2.737', 'H2 Pressure = 2.868', 'H2 Pressure = 3.000', ...
'Location', [0.86 0.1 0.1 0.92]);
for k = 1:8 % reprint numbers so they are on top of graph lines
tx = text(20+k*18, z(k,5)+0.1, num2str(ylin(k)));
end
set(gca, 'FontSize', 14);
grid;
set(findall(fig,'-property','FontSize'),'FontSize',18);
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

134
 APPENDIX F

SIMSCAPE CODE FOR ACTIVATION RESISTOR SUBCOMPONENT

135
component Ractivation
% R activation

% effective resistance of activation overpotential

nodes
p = foundation.electrical.electrical; % +:left
n = foundation.electrical.electrical; % -:right
end

inputs
H2_pressure = { 1, 'atm' }; % H2p:left
Air_pressure = { 1, 'atm' }; % Airp:left
end

outputs
effective_Res = {0, 'Ohm'}; % eRa:right
end

variables(Access = private)
i = { 0, 'A' }; % Current
v = { 0, 'V' }; % Voltage
end

parameters
Area = { 22.25, 'cm^2' } % cell active area
T = { 65 + 273.15, 'K' }; % cell temperature (K)
noc = { 20, '1' }; % number of cells
end

parameters(Hidden = true)
z1 = { -1.81, 'V' } % zeta_1 coefficient
z2 = { 0.00286, 'V/K' } % zeta_2 coefficient
z3 = { 7.6e-5, 'V/K' } % zeta_3 coefficient
z4 = { -0.78e-4, 'V/K' } % zeta_4 coefficient
end

function setup
if Area < {0.1, 'cm^2'}
error('cell active area must be greater than 0.1 cm^2');
end
if T < {(60 + 273.15), 'K'} || T > {(100 + 273.15), 'K'}
error('Temperature must be between 60 and 100 Celsius');
end
% caculate z2
z2 = {0.00286, 'V/K'} + {0.0002,'V/K'}*log(Area*{1,'1/cm^2'});
end

branches
i : p.i -> n.i;
end

equations
let
% concentration of H2 at the catalytic interface (mole/cm^3)
% note: make CH2 unitless ('1')
136
 CH2 = (H2_pressure * {1, '1/atm'}) / ...
(5.08e-6*exp(-498/(T*{1,'1/K'})));
in
let
% concentration of O2 at the catalytic interface (mole/cm^3)
% note: make CO2 unitless ('1')
CO2 = (0.2095 * Air_pressure * {1, '1/atm'}) / ...
(5.08e-6*exp(-498/(T*{1,'1/atm'})));
z2 = ({0.00286, 'V/K'} + ...
({0.0002,'V/K'}*log(Area*{1,'1/cm^2'})) + ...
({5.3e-5,'V/K'}*log(CH2)));
in
if i < {1e-29, 'A'}
v == -1*noc*((z1 + z2*T + z3*T*log(CO2) + ...
z4*T*log(1e-29)));
else
v == -1*noc*(z1 + z2*T + z3*T*log(CO2) + ...
z4*T*log(i));
end
end
end
% protect against division by zero
if i < {1e-12,'A'}
effective_Res == v / {1e-12, 'A'};
else
effective_Res == v/i;
end

v == p.v - n.v;
end
end

137
 APPENDIX G

MATLAB CODE TO CALCULATE CELL RESISTANCE

138
% cell_resistance_example.m
%
% Calculate the cell resistance for a PEMFC MEA over a temperature
% range of 40C to 90C and a current density range of 0.1 A/cm^2 to
% 1 A/cm^2. Plot results as a surface plot for each value of psi.
% Electronic resistance of cell, Rc = 0.02 ohms
% Active area, A = 20 cm^2
% Membrane thickness, t = 0.0051 cm. = 51um
Rc = 0.02; A = 20; t = 0.0051;
% create x-y grid (50 x 50)
xlin = linspace(40, 90, 50); % temperatures
ylin = linspace(0.1, 1.0, 50); % current densities
ulin = linspace(90, 40, 50);
vlin = linspace(1.0, 0.1, 50);
[x,y] = meshgrid(ulin, vlin);
% pre-allocate memory
z = zeros(50,50); % cell resistance
rho = zeros(50,50); % specific resistivity of Nafion membrane
num = zeros(50,50); % numerator
den = zeros(50,50); % denominator
% parameter psi (range 14 - 23)
psi_base = 13;
for w = 1:10
psi = psi_base + w;
for i = 1:50
for j = 1:50
Tk = xlin(i) + 273.15; % change temperature to Kelvin
cd = ylin(j); % current density

num(i,j) = 181.6*(1 + 0.03*cd + 0.062*(Tk/303)^2*cd^2.5);

den(i,j) = (psi - 0.634 - 3*cd)*exp(4.18*((Tk - 303)/Tk));
rho(i,j) = num(i,j)/den(i,j); % Nafion specific resistivity
z(i,j) = (rho(i,j)*t/A) + Rc;

end
end

% plot the results

colormap hsv
surf(x,y,z,'FaceColor','interp','EdgeColor', 'none', ...
'FaceLighting', 'gouraud');
if w == 1
hold on;
end
end
hold off; axis tight; view(-58,32);
title('Cell Resistance versus Temperature and Current Density');
xlabel('Temperature (C)'); ylabel('Current Density (A/cm^2)');
zlabel('Cell Resistance (Ohms)');
set(gca, 'FontSize', 14);
cbh = colorbar;
set(cbh, 'position', [0.91 0.38 0.03 0.4]);
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

139
 APPENDIX H

SIMSCAPE CODE FOR OHMIC RESISTOR SUBCOMPONENT

140
component Rohmic
% R ohmic

% effective resistance of ohmic losses

nodes
p = foundation.electrical.electrical; % +:left
n = foundation.electrical.electrical; % -:right
end

outputs
effective_Res = { 0, 'Ohm' }; % R:right
end

variables(Access = private)
i = { 0, 'A' }; % Current
v = { 0, 'V' }; % Voltage
end

parameters
Area = { 22.25, 'cm^2' }; % cell active area (cm^2)
T = { 65 + 273.15, 'K' }; % cell temperature (K)
l = { 0.051, 'cm' } % thickness of Nafion 112:2 (cm)
noc = { 20, '1' }; % number of cells
end

parameters(Hidden = true)
symbol_Psi = {23, '1'};
Rc = {0.04, 'Ohm'} % elecctronic cell resistance
end

function setup
if Area < {0.1, 'cm^2'}
error('cell active area must be greater than 0.1 cm^2');
end
if T < {(60 + 273.15), 'K'} || T > {(100 + 273.15), 'K'}
error('Temperature must be between 60 and 100 Celsius');
end
Rc = {0.04, 'Ohm*cm^2'} / Area;
end

branches
i : p.i -> n.i;
end

equations
let
Alpha = (((T*{1,'1/K'})/303)^2) * ...
(((i*{1,'1/A'})/(Area*{1,'1/cm^2'}))^2.5);
Beta = symbol_Psi - 0.634 - ...
3*((i*{1,'1/A'})/(Area*{1,'1/cm^2'}));
in
let
num = 36*(1 + ...
0.03*((i*{1,'1/A'})/(Area*{1,'1/cm^2'})) + ...
0.062*Alpha) * {1, 'Ohm'};
141
 den = Beta*exp(4.18 * ...
(((T*{1,'1/K'}) - 303)/(T*{1,'1/K'}))) * {1, '1/cm'};
in
let
Rho = num / den; % specific resistivity (ohm*cm)
in
let
% ionic resistance of cell membrane
Rm = (Rho * l)/Area;
in
v == i*noc*(Rm + Rc);
effective_Res == noc*(Rm + Rc);
end
end
end
end
v == p.v - n.v;
end
end

142
 APPENDIX I

SIMSCAPE CODE FOR CONCENTRATION RESISTOR SUBCOMPONENT

143
component Rconcentration
% R concentration

% effective resistance due to concentration losses

nodes
p = foundation.electrical.electrical; % +:left
n = foundation.electrical.electrical; % -:right
end

outputs
effective_Res = { 0, 'Ohm' }; % R:right
end

variables(Access = private)
i = { 0, 'A' }; % Current
v = { 0, 'V' }; % Voltage
end

parameters
Area = { 22.25, 'cm^2' }; % cell active area (cm^2)
Jmax = { 0.43, 'A/cm^2' }; % max current density (A/cm^2)
noc = { 20, '1' }; % number of cells
B = { 0.076, 'V' }; % concentration loss coefficient
end

function setup
if Area < {0.1, 'cm^2'}
error('cell active area must be greater than 0.1 cm^2');
end
if Jmax < {0.01, 'A/cm^2'} || Jmax > {10, 'A/cm^2'}
error('Jmax must be between 0.01 A/cm^2 and 10 A/cm^2');
end
end

branches
i : p.i -> n.i;
end

equations
v == p.v - n.v;

v == (-noc*B)*log( 1 - ((i/Area)/Jmax) );

