Power Grid Integration of GigaWatt-

scale Hydrogen Production

Master of Science Thesis

ENRIQUE VAN DYKE

Department of Electrical Engineering

Division of Electric Power Engineering
Chalmers University of Technology
Gothenburg, Sweden 2024
www.chalmers.se
 Master’s thesis 2024

POWER GRID INTEGRATION OF

GIGAWATT-SCALE HYDROGEN PRODUCTION

ENRIQUE VAN DYKE

Department of Electrical Engineering

Division of Electric Power Engineering
Chalmers University of Technology
Gothenburg, Sweden 2024
Design and analysis of industrial power plant with sets of electrolyzers that reach GigaWatt-
scale

ENRIQUE VAN DYKE

© ENRIQUE VAN DYKE , 2024.

Supervisor: Tarik Dervišić, Hitachi Energy

Examiner: Massimo Bongiorno, Electrical Engineering Department

Master’s Thesis 2024

Department of Electrical Engineering
Division of Electric Power Engineering
Chalmers University of Technology
SE-412 96 Gothenburg
Telephone +46 31 772 1000

Typeset in LATEX, template by Kyriaki Antoniadou-Plytaria

Printed by Chalmers Reproservice
Gothenburg, Sweden 2024

iv
Power Grid Integration of GigaWatt-scale Hydrogen Production
ENRIQUE VAN DYKE
Department of Electrical Engineering
Division of Electric Power Engineering
Chalmers University of Technology

Abstract
Climate change is one of the most pressing issues that humanity faces today. The urgent
need to mitigate green-house gas emissions has triggered an unprecedented expansion
of renewable energy sources, consequently, creating an opportunity to revolutionize our
energy systems. In this context, green hydrogen emerges as a potential candidate to
decarbonize hard-to-electrify industries, such as iron & steel, high-temperature heat ap-
plications, petrochemical, transport and aerospace sector.

This thesis endeavors to explore the technical feasibility of gigawatt scale hydrogen pro-
duction by examining a proposal of plant design, implementing a model for electrolyzer
systems available in the market and selecting an AC/DC topology for the rectifier.

The elected station is connected to the HVAC transmission grid through a step-down
distribution system with two voltage levels, 400 and 33 kV. This grid is assumed to have a
high short circuit capacity given the high demand of power and its design needs to comply
with grid code requirements and international standards (IEC, IEEE).

The driver for standardization and cost savings, while matching electrolyzer’s technology
with adequate electrical infrastructure, led to the decision of implementing alkaline water
electrolyzers. Besides, an accurate model definition required a solid understanding of the
electrical and thermal fundamentals of electrolysis.

The electrolyzer voltage and current requirements defined the criteria for selecting the
converter’s topology. The final choice was the 12-pulse thyristor rectifier given its high
efficiency, reliability and good control of the current. Low-order harmonic cancellation
excluding those harmonic orders with 12h±1 with h = 1, 2, 3 ... was an attractive feature,
but required the use of tuned-LC filters for the 11th-13th and 23rd-25th harmonics.

A look-up table is implemented for the electrolyzer model next with a per-unit current
controller for the 12-pulse thyristor rectifier. This set-up makes it easily reimplementable
in future projects, since it enables the implementation of this controller for any electrolyzer
system, independently of its specifications. All models and simulations are performed us-
ing the power system analysis software DIgSILENT PowerFactory.

Finally, the correct rating of different components used in the station and grid code
compliance was verified performing different simulations: a load flow analysis, a dynamic
ramp-up simulation to nominal capacity and lastly transient analysis of the station in
spontaneous events like load rejection.

Keywords: electrolyzer, alkaline, model, hydrogen, GW-scale, 12-PTR, station

v
Acknowledgements
First and foremost, I would like to express my sincere gratitude to my supervisor Tarik
Dervišić for his dedication and support during the realization of this thesis. Also, for
his invaluable patience and feedback. Additionally, this endeavor would not have been
possible without my examiner Massimo Bongiorno, who generously provided feedback and
his expertise in the subject.
I would also like to thank Björn Wickman for an insightful discussion on the electric
properties of fuel cells and electrolyzers, which was ultimately key in helping me build
the electrolyser modelling. I am also grateful to my colleague, Shemsedin who provided
guidance, introduced me to the PowerFactory modelling tool and provided moral support.
Special thank you to my work managers Joachim Andersson, Fredrik Sjögren and Leila
Manshaei who were always helpful and kindly offered their time to discuss a myriad of
topics but above all inspired me and provided personal advice.
A big shoutout to my whole Hitachi Energy’s Power Consulting team in Gothenburg,
especially Emily Baruffaldi and Kelly Antoniadou, without whom this would not have
been possible.
Many thanks to my family and friends. To my parents, for their invaluable guidance and
belief in me, my grandmother for her kind spirit and my sister Graciela for her priceless
support, enthusiasm and editing skills. To my friends Lander Vallejo and Gaizka Barrasa
for their late night feedback sessions and empathy. Lastly, I want to express my heartfelt
gratitude to Laura Martinez de Guinea, who stood by my side during the final completion
of this thesis. I cannot emphasize enough how much I owe you for your kind support and
help.