% protect against division by zero

if i < {1e-12,'A'}
effective_Res == v/{1e-12, 'A'};
else
effective_Res == v/i;
end
end
end

144
 APPENDIX J

SIMSCAPE CODE FOR DOUBLE LAYER CAPACITOR SUBCOMPONENT

145
component dlCap
% double-layer Capacitor

nodes
p = foundation.electrical.electrical; % p:top
n = foundation.electrical.electrical; % n:bottom
end

outputs
Vdl = {0, 'V'}; % Vdl:bottom
Idl = {0, 'A'}; % Idl:bottom
end

variables
i = { 0, 'A' }; % Current
v = { 0, 'V' }; % Voltage
end

parameters
C = { 1.0, 'uF' }; % Capacitance value
CIV = { 0.0, 'V' }; % initial voltage
end

function setup
if C <= 0
error('Capacitance Value must be positive');
end

% initial voltage to zero

v = { value = CIV, priority = priority.high };
end

branches
i : p.i -> n.i; % Through variable i from node p to node n
end

equations
v == p.v - n.v; % Across variable v from p to n
i == C*v.der; % Capacitor equation
Vdl == v;
Idl == i;
end
end

146
 APPENDIX K

SIMSCAPE CODE FOR FLOW RATE SUBCOMPONENT

147
component Flow
% calculate flow rate (SLPM)

inputs
I = {0, 'A'}; % I:left
end

outputs
H2_flow = {0, 'lpm'}; % H2f:right
Air_flow = {0, 'lpm'}; % Airf:right
end

parameters(Hidden = true);
F = { 96485, 'c/mol' }; % Faraday's constant
noc = { 12, '1' }; % number of cells
end

equations
% note: Simscape has an implicit conversion from liters per second
% to liters per minute. If you manually do this conversion
% as I tried, the SLPM value will be 60 times too large
% since Simscape does not recognize it.
% So, I had to comment out the Liter per second to
% Liters per minute conversion.
H2_flow == ((noc*I)/(2*F))*{24.46, 'l/mol'};%*{60,'s/min'};
Air_flow == ((noc*I)/(4*(0.2095)*F))*{24.46, 'l/mol'};%*{60,'s/min'};
end
end

148
 APPENDIX L

SIMSCAPE CODE FOR PEMFC COMPONENT (TOP LEVEL)

149
component MyPEMFC_2
% PEMFC model 2

% PEMFC model based on IEEE paper,

% "Dynamic and Steady-State Performance of PEM Fuel Cells
% under Various Loading Condition",
% Dong-Jing Lee & Li Wang, 2007, IEEE
nodes
p = foundation.electrical.electrical; %+:right
n = foundation.electrical.electrical; %-:right
end

inputs
H2_pressure = {1, 'atm'}; % H2p:right
Air_pressure = {1, 'atm'}; % Airp:right
end

outputs
H2_flow = {0, 'lpm'}; % H2c:left
Air_flow = {0, 'lpm'}; % Airc:left
sVoltage = {12, 'V'}; % sV:left
sCurrent = {0, 'A'}; % sC:left
eVr = {0, 'V'}; % Vrev:left
eRa = {0, 'Ohm'}; % eRact:left
eRo = {0, 'Ohm'}; % eRohm:left
eRc = {0, 'Ohm'}; % eRcon:left
Vdl = {0, 'V'}; % Vdl:left
Idl = {0, 'A'}; % Idl:left
end

parameters
c_noc = {20, '1'}; % number of cells
c_T = { (65 + 273.15), 'K' }; % temperature (K)
c_Area = { 22.25, 'cm^2' }; % cell active area (cm^2)
c_l = { 0.051, 'cm' }; % thickness of Nafion membrane (cm)
c_B = { 0.076, 'V' }; % concentration loss coefficient (V)
c_Jmax = { 0.43, 'A/cm^2' }; % max current density (A/cm^2)
c_CapVal = { 1.0, 'uF' }; % double-layer capacitor value
c_CIV = { 0.0, 'V' }; % capacitor initial voltage
end

components(Hidden = true)
Erev = Ch3Lib.Enernst(T = c_T, noc = c_noc);
Ract = Ch3Lib.Ractivation(noc=c_noc);
RLUT = Ch3Lib.RactLUT;
Rcon = Ch3Lib.Rconcentration(Area = c_Area,Jmax = c_Jmax,...
noc = c_noc, B = c_B);
Rohm = Ch3Lib.Rohmic(Area=c_Area, T=c_T, l=c_l, noc=c_noc);
Cap = Ch3Lib.dlCap(C = c_CapVal, CIV = c_CIV);
% sensors
CSense = foundation.electrical.sensors.current;
VSense = foundation.electrical.sensors.voltage;
% calculation block (calculate flow rates)
gFlow = Ch3Lib.Flow(noc = c_noc);
end
150
 function setup
% pass parameters from top level component to sub-components
Erev.T = c_T;
Erev.noc = c_noc;
% Ract.Area = c_Area;
% Ract.T = c_T;
Ract.noc = c_noc;
Rcon.Area = c_Area;
Rcon.B = c_B;
Rcon.Jmax = c_Jmax;
Rcon.noc = c_noc;
Rohm.Area = c_Area;
Rohm.T = c_T;
Rohm.l = c_l;
Rohm.noc = c_noc;
Cap.C = c_CapVal;
end

connections
connect(n, Erev.n, VSense.n);
connect(Erev.p, Ract.p, Cap.p);
connect(Ract.n, Rcon.p);
connect(Rcon.n, Rohm.p, Cap.n);
connect(Rohm.n, CSense.p);
connect(CSense.n, p, VSense.p);
connect(H2_pressure, Erev.H2_pressure, Ract.H2_pressure);
connect(Air_pressure, Erev.Air_pressure, Ract.Air_pressure);
connect(CSense.I, gFlow.I, sCurrent);
connect(VSense.V, sVoltage);
connect(gFlow.H2_flow, H2_flow);
connect(gFlow.Air_flow, Air_flow);
connect(Erev.Vrev, eVr);
connect(Rohm.effective_Res, eRo);
connect(Rcon.effective_Res, eRc);
connect(Ract.effective_Res, eRa);
connect(Cap.Vdl, Vdl);
connect(Cap.Idl, Idl);
end
end

151
 APPENDIX M

MATLAB CODE FOR DYNAMIC RESPONSE TEST

152
% PEMFC_Dynamic_Test.m
%
% Test the PEMFC for a pulsed current load for a range of
% double layer capacitance values.
cdl = 10^-6;
cdl_vals = [10^-6 0.1 0.25 0.5 1];
cm = [0 0 102; 204 0 204; 0 255 0; 255 128 0; 255 0 0];
cm = (1/255)*cm;
for i = 1:5
cdl = cdl_vals(i);
sim('MyPEMFC_2_Dynamics');
tv = simlog.PEMFC_model_2.sVoltage.series.time;
vv = simlog.PEMFC_model_2.sVoltage.series.values;
ti = simlog.PEMFC_model_2.sCurrent.series.time;
ii = simlog.PEMFC_model_2.sCurrent.series.values;
subplot(1,2,1)
plot(tv,vv, 'LineWidth', 2, 'Color', [cm(i,1) cm(i,2) cm(i,3)]);
if i == 1
hold on;
end
subplot(1,2,2)
plot(ti, ii, 'LineWidth', 2, 'Color', [cm(i,1) cm(i,2) cm(i,3)]);
if i == 1
hold on;
end
end
subplot(1,2,1)
hold off;
title('PEMFC Stack Voltage versus Time');
xlabel('Time (seconds)');
ylabel('Voltage (V)');
legend('C = 1 uF', 'C = 0.1 F', 'C = 0.25 F', 'C = 0.5 F', 'C = 1 F');
set(gca, 'FontSize', 14);
grid;
subplot(1,2,2)
hold off;
title('PEMFC Current versus Time');
xlabel('Time (seconds)');
ylabel('Voltage (V)');
legend('C = 1 uF', 'C = 0.1 F', 'C = 0.25 F', 'C = 0.5 F', 'C = 1 F');
set(gca, 'FontSize', 14);
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

153
 APPENDIX N

SIMSCAPE CODE FOR MPSS COMPONENTS

154
component RPMvsPower
% RPMvsPower

inputs
I = { 0, '1' }; % P:left
end

outputs
O = { 0, '1' }; % RPM:right
end

parameters(Size=variable)
xd = [0 25 50 75 100 125 150 175 200 225 250 275 300 325 350];
yd = [0 2168 2983 3597 4107 4552 4952 5316 5654 5969 6267 6549 6816 7073 7319];
end

equations
O == tablelookup(xd,yd,I,interpolation=cubic,extrapolation=nearest);
end
end

component TorquevsRPM
% TorquevsRPM

inputs
I = { 0, '1' }; % RPM:left
end

outputs
O = { 0, '1' }; % T:right
end

parameters(Size=variable)
xd = [0 2168 2983 3597 4107 4552 4952 5316 5654 5969 6267 6549 ...
6816 7073 7319];
yd = -(5.8/(2*3.14159))*[0 0.001835 0.002668 0.003318 0.003875 0.004370 ...
0.004821 0.005239 0.005630 0.005999 0.006349 0.006683 ...
0.007005 0.007313 0.007611];
end

equations
O == tablelookup(xd, yd, I, interpolation=cubic, ...
extrapolation=nearest);
end
end

component TorqueLoad
% Torque_Load

% idel torque source (use as load)

nodes
C = foundation.mechanical.rotational.rotational; % C:bottom
R = foundation.mechanical.rotational.rotational; % R:top
155
 end

inputs
S = { 0, 'N*m'}; % S:bottom
end

variables(Access = protected)
w = { 0, 'rpm' }; % angular velocity (across variable)
t = { 0, 'N*m' }; % torque (through variable)
end

branches
t : C.t -> R.t; % through variable
end

equations
w == C.w - R.w;
t == S;
end
end

156
 APPENDIX O

OBSERVER_1 DIAGRAM

157
Figure 60. Observer_1 Diagram

158
 APPENDIX P

OBSERVER_2 DIAGRAM

159
Figure 61. Observer_2 Diagram

160
 APPENDIX Q

HYBRID POWER SYSTEM MODEL CONFIGURATION PARAMETERS

161
Figure 62. Hybrid Power System Model Configuration Parameters

162
 APPENDIX R

FLIGHT PROFILE AUTOMATION SCRIPT

163
%
% flight_profile_03.m
%
% Flight Time: 6 hours
% (21,600 seconds + 200 seconds idle after landing)
TFT = 6*60*60 + 200; % total flight time
% Fuel Cell parameters
fcnoc = 20; % number of cells in the fuel cell
fcF = 96485; % Faraday's constant (coulombs/mole)
% Input gas pressures to fuel cell
h2p = 0.5; % hydrogen pressure (atm)
airp = 1.0; % air pressure (atm)
% Hydrogen fuel stored
h2Fuel = 400; % total hydrogen fuel available (Liters at STP)
% Battery parameters
BatInitSOC = 100; % initial battery state of charge
BatCapacity = 8; % Battery Capacity
% Battery charger
charger_on = 0; % turn decent charger on or off (0 = off, 1 = on)
charger_val = 0.4; % current setpoint for charger
% Hybrid parameter: zero = no battery assist during cruise
hybrid = 0.0; % hybrid level -> adjust CAOC setpoint

% update flight profile

throttle = [0 0;
100 0;
140 1.0; % 3 minute take-off
320 1.0;
340 0.6; % 4 minute climb
480 0.6;
500 0.25; % 58.3 minute cruise
4000 0.25;
4020 0.4; % 4 minute climb
4260 0.4;
4280 0.25;
7000 0.25; % start wind gusts
7010 0.35;
7040 0.35;
7050 0.25;
7100 0.25; % wind gusts
7110 0.35;
7140 0.35;
7150 0.25;
7200 0.25; % wind gusts
7210 0.35;
7240 0.35;
7250 0.25;
7300 0.25; % wind gusts
7310 0.35;
7340 0.35;
7350 0.25;
7400 0.25; % wind gusts
7410 0.35;
7440 0.35;
7450 0.25; % end wind gusts
164
 21200 0.25; % prepare for descent, meneuver to line up for landing
21220 0.3; % 30% power for 1 minute
21280 0.3;
21300 0.2; % descent for landing
21550 0.2; % landed, aircraft now on the ground
21570 0.3; % taxi start
21670 0.3; % taxi end
21680 0.1; % idle start
21780 0.1; % idel end
21800 0.0]; % shut down motor

signalbuilder('PowerManagement_08/Flight_Profile','set', ...
'profile_01', 'group_01', throttle(:,1), throttle(:,2));

% Battery Assist On/Off signal (turn on during cruise, otherwise off)

bas_profile = [ 0 0;
500 0;
501 1;
21300 1;
21301 0;
21800 0];
signalbuilder('PowerManagement_08/Controller/BAS_signal', 'set', ...
'BAS_signal_01', 'Group_01', ...
bas_profile(:,1), bas_profile(:,2));

% Fuel Cell BOP load

bop_res = [0 18;
2050 18;
2060 11; % fan on
2240 11;
2250 18; % fan off
4990 18;
5000 11; % fan on
5180 11;
5190 18; % fan off
9000 18;
9010 11; % fan on
9190 11;
9200 18; % fan off
14000 18;
14010 11; % fan on
14190 11;
14200 18; % fan off
21800 18];

signalbuilder('PowerManagement_08/Balance_of_Plant/Fuel_Cell_BOP', ...
'set', 'bop_profile_01', 'group_01', ...
bop_res(:,1), bop_res(:,2));

165
 APPENDIX S

MATLAB SCRIPT TO GENERATE FIGURE 45

166
% fp03_run1and2_plot_voltage_and_current.m
%
% Run PowerManagement_08.slx for hybrid parameters h = {0, 1.5} and
% plot the voltage and current responses on the same graphs.
% m-file dependency: flight_profile_03.m
%
cm = [0 0 0; 0 0 250; 250 0 0; 0 140 0]; % color matrix
cm = (1/255)*cm; % normalize color matrix
flight_profile_03; % run flight_profile_03.m (sets hybrid = 0)
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
vfc_h0p0_t = simlog.PEMFC_model_2.sVoltage.series.time;
vfc_h0p0_v = simlog.PEMFC_model_2.sVoltage.series.values;
ifc_h0p0_t = simlog.PEMFC_model_2.sCurrent.series.time;
ifc_h0p0_v = simlog.PEMFC_model_2.sCurrent.series.values;
vbat_h0p0_t = simlog.LiPo_4S.BatteryVoltage.series.time;
vbat_h0p0_v = simlog.LiPo_4S.BatteryVoltage.series.values;
ibat_h0p0_t = simlog.LiPo_4S.BatteryCurrent.series.time;
ibat_h0p0_v = simlog.LiPo_4S.BatteryCurrent.series.values;
hybrid = 1.5; % change the hybrid parameter to 1.5
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
vfc_h1p5_t = simlog.PEMFC_model_2.sVoltage.series.time;
vfc_h1p5_v = simlog.PEMFC_model_2.sVoltage.series.values;
ifc_h1p5_t = simlog.PEMFC_model_2.sCurrent.series.time;
ifc_h1p5_v = simlog.PEMFC_model_2.sCurrent.series.values;
vbat_h1p5_t = simlog.LiPo_4S.BatteryVoltage.series.time;
vbat_h1p5_v = simlog.LiPo_4S.BatteryVoltage.series.values;
ibat_h1p5_t = simlog.LiPo_4S.BatteryCurrent.series.time;
ibat_h1p5_v = simlog.LiPo_4S.BatteryCurrent.series.values;
% plot the results
fh = figure;
subplot(1,2,1)
plot(vfc_h0p0_t/60, vfc_h0p0_v, 'LineWidth', 2, 'Color', cm(1,:));
hold on;
plot(vbat_h0p0_t/60, vbat_h0p0_v, 'LineWidth', 2, 'Color', cm(2,:));
plot(vfc_h1p5_t/60, vfc_h1p5_v, 'LineWidth', 2, 'Color', cm(3,:));
plot(vbat_h1p5_t/60, vbat_h1p5_v, 'LineWidth', 2, 'Color', cm(4,:));
hold off;
title('Fuel Cell and Battery Voltage versus Time');
xlabel('Time (minutes)');
ylabel('Voltage (V)');
legend('Fuel Cell Voltage, hybrid = 0', 'Battery Voltage, hybrid = 0', ...
'Fuel Cell Voltage, hybrid = 1.5', 'Battery Voltage, hybrid = 1.5');
set(gca, 'FontSize', 14);
grid;
subplot(1,2,2)
plot(ifc_h0p0_t/60, ifc_h0p0_v, 'LineWidth', 2, 'Color', cm(1,:));
hold on;
plot(ibat_h0p0_t/60, ibat_h0p0_v, 'LineWidth', 2, 'Color', cm(2,:));
plot(ifc_h1p5_t/60, ifc_h1p5_v, 'LineWidth', 2, 'Color', cm(3,:));
plot(ibat_h1p5_t/60, ibat_h1p5_v, 'LineWidth', 2, 'Color', cm(4,:));
hold off;
title('Fuel Cell and Battery Current versus Time');
xlabel('Time (minutes)');
167
 ylabel('Current (A)');
legend('Fuel Cell Current, hybrid = 0', 'Battery Current, hybrid = 0', ...
'Fuel Cell Current, hybrid = 1.5', 'Battery Current, hybrid = 1.5');
set(gca, 'FontSize', 14);
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