Enrique Van Dyke, Gothenburg, August 2024

vii
 Contents

List of Figures xi

List of Tables xiii

List of Acronyms xiv

Nomenclature xvii

1 Introduction 1
1.1 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Societal, ethical and ecological aspects . . . . . . . . . . . . . . . . . . . . 5
1.5 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 State of the art 7

2.1 Alkaline Water Electrolyzer (AWE) . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Proton Exchange Membrane Electrolyzer (PEM) . . . . . . . . . . . . . . . 10
2.3 Economic and Performance Evaluation . . . . . . . . . . . . . . . . . . . . 11
2.3.1 Investment costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.2 Operating conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.3 Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Thermodynamic and Electrochemical Analyses 15

3.1 Thermodynamics applied to chemical reactions . . . . . . . . . . . . . . . . 15
3.2 Reversible and Thermoneutral Potentials . . . . . . . . . . . . . . . . . . . 18
3.3 Polarizations and efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.1 Activation losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.2 Ohmic losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.3 Concentration losses . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Faraday’s Law of Electrolysis . . . . . . . . . . . . . . . . . . . . . . . . . 23

4 Alkaline Water Electrolyzer Modeling in PowerFactory 27

4.1 Model definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Power increase from kW electrolyzer to MW-level . . . . . . . . . . . . . . 30
4.3 Dynamic Model Implementation in PowerFactory . . . . . . . . . . . . . . 31

ix
Contents

5 Converter-load interaction and control 35

5.1 Prestudy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.2 Power converters in PowerFactory . . . . . . . . . . . . . . . . . . . . . . . 37
5.3 6 and 12 pulse three phase thyristor-based rectifiers . . . . . . . . . . . . . 38
5.4 Controller Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.4.1 Process model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.4.2 Controller design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.4.3 System analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.4.4 Power consumption analysis . . . . . . . . . . . . . . . . . . . . . . 48

6 Structure Design for Electrolyzer stations at GW-scale 51

6.1 Station Design Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.2 Model implementation of different structure designs . . . . . . . . . . . . . 55
6.3 Demand Connection Code . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.3.1 Harmonics and Distortion . . . . . . . . . . . . . . . . . . . . . . . 57
6.3.2 Power Factor Correction . . . . . . . . . . . . . . . . . . . . . . . . 58
6.3.3 Voltage Regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.4 Power Plant Electrical Components Design . . . . . . . . . . . . . . . . . . 59
6.4.1 Grid Characterization . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.4.2 Transformers Rating Estimation . . . . . . . . . . . . . . . . . . . . 60
6.4.3 Harmonic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.4.4 Reactive Power Compensation . . . . . . . . . . . . . . . . . . . . . 65
6.4.5 STATCOM Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

7 Simulation cases 71
7.1 Load flow analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.2 Ramp-up simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.3 Load Rejection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
7.3.1 Partial Load Rejection . . . . . . . . . . . . . . . . . . . . . . . . . 78
7.3.2 Full Load Rejection . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

8 Conclusions 83
8.1 Results from present work . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
8.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

x
 List of Figures

2.1 Electrolyzer classification according to electrolyte and operating temperature. 8

2.2 AWE cell working principle [37]. . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Typical AWE system architecture [44]. . . . . . . . . . . . . . . . . . . . . 9
2.4 PEM cell working principle [37]. . . . . . . . . . . . . . . . . . . . . . . . . 10
2.5 Typical PEM electrolyser system architecture [44]. . . . . . . . . . . . . . . 10
2.6 Overvoltage for AWE and PEM in the same current density range [42]. . . 12

3.1 Control volume of the electrolyzer [51]. . . . . . . . . . . . . . . . . . . . . 15

3.2 Reversible and thermoneutral potential versus temperature at atmospheric
pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3 Effects of operating pressure and temperature on the reversible voltage. . . 20
3.4 Pressure and temperature effect on reversible and thermoneutral potentials
[53]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5 Cell voltage dependence on temperature and pressure [54]. . . . . . . . . . 23
3.6 Test schematics, applied currents and cathode’s pressure [59]. . . . . . . . . 24
3.7 Lumped Parameter Circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.8 Parasitic current distribution and individual cell voltage in a thirty-cell
stack [60]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4.1 Characteristic IU curves and Voltage Efficiency. . . . . . . . . . . . . . . . 29

4.2 Faraday Efficiency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 Scale-up from kW to MW level electrolyzer. . . . . . . . . . . . . . . . . . 31
4.4 AWE system composite frame (BlkDef. Test Voltage Source). . . . . . . . 33
4.5 AWE IU curves block (BlkDef. IU curves). . . . . . . . . . . . . . . . . . . 34
4.6 Hydrogen production block (BlkDef. Faraday Law). . . . . . . . . . . . . . 34

5.1 Active Components Ratings in Power Electronics [67]. . . . . . . . . . . . . 35

5.2 PWM-CSRs [68]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.3 Three-level Neutral Point Clamping converter. . . . . . . . . . . . . . . . . 37
5.4 Power Electronic Devices available in PowerFactory. . . . . . . . . . . . . . 38
5.5 PowerFactory three-phase LCC rectifier model. . . . . . . . . . . . . . . . . 39
5.6 Normalized DC voltage Vd /Vd0 as function of the firing angle α assuming
no commutation inductance Lc [71]. . . . . . . . . . . . . . . . . . . . . . . 40
5.7 Phase voltages, currents and DC voltage of a 6-PTR working with a firing
angle α=30o and a commutation angle µ=20o . . . . . . . . . . . . . . . . . 40
5.8 12-pulse thyristor rectifier [70]. . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.9 Harmonic distortion distribution for 6 and 12-PTR topologies. . . . . . . . 42
5.10 Analogy of electrolyzer with resistance and voltage source. . . . . . . . . . 43

xi
List of Figures

5.11 Current controller for RL circuit. . . . . . . . . . . . . . . . . . . . . . . . 44

5.12 Introduction of feed-forward D̂ in the current controller. . . . . . . . . . . 44
5.13 Proportional-integral current controller with feed-forward and anti-windup
limiter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.14 Current controller block definition (BlkDef. Current Control). . . . . . . . 45
5.15 Time-domain response of the system with different PI parameters. . . . . . 47
5.16 Time-domain response of the system with different operating temperatures. 47
5.17 Stack voltage Vpu and firing angle α at different temperatures. . . . . . . . 48
5.18 Grid single bus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.19 Active PAC and reactive QAC power consumption as function of loading level. 49
5.20 Line current wave-forms in a 6-PTR as function of firing angle α assuming
no commutation reactance [71]. . . . . . . . . . . . . . . . . . . . . . . . . 50
5.21 Efficiency Chart. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

6.1 Distribution grids for an electrolyzer station comprised of system modules

[56]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6.2 Electrolyzer station base line diagram [56]. . . . . . . . . . . . . . . . . . . 53
6.3 Selected SLD for station design. . . . . . . . . . . . . . . . . . . . . . . . . 54
6.4 A485 and A3880 modules [44]. . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.5 A3880 4s2p electrolyzer system. . . . . . . . . . . . . . . . . . . . . . . . . 55
6.6 Grouping electrolyzer systems into electrolyzer modules. . . . . . . . . . . 56
6.7 Current harmonic distortion distribution at 33kV and 400kV busbar without
filtering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.8 Harmonic Filtering [79]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.9 Single-tuned filter [80]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.10 Frequency sweep of filter impedance. . . . . . . . . . . . . . . . . . . . . . 64
6.11 Current harmonic distortion distribution at 33kV and 400kV busbar with
filtering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.12 Voltage harmonic distortion distribution at PCC with filtering. . . . . . . . 65
6.13 Hybrid system composed of STATCOM, passive filter, 12-PTR and elec-
trolyzer [83]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.14 Preliminary load flow calculation. . . . . . . . . . . . . . . . . . . . . . . . 67
6.15 STATCOM Single Line Diagram and Composite Frame. . . . . . . . . . . . 68
6.16 STATCOM Voltage Control Block Diagram. . . . . . . . . . . . . . . . . . 69