168
 APPENDIX T

MATLAB SCRIPT TO GENERATE FIGURE 46

169
% fp03_run1and2_plot_Power_and_Wh.m
%
% Run PowerManagement_08.slx for hybrid parameters h = {0, 1.5} and
% plot the Fuel Cell, Battery, Load, Power, and Q1 Power on the same graph.
% m-file dependency: flight_profile_03.m
%
cm = [0 0 0; 0 0 250; 250 0 0; 0 140 0]; % color matrix
cm = (1/255)*cm; % normalize color matrix
flight_profile_03; % run flight_profile_03.m (sets hybrid = 0)
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
% Fuel Cell and Battery Voltage and Current
vfc_h0p0_t = simlog.PEMFC_model_2.sVoltage.series.time;
vfc_h0p0_v = simlog.PEMFC_model_2.sVoltage.series.values;
ifc_h0p0_t = simlog.PEMFC_model_2.sCurrent.series.time;
ifc_h0p0_v = simlog.PEMFC_model_2.sCurrent.series.values;
vbat_h0p0_t = simlog.LiPo_4S.BatteryVoltage.series.time;
vbat_h0p0_v = simlog.LiPo_4S.BatteryVoltage.series.values;
ibat_h0p0_t = simlog.LiPo_4S.BatteryCurrent.series.time;
ibat_h0p0_v = simlog.LiPo_4S.BatteryCurrent.series.values;
FC_Power_h0p0 = vfc_h0p0_v .* ifc_h0p0_v; % calculate fuel cell power
Bat_Power_h0p0 = vbat_h0p0_v .* ibat_h0p0_v; % calculate battery power
% Elecrical load power and Mechanical load power
eLoad_h0p0_t = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.time;
eLoad_h0p0_v = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.values;
mLoad_h0p0_t = simlog.Motor_Prop_Subsystem.Mechanical_Product.O.series.time;
mLoad_h0p0_v = simlog.Motor_Prop_Subsystem.Mechanical_Product.O.series.values;
% Q1 Vds and current
q1vs_h0p0_t = simlog.Controller.Q1.S.v.series.time;
q1vs_h0p0_v = simlog.Controller.Q1.S.v.series.values;
q1vd_h0p0_t = simlog.Controller.Q1.D.v.series.time;
q1vd_h0p0_v = simlog.Controller.Q1.D.v.series.values;
q1vds_h0p0 = q1vs_h0p0_v - q1vd_h0p0_v; % calculate voltage drop of Q1
q1power_h0p0 = q1vds_h0p0 .* ibat_h0p0_v; % calculate Q1 power
hybrid = 1.5; % change the hybrid parameter to 1.5
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
% Fuel Cell and Battery Voltage and Current
vfc_h1p5_t = simlog.PEMFC_model_2.sVoltage.series.time;
vfc_h1p5_v = simlog.PEMFC_model_2.sVoltage.series.values;
ifc_h1p5_t = simlog.PEMFC_model_2.sCurrent.series.time;
ifc_h1p5_v = simlog.PEMFC_model_2.sCurrent.series.values;
vbat_h1p5_t = simlog.LiPo_4S.BatteryVoltage.series.time;
vbat_h1p5_v = simlog.LiPo_4S.BatteryVoltage.series.values;
ibat_h1p5_t = simlog.LiPo_4S.BatteryCurrent.series.time;
ibat_h1p5_v = simlog.LiPo_4S.BatteryCurrent.series.values;
FC_Power_h1p5 = vfc_h1p5_v .* ifc_h1p5_v; % calculate fuel cell power
Bat_Power_h1p5 = vbat_h1p5_v .* ibat_h1p5_v; % calculate battery power
% Electrical load power and Mechanical load power
eLoad_h1p5_t = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.time;
eLoad_h1p5_v = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.values;
mLoad_h1p5_t = simlog.Motor_Prop_Subsystem.Mechanical_Product.O.series.time;
mLoad_h1p5_v = simlog.Motor_Prop_Subsystem.Mechanical_Product.O.series.values;
% Q1 Vds and current
q1vs_h1p5_t = simlog.Controller.Q1.S.v.series.time;
170
q1vs_h1p5_v = simlog.Controller.Q1.S.v.series.values;
q1vd_h1p5_t = simlog.Controller.Q1.D.v.series.time;
q1vd_h1p5_v = simlog.Controller.Q1.D.v.series.values;
q1vds_h1p5 = q1vs_h1p5_v - q1vd_h1p5_v; % calculate voltage drop of Q1
q1power_h1p5 = q1vds_h1p5 .* ibat_h1p5_v; % calculate Q1 power
% plot the results
fh = figure;
subplot(1,2,1)
plot(vfc_h0p0_t/60, FC_Power_h0p0, 'LineWidth', 2, 'Color', cm(1,:));
hold on;
plot(vbat_h0p0_t/60, Bat_Power_h0p0, 'LineWidth', 2, 'Color', cm(2,:));
plot(eLoad_h0p0_t/60, eLoad_h0p0_v, 'LineWidth', 2, 'Color', cm(3,:));
plot(ibat_h0p0_t/60, q1power_h0p0, 'LineWidth', 2, 'Color', cm(4,:));
hold off;
title('Electrical Power, hybrid = 0 (Total Energy = 583.5 Wh)');
xlabel('Time (minutes)');
ylabel('Power (W)');
legend('Fuel Cell Power', 'Battery Power', 'Load Power', 'Q1 Power');
set(gca, 'FontSize', 14);
grid;
subplot(1,2,2)
plot(vfc_h1p5_t/60, FC_Power_h1p5, 'LineWidth', 2, 'Color', cm(1,:));
hold on;
plot(vbat_h1p5_t/60, Bat_Power_h1p5, 'LineWidth', 2, 'Color', cm(2,:));
plot(eLoad_h1p5_t/60, eLoad_h1p5_v, 'LineWidth', 2, 'Color', cm(3,:));
plot(ibat_h1p5_t/60, q1power_h1p5, 'LineWidth', 2, 'Color', cm(4,:));
hold off;
title('Electrical Power, hybrid = 1.5 (Total Energy = 585.2 Wh)');
xlabel('Time (minutes)');
ylabel('Power (W)');
legend('Fuel Cell Power', 'Battery Power', 'Load Power', 'Q1 Power');
set(gca, 'FontSize', 14);
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

171
 APPENDIX U

MATLAB SCRIPT TO GENERATE FIGURE 47

172
% fp03_run1and2_plot_BatSOC_and_Hydrogen.m
%
% Run PowerManagement_08.slx for hybrid parameters h = {0, 1.5} and
% plot the battery SOC and hydrogen consumed on the same graphs.
% m-file dependency: flight_profile_03.m
%
cm = [0 0 0; 0 0 250; 250 0 0; 0 140 0]; % color matrix
cm = (1/255)*cm; % normalize color matrix
flight_profile_03; % run flight_profile_03.m (sets hybrid = 0)
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
soc_h0p0_t = simlog.LiPo_4S.SOC.series.time;
soc_h0p0_v = simlog.LiPo_4S.SOC.series.values;
hyd_h0p0_t = simlog.Observer_1.H2Liters.O.series.time;
hyd_h0p0_v = simlog.Observer_1.H2Liters.O.series.values;
hybrid = 1.5; % change the hybrid parameter to 1.5
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
soc_h1p5_t = simlog.LiPo_4S.SOC.series.time;
soc_h1p5_v = simlog.LiPo_4S.SOC.series.values;
hyd_h1p5_t = simlog.Observer_1.H2Liters.O.series.time;
hyd_h1p5_v = simlog.Observer_1.H2Liters.O.series.values;
% plot the results
fh = figure;
subplot(1,2,1)
plot(soc_h0p0_t/60, soc_h0p0_v, 'LineWidth', 2, 'Color', cm(1,:));
hold on;
plot(soc_h1p5_t/60, soc_h1p5_v, 'LineWidth', 2, 'Color', cm(3,:));
hold off;
title('Battery SOC versus Time');
xlabel('Time (minutes)');
ylabel('SOC (%)');
legend('Battery SOC, hybrid = 0', 'Battery SOC, hybrid = 1.5');
set(gca, 'FontSize', 14);
grid;
subplot(1,2,2)
plot(hyd_h0p0_t/60, hyd_h0p0_v, 'LineWidth', 2, 'Color', cm(1,:));
hold on;
plot(hyd_h1p5_t/60, hyd_h1p5_v, 'LineWidth', 2, 'Color', cm(3,:));
hold off;
title('Hydrogen Consumed versus Time');
xlabel('Time (minutes)');
ylabel('Hydrogen (Liters)');
legend('Hydrogen Consumed, hybrid = 0', 'Hydrogen Consumed, hybrid = 1.5');
set(gca, 'FontSize', 14);
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