7.1 Ramp-up power flow in the External Grid. . . . . . . . . . . . . . . . . . . 74

7.2 Ramp-up External Grid excitation voltage and PCC voltage. . . . . . . . . 74
7.3 Ramp-up STATCOM load flow, current and voltage in p.u. . . . . . . . . . 75
7.4 Ramp-up power flow in a three-winding 400/33kV transformer. . . . . . . . 76
7.5 Ramp-up power flow in adjacent electrolyzer modules and voltage at 33 kV
busbar. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
7.6 Ramp-up electrolyzer dynamic response. . . . . . . . . . . . . . . . . . . . 77
7.7 Power flow in the External Grid under partial load rejection . . . . . . . . 78
7.8 External Grid excitation voltage and PCC voltage under partial load re-
jection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
7.9 STATCOM load flow, current and voltage in p.u. under partial load rejection. 79
7.10 Power flow in the External Grid under total load rejection. . . . . . . . . . 80
7.11 External Grid excitation voltage and PCC voltage under total load rejection. 80
7.12 STATCOM load flow, current and voltage in p.u. under total load rejection. 81

xii
 List of Tables

2.1 Water Electrolysis Technology Description [37],[38],[39]. . . . . . . . . . . . 8

2.2 Performance comparison [38], [37]. . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Performance comparison [38], [37]. . . . . . . . . . . . . . . . . . . . . . . . 14

3.1 Sign criteria for heat transfer and work. . . . . . . . . . . . . . . . . . . . . 16

3.2 Thermodynamic variables. . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 Thermodynamic properties for liquid water, hydrogen and oxygen under
standard conditions [51]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.4 Impact of an adiabatic system and reversible process on electrolysis. . . . . 18
3.5 Enthalpy for water, hydrogen and oxygen at atmospheric pressure versus
temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4.1 PHOEBUS AWE operation ratings and design specifications [53]. . . . . . 27

4.2 PHOEBUS AWE I-U curves and Faraday Efficiency parameters [43],[62]. . 28
4.3 Series A Nel Hydrogen Atmospheric Alkaline Water Electrolyser [44]. . . . 30
4.4 Deductions of Series A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.5 A485 assumptions and operation ratings. . . . . . . . . . . . . . . . . . . . 31

5.1 Implemented Harmonic Distortion in 12-PTR [72]. . . . . . . . . . . . . . . 42

5.2 Total Harmonic Disorder (THD) for 6 and 12 PTR topologies. . . . . . . . 42
5.3 Candidates for final controller. . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.4 Rise time at ambient and operating temperature. . . . . . . . . . . . . . . 48

6.1 Structure comparison for cases 1-4 [56]. . . . . . . . . . . . . . . . . . . . . 53

6.2 Current distortion limits for systems rated above 161 kV IEEE Std 519-
2014 [76]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.3 Individual harmonic order (odd harmonics) IEC TR 61000-3-6 [77]. . . . . 57
6.4 Grid Characterization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.5 Transformers Rating Sheet. . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.6 Specifications for 11th and 23rd tuned band-pass filters at 33kV busbar. . . 63
6.7 Preliminary specifications for STATCOM and filters. . . . . . . . . . . . . 67
6.8 STATCOM controller parameter definition. . . . . . . . . . . . . . . . . . . 69
6.9 PLL parameter definition. . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

7.1 External Grid load flow with 400 kV nominal voltage. . . . . . . . . . . . . 72

7.2 Three-winding transformers load flow results at 400 kV terminal. . . . . . . 72
7.3 11th and 23rd harmonic filters. . . . . . . . . . . . . . . . . . . . . . . . . . 72
7.4 STATCOM reactive power flow and 33kV busbar voltage. . . . . . . . . . . 73
7.5 Ramp-up sequence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

xiii
xiv
 List of Acronyms

Below is the list of acronyms that have been used throughout this thesis listed in alpha-
betical order:

AEM Anion Exchange Membrane

AWE Alkaline Water Electrolyzer
BP Band Pass
CAPEX Capital Expenditure
CCUS Carbon Capture, Utilization and Storage
CHP Combined Heat and Power
CV Control Volume
DCC Demand Connection Code
GDL Gas Diffusion Layer
GW Giga Watt
H2 Hydrogen
HTE High Temperature Electrolyzer
HVAC High Voltage Alternating Current
HVDC High Voltage Direct Current
IEA International Energy Agency
IEC International Electrotechnical Commission
IEEE Institute of Electrical and Electronics Engineers
IGBT Insulated-Gate Bipolar Transistor
IMC Internal Model Controller
kW Kilo Watt
LCOE Levelized Cost of Electricity
LCOH Levelized Cost of Hydrogen
MEA Membrane Electrode Assembly
MW Mega Watt
PCC Point of Common Coupling
PEM Proton Exchange Membrane
PFSA Perfluorosulfonic acid
PF Power Factor
PI Proportional-Integral
PLL Phase-Locked Loop
PTL Porous Transport Layer
POX Partial Oxidation
PTR Pulse Thyristor Rectifier

xv
PWM Pulse Width Modulation
RMS Root Mean Square
SLD Single line diagram
SMR Steam Methane Reforming
SPE Solid Polymer Electrolyte
SOE Solid Oxide Electrolyzer
STATCOM Static Synchronous Compensator
SVC Static VAr Compensator
TDD Total Demand Distortion
THD Total Harmonic Distortion
VSC Voltage Source Converter
 Nomenclature

Below is the nomenclature of parameters and variables that have been used throughout
this thesis.

Variables
Er , Eth Reversible and Thermoneutral Potential V
Ecell , Eloss Cell voltage and over-voltage V
T Temperature K
P Pressure bar
∆h Net specific enthalpy kJ mol-1
∆g Net specific Gibbs energy kJ mol-1
∆s Net specific entropy J (mol K)-1
hof Standard enthalpy of formation kJ mol-1
h Specific enthalpy (T , P ) kJ mol-1
ho Standard specific enthalpy kJ mol-1
s Specific entropy (T , P ) J (mol K)-1
sgen Specific generated entropy J (mol K)-1
qCV Specific heat transfer kJ mol-1
wCV Specific work kJ mol-1
z Number of electron moles -
F Faraday constant C mol-1
R Universal gas constant J (mol · K)-1
∆N Net number of gaseous moles -
Jcell Cell current density A m-2
J0,an , J0,ca Anode and cathode exchange current density A m-2
ηF Faraday or current efficiency -
ṁH2 , ṁideal,H2 Real and ideal hydrogen mass flow kg/s
MH2 Hydrogen molecular weight kg mol-1
nseries Number of electron moles -
r1 Ohmic parameter 1 Ω m2
r2 Ohmic parameter 2 Ω m2 /o C
s Coefficient of electrode overvoltage V
t1 Overvoltage coefficient 1 m2 /A
t2 Overvoltage coefficient 2 m2 o C/A
t3 Overvoltage coefficient 3 m2 o C2 /A
a1 Faraday efficiency coefficient 1 %

xvii
 a2 Faraday efficiency coefficient 2 A/m2
a3 Faraday efficiency coefficient 3 A (m2 o C)-1
a4 Faraday efficiency coefficient 4 A2 /m4
a5 Faraday efficiency coefficient 5 A2 (m4 o C)-1
α Firing angle rad
Lc Commutation inductance mH
Vd0 Ideal no-load DC voltage V
∆Vd Voltage drop in the commutation reactance V
Vd 6-PTR DC voltage V
VLL RMS line to line voltage V
Id 6-PTR DC current A
IˆA(1) Peak fundamental current A
IˆA(h) Peak current for its corresponding harmonic order A
h Harmonic order -
Rload Electrolyzer equivalent resistance Ω
rload Electrolyzer equivalent resistance in per unit pu
Lload Electrolyzer equivalent inductance mH
lload Electrolyzer equivalent inductance in per unit pu
Rbase Electrolyzer base resistance Ω
trise Rise time s
KP Controller integration constant -
KI Controller proportional constant -
αc Bandwidth rad/s
iref Reference current pu
Vbus Bus voltage of the electrolyzer systems kV
In,bus Nominal bus current of electrolyzer systems kA
ISCC Short-circuit current kA
Sn,T Rated apparent power of transformers MVA
Nsys,total Total number of electrolyzer systems -
Nsys,branch Number of electrolyzer systems per branch -
Pn,station Total station active power capacity MW
Pn,sys Individual active power demand of each electro- MW
lyzer system
Sn,branch Apparent power demand from each transformer MVA
branch
Sn,sys Individual apparent power demand of each electro- MVA
lyzer system
′′
Sk Grid’s short-circuit capacity MVA
′′
Ik Grid’s short-circuit current kA
Z̄1 Grid’s Thevenin positive sequence impedance Ω
R1 Grid’s Thevenin resistance Ω
X1 Grid’s Thevenin reactance Ω
Unom Grid’s nominal voltage kV
Sn,2T R Rated apparent power of 33/3.11kV transformers MVA
Sn,3T R Rated apparent power of 400/33kV 3-winding MVA
transformers
cos φsys Electrolyzer system power factor -
Qf Filter rated reactive power MVAr