173
 APPENDIX V

MATLAB SCRIPT TO GENERATE FIGURE 48

174
% fp03_run1and2_plot_Q1_Power.m
%
% Run PowerManagement_08.slx for hybrid parameters h = {0, 1.5} and
% plot the Q1 Power for h = {0, 1.5} on the same graph.
% Use zoom plot within plot to zoom into to area of interest [0 200 -2 8].
% m-file dependency: flight_profile_03.m
%
cm = [0 0 0; 0 0 250; 250 0 0; 0 140 0]; % color matrix
cm = (1/255)*cm; % normalize color matrix
flight_profile_03; % run flight_profile_03.m (sets hybrid = 0)
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
% Battery Current
ibat_h0p0_t = simlog.LiPo_4S.BatteryCurrent.series.time;
ibat_h0p0_v = simlog.LiPo_4S.BatteryCurrent.series.values;
% Q1 Vds and current
q1vs_h0p0_t = simlog.Controller.Q1.S.v.series.time;
q1vs_h0p0_v = simlog.Controller.Q1.S.v.series.values;
q1vd_h0p0_t = simlog.Controller.Q1.D.v.series.time;
q1vd_h0p0_v = simlog.Controller.Q1.D.v.series.values;
q1vds_h0p0 = q1vs_h0p0_v - q1vd_h0p0_v; % calculate voltage drop of Q1
q1power_h0p0 = q1vds_h0p0 .* ibat_h0p0_v; % calculate Q1 power
hybrid = 1.5; % change the hybrid parameter to 1.5
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
% Battery Current
ibat_h1p5_t = simlog.LiPo_4S.BatteryCurrent.series.time;
ibat_h1p5_v = simlog.LiPo_4S.BatteryCurrent.series.values;
% Q1 Vds and current
q1vs_h1p5_t = simlog.Controller.Q1.S.v.series.time;
q1vs_h1p5_v = simlog.Controller.Q1.S.v.series.values;
q1vd_h1p5_t = simlog.Controller.Q1.D.v.series.time;
q1vd_h1p5_v = simlog.Controller.Q1.D.v.series.values;
q1vds_h1p5 = q1vs_h1p5_v - q1vd_h1p5_v; % calculate voltage drop of Q1
q1power_h1p5 = q1vds_h1p5 .* ibat_h1p5_v; % calculate Q1 power
% plot the results
fh = figure;
plot(ibat_h0p0_t/60, q1power_h0p0, 'LineWidth', 2, 'Color', cm(1,:));
hold on;
plot(ibat_h1p5_t/60, q1power_h1p5, 'LineWidth', 2, 'Color', cm(3,:));
hold off;
title('Q1 Power versus Time');
xlabel('Time (minutes)');
ylabel('Q1 Power (W)');
axis([0 370 -10 110]);
legend('hybrid = 0', 'hybrid = 1.5');
set(gca, 'FontSize', 14);
grid;
% zoom plot within same plot
axes('position', [0.25 0.4 0.5 0.5]);
box on; % put a box around new pair of axes
plot(ibat_h0p0_t/60, q1power_h0p0, 'LineWidth', 2, 'Color', cm(1,:));
hold on;
plot(ibat_h1p5_t/60, q1power_h1p5, 'LineWidth', 2, 'Color', cm(3,:));
hold off;
175
axis([0 200 -2 8]);
set(gca, 'FontSize', 18);
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

176
 APPENDIX W

CALCULATE CHANGE IN TOTAL ENERGIES BETWEEN SIMULATION RUNS

177
% fp03_run1and2_Calculate_Total_Energy.m
%
% Run PowerManagement_08.slx for hybrid parameters h = {0, 1.5} and
% calculate the difference in energy produced by the sources, consumed
% by Q1 and the load during the flight profiles when h = {0, 1.5}.
% m-file dependency: flight_profile_03.m
%
cm = [0 0 0; 0 0 250; 250 0 0; 0 140 0]; % color matrix
cm = (1/255)*cm; % normalize color matrix
% h = 0 simulation run
flight_profile_03; % run flight_profile_03.m (sets hybrid = 0)
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
% Fuel Cell Voltage
vfc_h0p0_t = simlog.PEMFC_model_2.sVoltage.series.time;
vfc_h0p0_v = simlog.PEMFC_model_2.sVoltage.series.values;
% Fuel Cell Current
ifc_h0p0_t = simlog.PEMFC_model_2.sCurrent.series.time;
ifc_h0p0_v = simlog.PEMFC_model_2.sCurrent.series.values;
% Battery Voltage
vbat_h0p0_t = simlog.LiPo_4S.BatteryVoltage.series.time;
vbat_h0p0_v = simlog.LiPo_4S.BatteryVoltage.series.values;
% Battery Current
ibat_h0p0_t = simlog.LiPo_4S.BatteryCurrent.series.time;
ibat_h0p0_v = simlog.LiPo_4S.BatteryCurrent.series.values;
% Q1 Vds and current
q1vs_h0p0_t = simlog.Controller.Q1.S.v.series.time;
q1vs_h0p0_v = simlog.Controller.Q1.S.v.series.values;
q1vd_h0p0_t = simlog.Controller.Q1.D.v.series.time;
q1vd_h0p0_v = simlog.Controller.Q1.D.v.series.values;
q1vds_h0p0 = q1vs_h0p0_v - q1vd_h0p0_v; % calculate voltage drop of Q1
q1power_h0p0 = q1vds_h0p0 .* ibat_h0p0_v; % calculate Q1 power
% get motor-propeller load power data
eLoad_h0p0_t = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.time;
eLoad_h0p0_v = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.values;
% get BOP load power data
vBOP_h0p0_t = simlog.Balance_of_Plant.BOP_Power.v1.series.time;
vBOP_h0p0_v = simlog.Balance_of_Plant.BOP_Power.v1.series.values;
iBOP_h0p0_t = simlog.Balance_of_Plant.BOP_Power.i1.series.time;
iBOP_h0p0_v = simlog.Balance_of_Plant.BOP_Power.i1.series.values;
% h = 1.5 simulation run
hybrid = 1.5; % change the hybrid parameter to 1.5
sim('PowerManagement_08.slx', 'StopTime', num2str(21800)); % run simulation
% extract data
% Fuel Cell Voltage
vfc_h1p5_t = simlog.PEMFC_model_2.sVoltage.series.time;
vfc_h1p5_v = simlog.PEMFC_model_2.sVoltage.series.values;
% Fuel Cell Current
ifc_h1p5_t = simlog.PEMFC_model_2.sCurrent.series.time;
ifc_h1p5_v = simlog.PEMFC_model_2.sCurrent.series.values;
% Battery Voltage
vbat_h1p5_t = simlog.LiPo_4S.BatteryVoltage.series.time;
vbat_h1p5_v = simlog.LiPo_4S.BatteryVoltage.series.values;
% Battery Current
ibat_h1p5_t = simlog.LiPo_4S.BatteryCurrent.series.time;
178
ibat_h1p5_v = simlog.LiPo_4S.BatteryCurrent.series.values;
% Q1 Vds and current
q1vs_h1p5_t = simlog.Controller.Q1.S.v.series.time;
q1vs_h1p5_v = simlog.Controller.Q1.S.v.series.values;
q1vd_h1p5_t = simlog.Controller.Q1.D.v.series.time;
q1vd_h1p5_v = simlog.Controller.Q1.D.v.series.values;
q1vds_h1p5 = q1vs_h1p5_v - q1vd_h1p5_v; % calculate voltage drop of Q1
q1power_h1p5 = q1vds_h1p5 .* ibat_h1p5_v; % calculate Q1 power
% get motor-propeller load data
eLoad_h1p5_t = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.time;
eLoad_h1p5_v = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.values;
% get BOP load power data
vBOP_h1p5_t = simlog.Balance_of_Plant.BOP_Power.v1.series.time;
vBOP_h1p5_v = simlog.Balance_of_Plant.BOP_Power.v1.series.values;
iBOP_h1p5_t = simlog.Balance_of_Plant.BOP_Power.i1.series.time;
iBOP_h1p5_v = simlog.Balance_of_Plant.BOP_Power.i1.series.values;
% Calculate the difference in total energy consumed by Q1 during flight
% profiles h = 0 and h = 1.5.
% note: h = 0 and h = 1.5 simulation runs have different number of data
% points. Data from each run must be integrated separately.
m = length(ibat_h0p0_t);
n = length(ibat_h1p5_t);
Q1_Energy_h0p0 = zeros(1,m);
Q1_Energy_Total_h0p0 = 0;
Q1_Energy_h1p5 = zeros(1,n);
Q1_Energy_Total_h1p5 = 0;
delta_t = 0;
delta_Q1_Energy = 0;
% Calculate total energy consumed by Q1 for h = 0 run.
for i = 2:m
delta_t = ibat_h0p0_t(i) - ibat_h0p0_t(i-1);
y = 0.5*(q1power_h0p0(i) + q1power_h0p0(i-1));
Q1_Energy_h0p0(i) = y * delta_t; % Joules
Q1_Energy_Total_h0p0 = Q1_Energy_Total_h0p0 + Q1_Energy_h0p0(i);
end
% Calculate total energy consumed by Q1 for h = 1.5 run.
for i = 2:n
delta_t = ibat_h1p5_t(i) - ibat_h1p5_t(i-1);
y = 0.5*(q1power_h1p5(i) + q1power_h1p5(i-1));
Q1_Energy_h1p5(i) = y * delta_t; % Joules
Q1_Energy_Total_h1p5 = Q1_Energy_Total_h1p5 + Q1_Energy_h1p5(i);
end
delta_Q1_Energy = Q1_Energy_Total_h1p5 - Q1_Energy_Total_h0p0;
delta_Q1_Energy = delta_Q1_Energy / 3600; % convert to Watt-hours
% display the difference in Q1 total energy
delta_Q1_Energy
% Next, calculate total energy consumed by the motor-propeller load for h = 0
m = length(eLoad_h0p0_t);
Load_Energy_h0p0 = zeros(1,m);
Load_Energy_Total_h0p0 = 0;
for i = 2:m
delta_t = eLoad_h0p0_t(i) - eLoad_h0p0_t(i-1);
y = 0.5*(eLoad_h0p0_v(i) + eLoad_h0p0_v(i-1));
Load_Energy_h0p0(i) = y * delta_t; % Joules
Load_Energy_Total_h0p0 = Load_Energy_Total_h0p0 + Load_Energy_h0p0(i);
179
end
% Now, calculate total energy consumed by the motor-propeller load for h = 1.5
n = length(eLoad_h1p5_t);
Load_Energy_h1p5 = zeros(1,n);
Load_Energy_Total_h1p5 = 0;
for i = 2:n
delta_t = eLoad_h1p5_t(i) - eLoad_h1p5_t(i-1);
y = 0.5*(eLoad_h1p5_v(i) + eLoad_h1p5_v(i-1));
Load_Energy_h1p5(i) = y * delta_t; % Joules
Load_Energy_Total_h1p5 = Load_Energy_Total_h1p5 + Load_Energy_h1p5(i);
end
delta_Load = Load_Energy_Total_h1p5 - Load_Energy_Total_h0p0;
delta_Load = delta_Load / 3600; % convert to Watt-hours
delta_Load

% Next, calculate the total energy consumed by the BOP load for h = 0
m = length(iBOP_h0p0_t);
BOP_Energy_h0p0 = zeros(1,m);
BOP_Energy_Total_h0p0 = 0;
for i = 2:m
delta_t = iBOP_h0p0_t(i) - iBOP_h0p0_t(i-1);
y = 0.5*(vBOP_h0p0_v(i) + vBOP_h0p0_v(i-1));
z = 0.5*(iBOP_h0p0_v(i) + iBOP_h0p0_v(i-1));
BOP_Energy_h0p0(i) = (y * z) * delta_t; % Joules
BOP_Energy_Total_h0p0 = BOP_Energy_Total_h0p0 + BOP_Energy_h0p0(i);
end

% Now, calculate the total energy consumed by the BOP load for h = 1.5
n = length(iBOP_h1p5_t);
BOP_Energy_h1p5 = zeros(1,n);
BOP_Energy_Total_h1p5 = 0;
for i = 2:n
delta_t = iBOP_h1p5_t(i) - iBOP_h1p5_t(i-1);
y = 0.5*(vBOP_h1p5_v(i) + vBOP_h1p5_v(i-1));
z = 0.5*(iBOP_h1p5_v(i) + iBOP_h1p5_v(i-1));
BOP_Energy_h1p5(i) = (y * z) * delta_t; % Joules
BOP_Energy_Total_h1p5 = BOP_Energy_Total_h1p5 + BOP_Energy_h1p5(i);
end
delta_BOP = BOP_Energy_Total_h1p5 - BOP_Energy_Total_h0p0;
delta_BOP = delta_BOP / 3600; % change to Watt-hours
delta_BOP

% Next, calculate the total energy delivered by the sources for h = 0.

m = length(ifc_h0p0_t);
FC_Energy_h0p0 = zeros(1,m);
BT_Energy_h0p0 = zeros(1,m);
Sources_Energy_h0p0 = zeros(1,m);
Sources_Energy_Total_h0p0 = 0;
for i = 2:m
delta_t = ifc_h0p0_t(i) - ifc_h0p0_t(i-1);
y = 0.5*(vfc_h0p0_v(i) + vfc_h0p0_v(i-1));
z = 0.5*(ifc_h0p0_v(i) + ifc_h0p0_v(i-1));
FC_Energy_h0p0(i) = (y * z) * delta_t; % Joules
y = 0.5*(vbat_h0p0_v(i) + vbat_h0p0_v(i-1));
z = 0.5*(ibat_h0p0_v(i) + ibat_h0p0_v(i-1));
180
 BT_Energy_h0p0(i) = (y * z) * delta_t; % Joules
Sources_Energy_h0p0(i) = FC_Energy_h0p0(i) + BT_Energy_h0p0(i);
Sources_Energy_Total_h0p0 = Sources_Energy_Total_h0p0 + Sources_Energy_h0p0(i);
end
% Now, calculate the total energy delivered by the source for h = 1.5.
n = length(ifc_h1p5_t);
FC_Energy_h1p5 = zeros(1,n);
BT_Energy_h1p5 = zeros(1,n);
Sources_Energy_h1p5 = zeros(1,n);
Sources_Energy_Total_h1p5 = 0;
for i = 2:n
delta_t = ifc_h1p5_t(i) - ifc_h1p5_t(i-1);
y = 0.5*(vfc_h1p5_v(i) + vfc_h1p5_v(i-1));
z = 0.5*(ifc_h1p5_v(i) + ifc_h1p5_v(i-1));
FC_Energy_h1p5(i) = (y * z)* delta_t; % Joules
y = 0.5*(vbat_h1p5_v(i) + vbat_h1p5_v(i-1));
z = 0.5*(ibat_h1p5_v(i) + ibat_h1p5_v(i-1));
BT_Energy_h1p5(i) = (y * z) * delta_t; % Joules
Sources_Energy_h1p5(i) = FC_Energy_h1p5(i) + BT_Energy_h1p5(i);
Sources_Energy_Total_h1p5 = Sources_Energy_Total_h1p5 + Sources_Energy_h1p5(i);
end
delta_Sources = Sources_Energy_Total_h1p5 - Sources_Energy_Total_h0p0;
delta_Sources = delta_Sources / 3600; % convert to Watt-hours
delta_Sources

181
 APPENDIX X

THRUST TEST BENCH DATA

182
Motor: AXI 4120/14
Electronic Speed Controller: JETI Advance 70 Pro (5V < Vin < 15V)
Radio Receiver: Spektrum AR7000
Radio Transmitter: Spektrum DX7
Power Sources: DC power supply BK Precision 1670A, Motorcycle battery 12V

Figure 63. Oscilloscope Capture of ESC Voltage Signals Sent to Motor

Figure 64. Oscilloscope Capture of ESC Voltage and Current Motor Signals (zoomed in)

183
Figure 65. Motor Measurements with DC Power Supply and No Propeller

184
Figure 66. Motor Measurements with DC Power Supply with Propeller

185
Figure 67. Motor Measurements with Motorcycle Battery with Propeller

186
 APPENDIX Y

SIMULATION OF SIX HOUR FLIGHT FOR DIFFERENT BATTERY CAPACITIES

187
% HPMC_Battery_Capacity.m
%
% Run the Power_Management_08.slx model with flight_profile_03.
% The battery capacity is varied from 8 Ah to 12 Ah, and the
% amount of hydrogen consumed during the flight profile is recorded.
% Plot the battery SOC and hydrogen consumption curves for each run.
% file dependency: flight_profile_03.m
cm = [0 0 0; 0 0 250; 250 0 0; 0 140 0; 244 160 64]; % color matrix
cm = (1/255)*cm; % normalize color matrix
flight_profile_03; % generate Matlab variables for simulation
hybrid = 1.5; % set the hybrid parameter to 1.5
h2_used = zeros(1,5); % store the h2 used for each simulation
bat_min_SOC = zeros(1,5); % record minimum battery SOC for each simulation
for i = 1:5
BatCapacity = 8 + (i-1); % set the battery capacity
% run the simulation
sim('PowerManagement_08.slx', 'StopTime', num2str(21800));
% extract the data
% battery SOC
bSOC_t = simlog.LiPo_4S.SOC.series.time;
bSOC_v = simlog.LiPo_4S.SOC.series.values;
% hydrogen consumed
hc_t = simlog.Observer_1.H2Liters.O.series.time;
hc_v = simlog.Observer_1.H2Liters.O.series.values;
% total energy consumed
h2_used(i) = max(hc_v);
bat_min_SOC(i) = min(bSOC_v);
% create plot
subplot(1,2,1)
plot(bSOC_t/60, bSOC_v, 'LineWidth', 2, 'Color', cm(i,:));
if i == 1
hold on;
end
subplot(1,2,2)
plot(hc_t/60, hc_v, 'LineWidth', 2, 'Color', cm(i,:));
if i == 1
hold on;
end
end
% generate strings for legend
bat_8 = strcat('Battery: 8 Ah, H2: ', num2str(h2_used(1),'%.1f'));
bat_9 = strcat('Battery: 9 Ah, H2: ', num2str(h2_used(2),'%.1f'));
bat_10 = strcat('Battery: 10 Ah, H2: ', num2str(h2_used(3),'%.1f'));
bat_11 = strcat('Battery: 11 Ah, H2: ', num2str(h2_used(4),'%.1f'));
bat_12 = strcat('Battery: 12 Ah, H2: ', num2str(h2_used(5),'%.1f'));

subplot(1,2,1)
hold off;
title('Battery SOC versus Time');
xlabel('Time (minutes)');
ylabel('SOC (%)');
legend(bat_8, bat_9, bat_10, bat_11, bat_12);
set(gca, 'FontSize', 14);
grid;
subplot(1,2,2)
188
 hold off;
title('Hydrogen Consumed versus Time');
xlabel('Time (minutes)');
ylabel('Hydrogen Consumed (L)');
legend(bat_8, bat_9, bat_10, bat_11, bat_12);
set(gca, 'FontSize', 14);
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);

189
 APPENDIX Z

MATLAB SCRIPT TO CALCULATE TAKEOFF AND LANDING ENERGY

190
% Energy_Analysis.m
%
% Integrate Load Electric Power and divide by time to obtain
% a graph of energy use per minute (Wh/min)
Motor_t = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.time;
Motor_v = simlog.Motor_Prop_Subsystem.Electrical_Product.O.series.values;
BOP_t = simlog.Balance_of_Plant.BOP_Power.v1.series.time;
BOP_voltage = simlog.Balance_of_Plant.BOP_Power.v1.series.values;
BOP_current = simlog.Balance_of_Plant.BOP_Power.i1.series.values;
BOP_v = BOP_voltage .* BOP_current;
Power_v = Motor_v + BOP_v;
% Plot the electrical load
plot(Motor_t/60, Power_v/60, 'LineWidth', 2); % (Wh/min)
title('Electrical Load Power versus Time');
xlabel('Time (minutes)');
ylabel('Power (Wh/min)');
set(gca, 'FontSize', 18);
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);
% Calculate energy used during takeoff phase of flight profile
take_off_energy = 0;
m = find(Motor_t == 480); % get index where t = 480 seconds
for i = 2:m
delta_t = Motor_t(i) - Motor_t(i-1);
y = 0.5*(Power_v(i) + Power_v(i-1));
energy = y * delta_t; % Joules
Wh = energy / 3600;
take_off_energy = take_off_energy + Wh;
end
% Calculate energy used during landing phase of flight profile
landing_energy = 0;
n = find(Motor_t == 21200);
l = length(Motor_t);
for i = (n+1):l
delta_t = Motor_t(i) - Motor_t(i-1);
y = 0.5*(Power_v(i) + Power_v(i-1));
energy = y * delta_t; % Joules
Wh = energy / 3600;
landing_energy = landing_energy + Wh;
end
% display the results
[take_off_energy landing_energy]

191
 APPENDIX AA

MATLAB SCRIPT TO PLOT FLIGHT TIME VERSUS POWER SYSTEM WEIGHT

192
% Weight_FlightTime.m
%
% Compare flight time for various hybrid power system configurations.
% All configurations (except config_1) use all three sizes of Ninja
% compressed air tanks to store hydrogen (Ninja Pro v2 50, 68, & 90).
%
% config_1 = battery only
% config_2 = Hybrid with current assist, H-100 fuel cell, 6Ah battery
% config_3 = Hybrid with current assist, H-100 fuel cell, 11Ah battery
% config_4 = Hybrid with current assist, H-100 fuel cell, 17Ah battery
% config_5 = Hybrid with current assist, AST01-01 fuel cell, 6Ah battery
% config_6 = Hybrid with current assist, AST01-01 fuel cell, 11Ah battery
% config_7 = Hybrid with current assist, AST01-01 fuel cell, 17Ah battery
% config_8 = H-500 fuel cell only
% config_9 = Hybrid with no current assist, H-100 fuel cell, 6Ah battery
% using a 220g DC-to-DC converter (Conventional Hybrid)
%
% color matrix
cm = [0 0 0; 0 0 250; 250 0 0; 0 140 0; 244 160 64; 170 0 151];
cm = (1/255)*cm; % normalize color matrix
% Battery Data
Battery = [0.570 6; 0.850 11; 1.294 17]; % [kg Ah]
% Fuel Cell Data
FC_Weight = [0.55 1.69 2.92]; % [AST01-01 H-100 H-500] in kg
% DC-to-DC converter weight (400W => 0.220 kg)
DCC_Weight = 0.220; % (kg)
% Hydrogen Storage Tank Data
Ninja_50 = [0.936 166]; % [kg L]
Ninja_68 = [1.19 225.8]; % [kg L]
Ninja_90 = [1.45 298.9]; % [kg L]
% Second Stage Pressure Regulator
% Beswick engineering "Ultra-Miniature Single Stage Piston Regulator"
PR_Weight = 0.0062; % Pressure Regulator weight (Beswick)

% calculate config_1: Battery only

wt_1 = zeros(1,6);
ft_1 = zeros(1,6);
wt_1(1) = Battery(1,1);
ft_1(1) = 18;
wt_1(2) = wt_1(1);
ft_1(2) = ((10*Battery(1,2) - 44.9)/1.51) + 18;
wt_1(3) = Battery(2,1);
ft_1(3) = ft_1(2);
wt_1(4) = wt_1(3);
ft_1(4) = ((10*Battery(2,2) - 44.9)/1.51) + 18;
wt_1(5) = Battery(3,1);
ft_1(5) = ft_1(4);
wt_1(6) = wt_1(5);
ft_1(6) = ((10*Battery(3,2) - 44.9)/1.51) + 18;
% calculate battery only lines
slope = zeros(1,3);
intercept = zeros(1,3);
ePt = zeros(2,8);
for i = 1:3
num = Battery(i,1);
193
 t1 = ((10*Battery(i,2) - 44.9)/1.51)+18;
t2 = ((20*Battery(i,2) - 44.9)/1.51)+18;
den = t2 - t1;
slope(i) = num/den;
b = Battery(i,1) - (slope(i)*t1);
intercept(i) = b;
% calculate endpoints
ePt(1,i) = (0.5 - b)/slope(i);
ePt(2,i) = (5.0 - b)/slope(i);
end

% calculate config_2: Hybrid with current assist, H-100 fuel cell, 6Ah battery
wt_2 = zeros(1,6);
ft_2 = zeros(1,6);
wt_2(1) = Battery(1,1) + FC_Weight(2) + Ninja_50(1) + PR_Weight;
ft_2(1) = 18;
wt_2(2) = wt_2(1);
ft_2(2) = ((1.55*Ninja_50(2) + 10*Battery(1,2) - 44.9)/1.51) + 18;
wt_2(3) = Battery(1,1) + FC_Weight(2) + Ninja_68(1) + PR_Weight;
ft_2(3) = ft_2(2);
wt_2(4) = wt_2(3);
ft_2(4) = ((1.55*Ninja_68(2) + 10*Battery(1,2) - 44.9)/1.51) + 18;
wt_2(5) = Battery(1,1) + FC_Weight(2) + Ninja_90(1) + PR_Weight;
ft_2(5) = ft_2(4);
wt_2(6) = wt_2(5);
ft_2(6) = ((1.55*Ninja_90(2) + 10*Battery(1,2) - 44.9)/1.51) + 18;

% calculate config_3: Hybrid with current assist, H-100 fuel cell, 11Ah battery
wt_3 = zeros(1,6);
ft_3 = zeros(1,6);
wt_3(1) = Battery(2,1) + FC_Weight(2) + Ninja_50(1) + PR_Weight;
ft_3(1) = 18;
wt_3(2) = wt_3(1);
ft_3(2) = ((1.55*Ninja_50(2) + 10*Battery(2,2) - 44.9)/1.51) + 18;
wt_3(3) = Battery(2,1) + FC_Weight(2) + Ninja_68(1) + PR_Weight;
ft_3(3) = ft_3(2);
wt_3(4) = wt_3(3);
ft_3(4) = ((1.55*Ninja_68(2) + 10*Battery(2,2) - 44.9)/1.51) + 18;
wt_3(5) = Battery(2,1) + FC_Weight(2) + Ninja_90(1) + PR_Weight;
ft_3(5) = ft_3(4);
wt_3(6) = wt_3(5);
ft_3(6) = ((1.55*Ninja_90(2) + 10*Battery(2,2) - 44.9)/1.51) + 18;

% calculate config_4: Hybrid with current assist, H-100 fuel cell, 17Ah battery
wt_4 = zeros(1,6);
ft_4 = zeros(1,6);
wt_4(1) = Battery(3,1) + FC_Weight(2) + Ninja_50(1) + PR_Weight;
ft_4(1) = 18;
wt_4(2) = wt_4(1);
ft_4(2) = ((1.55*Ninja_50(2) + 10*Battery(3,2) - 44.9)/1.51) + 18;
wt_4(3) = Battery(3,1) + FC_Weight(2) + Ninja_68(1) + PR_Weight;
ft_4(3) = ft_4(2);
wt_4(4) = wt_4(3);
ft_4(4) = ((1.55*Ninja_68(2) + 10*Battery(3,2) - 44.9)/1.51) + 18;
wt_4(5) = Battery(3,1) + FC_Weight(2) + Ninja_90(1) + PR_Weight;
194
ft_4(5) = ft_4(4);
wt_4(6) = wt_4(5);
ft_4(6) = ((1.55*Ninja_90(2) + 10*Battery(3,2) - 44.9)/1.51) + 18;

% calculate config_5: Hybrid with current assist, AST01-01 fuel cell, 6Ah battery
wt_5 = zeros(1,6);
ft_5 = zeros(1,6);
wt_5(1) = Battery(1,1) + FC_Weight(1) + Ninja_50(1) + PR_Weight;
ft_5(1) = 18;
wt_5(2) = wt_5(1);
ft_5(2) = ((1.55*Ninja_50(2) + 10*Battery(1,2) - 44.9)/1.51) + 18;
wt_5(3) = Battery(1,1) + FC_Weight(1) + Ninja_68(1) + PR_Weight;
ft_5(3) = ft_5(2);
wt_5(4) = wt_5(3);
ft_5(4) = ((1.55*Ninja_68(2) + 10*Battery(1,2) - 44.9)/1.51) + 18;
wt_5(5) = Battery(1,1) + FC_Weight(1) + Ninja_90(1) + PR_Weight;
ft_5(5) = ft_5(4);
wt_5(6) = wt_5(5);
ft_5(6) = ((1.55*Ninja_90(2) + 10*Battery(1,2) - 44.9)/1.51) + 18;

% calculate config_6: Hybrid with current assist, AST01-01 fuel cell, 11Ah battery
wt_6 = zeros(1,6);
ft_6 = zeros(1,6);
wt_6(1) = Battery(2,1) + FC_Weight(1) + Ninja_50(1) + PR_Weight;
ft_6(1) = 18;
wt_6(2) = wt_6(1);
ft_6(2) = ((1.55*Ninja_50(2) + 10*Battery(2,2) - 44.9)/1.51) + 18;
wt_6(3) = Battery(2,1) + FC_Weight(1) + Ninja_68(1) + PR_Weight;
ft_6(3) = ft_6(2);
wt_6(4) = wt_6(3);
ft_6(4) = ((1.55*Ninja_68(2) + 10*Battery(2,2) - 44.9)/1.51) + 18;
wt_6(5) = Battery(2,1) + FC_Weight(1) + Ninja_90(1) + PR_Weight;
ft_6(5) = ft_6(4);
wt_6(6) = wt_6(5);
ft_6(6) = ((1.55*Ninja_90(2) + 10*Battery(2,2) - 44.9)/1.51) + 18;

% calculate config_7: Hybrid with current assist, AST01-01 fuel cell, 17Ah battery
wt_7 = zeros(1,6);
ft_7 = zeros(1,6);
wt_7(1) = Battery(3,1) + FC_Weight(1) + Ninja_50(1) + PR_Weight;
ft_7(1) = 18;
wt_7(2) = wt_7(1);
ft_7(2) = ((1.55*Ninja_50(2) + 10*Battery(3,2) - 44.9)/1.51) + 18;
wt_7(3) = Battery(3,1) + FC_Weight(1) + Ninja_68(1) + PR_Weight;
ft_7(3) = ft_7(2);
wt_7(4) = wt_7(3);
ft_7(4) = ((1.55*Ninja_68(2) + 10*Battery(3,2) - 44.9)/1.51) + 18;
wt_7(5) = Battery(3,1) + FC_Weight(1) + Ninja_90(1) + PR_Weight;
ft_7(5) = ft_7(4);
wt_7(6) = wt_7(5);
ft_7(6) = ((1.55*Ninja_90(2) + 10*Battery(3,2) - 44.9)/1.51) + 18;

% calculate config_8: H-500 fuel cell only

wt_8 = zeros(1,6);
ft_8 = zeros(1,6);
195
wt_8(1) = FC_Weight(3) + Ninja_50(1) + PR_Weight;
ft_8(1) = 18;
wt_8(2) = wt_8(1);
ft_8(2) = ((1.55*Ninja_50(2) - 44.9)/1.51) + 18;
wt_8(3) = FC_Weight(3) + Ninja_68(1) + PR_Weight;
ft_8(3) = ft_8(2);
wt_8(4) = wt_8(3);
ft_8(4) = ((1.55*Ninja_68(2) - 44.9)/1.51) + 18;
wt_8(5) = FC_Weight(3) + Ninja_90(1) + PR_Weight;
ft_8(5) = ft_8(4);
wt_8(6) = wt_8(5);
ft_8(6) = ((1.55*Ninja_90(2) - 44.9)/1.51) + 18;

% calculate config_9: Hybrid with no current assist, H-100 fuel cell, 6Ah battery
% using a 220g DC-to-DC converter (Conventional Hybrid)
wt_9 = zeros(1,6);
ft_9 = zeros(1,6);
wt_9(1) = Battery(1,1) + FC_Weight(2) + Ninja_50(1) + PR_Weight + DCC_Weight;
ft_9(1) = 18;
wt_9(2) = wt_9(1);
ft_9(2) = ((1.55*Ninja_50(2) - 44.9)/1.51) + 18;
wt_9(3) = Battery(1,1) + FC_Weight(2) + Ninja_68(1) + PR_Weight + DCC_Weight;
ft_9(3) = ft_9(2);
wt_9(4) = wt_9(3);
ft_9(4) = ((1.55*Ninja_68(2) - 44.9)/1.51) + 18;
wt_9(5) = Battery(1,1) + FC_Weight(2) + Ninja_90(1) + PR_Weight + DCC_Weight;
ft_9(5) = ft_9(4);
wt_9(6) = wt_9(5);
ft_9(6) = ((1.55*Ninja_90(2) - 44.9)/1.51) + 18;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% plot flight time versus weight
plot(ft_1, wt_1, 'LineWidth', 2, 'Marker', '*', 'Color', cm(1,:));
hold on;
plot(ft_2, wt_2, 'LineWidth', 2, 'Marker', '*', 'Color', cm(2,:));
plot(ft_3, wt_3, 'LineWidth', 2, 'Marker', '*', 'Color', cm(3,:));
plot(ft_4, wt_4, 'LineWidth', 2, 'Marker', '*', 'Color', cm(4,:));

plot(ft_5, wt_5, 'LineWidth', 2, 'LineStyle', '--', 'Marker', 'd', 'Color', cm(2,:));

plot(ft_6, wt_6, 'LineWidth', 2, 'LineStyle', '--', 'Marker', 'd', 'Color', cm(3,:));
plot(ft_7, wt_7, 'LineWidth', 2, 'LineStyle', '--', 'Marker', 'd', 'Color', cm(4,:));

plot(ft_8, wt_8, 'LineWidth', 2, 'Marker', '+', 'Color', cm(5,:));

plot(ft_9, wt_9, 'LineWidth', 3, 'Marker', 'o', 'Color', cm(6,:));
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Plot battery lines
yb = zeros(1,10);
for i = 1:3
xb = linspace(ePt(1,i),ePt(2,i), 10);
yb = slope(i)*xb + intercept(i);
plot(xb,yb, '-', 'color', [120 100 80]*(1/255));
end
% Plot points that represent number of batteries
% on each battery extrapolation line.
SixAh = zeros(2,8);
196
wt_pt = Battery(1,1);
for i = 1:8
SixAh(1,i) = ((i*10*Battery(1,2) - 44.9)/1.51)+18;
SixAh(2,i) = i * wt_pt;
end
plot(SixAh(1,:), SixAh(2,:), 'o');

ElevenAh = zeros(2,5);
wt_pt = Battery(2,1);
for i = 1:5
ElevenAh(1,i) = ((i*10*Battery(2,2) - 44.9)/1.51)+18;
ElevenAh(2,i) = i * wt_pt;
end
plot(ElevenAh(1,:), ElevenAh(2,:), 'o');

SeventeenAh = zeros(2,3);
wt_pt = Battery(3,1);
for i = 1:3
SeventeenAh(1,i) = ((i*10*Battery(3,2) - 44.9)/1.51)+18;
SeventeenAh(2,i) = i * wt_pt;
end
plot(SeventeenAh(1,:), SeventeenAh(2,:), 'o');

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
hold off;
title('Power System Weight versus Flight Time');
xlabel('Flight Time (minutes)');
ylabel('Power System Weight (kg)');
set(gca, 'FontSize', 18);
lh = legend('Battery Only', ...
'Hybrid: H-100, 6Ah Battery (Current Assist)', ...
'Hybrid: H-100, 11Ah Battery (Current Assist)', ...
'Hybrid: H-100, 17Ah Battery (Current Assist)', ...
'Hybrid: AST01-01, 6AhBattery (Current Assist)', ...
'Hybrid: AST01-01, 11AhBattery (Current Assist)', ...
'Hybrid: AST01-01, 17AhBattery (Current Assist)', ...
'H-500 Only', ...
'DC converter Hybrid: H-100, 6Ah Battery (No Current Assist)', ...
'Location', 'southeast');
grid;
set(gcf, 'color', 'w');
set(findall(gcf, 'type', 'text'), 'FontSize', 18);
set(lh, 'FontSize', 18);

197