xviii
Vf Filter rated voltage kV
fr Filter resonant frequency Hz
q Filter quality factor -
XL , XC Filter inductive and capacitive reactances Ω
Lf , Cf , Rf Filter inductance, capacitance and resistance mH,µF,Ω
Qmax , Qmin Station maximum and minimum reactive power MVAr
demand
Qst STATCOM rated reactive power MVAr
Q3T R,max Maximum reactive power flow through 400/33kV MVAr
3-winding transformer
Q3T R,min Minimum reactive power flow through 400/33kV MVAr
3-winding transformer

xix
xx
 1
Introduction

Climate change is one of the biggest challenges that humanity faces in the 21st cen-
tury. Continuous escalation of greenhouse gas (GHG) emissions caused by technology
development, population growth and well-being improvement since the First Industrial
Revolution has become a global threat to international security and stability [1].

Although the first world conference about the environment was in 1972 held by the United
Nations in Stockholm [2], it was not until 1997 when the first international agreement
for reducing GHG emissions was adopted: the Kyoto Protocol. In this treaty [3], the
different parties recognized that global warming was happening and that human-source
CO2 emissions were causing it. This delay was caused by disinformation, uncertainty and
conflicts of interest in the matter. For example, in the 70’s ExxonMobil publicly misled
climate science while simultaneously having internal assessments forecasting a continuous
increase in global temperatures [4], [5]. After that came the well-known Paris Agreement
in 2015, where all parties committed to reduce their emissions with differences between
developing and developed countries. The headline of the agreement was to avoid a global
temperature increase below 2°C, preferably 1.5°C, compared to pre-industrial levels and
reaching a zero emissions scenario by 2050 [6].

Since then, some major events happened like USA withdrawal from Paris Agreement
in 2017 during Donald Trump’s Presidency [7] and its return in 2021 with Joe Biden
[8]. COVID19 prompted a temporary reduction in carbon emissions [9]. In 2022, Russia’s
invasion of Ukraine accelerated the energy crisis in Europe, pressing on oil and gas supplies
already stressed due to the fast recovery of the economy post-pandemic [10] and finally,
pending to see the effects of the Israel-Gaza conflict currently underway.

China has led the growth in energy demand worldwide during the last two decades with its
subsequent repercussion in CO2 emissions. However in 2023, China’s population growth
came to a halt [11] whilst India will surpass it by 2025 [12]. This shift joined with India’s
industrialization represents an opportunity to reduce CO2 emissions as India increases its
share of renewables (35% by 2030, 15% PV solar[13]) and adopts cleaner energy practices
and sustainable development strategies.

The Paris Agreement led to increased investment in renewable energy and energy effi-
ciency measures, including electrification of transportation, industry, and buildings. This
electrification trend led to a wave of deployment of electric vehicles and the growth of
renewable energy sources such as wind and solar power, which has contributed actively to
reduce GHG emissions. However, there are specific sectors and activities where emissions
are significant and electrification is not always the way to go, such as agriculture and the

1
1. Introduction

chemical, cement, steel and aerospace industries. This is where hydrogen comes in and
plays a key role in the green and sustainable energy system of the future.

According to the International Energy Agency (IEA) [14], electricity and heat generation
accounted for roughly 40% of total emissions in 2022, transport 23%, industry 23%,
buildings 10% and other activities 5%. From industry [15], iron & steel account for 28%,
cement 27% and petrochemical 15%. Hydrogen has a great potential to substitute fossil
fuels in many of these sectors such as iron & steel, petrochemical and transport.

Global hydrogen demand reached 94 megatonnes (Mt) in 2021 where 42% was employed
in refining, 36% in ammonia production, 16% in methanol production and 6% in iron &
steel production [16]. Unfortunately, as it is shown in the “current status of hydrogen
production” of the Global Hydrogen Review 2022 [17], natural gas is the main route
to produce hydrogen via steam methane reforming (SMR) representing 62%. The second
major contributor is coal via Partial Oxidation (POX) accounted for 19%, practice mainly
done in China, and lastly hydrogen as by-product of reforming at refineries 18%.

Overall, 99% of all hydrogen produced in 2021 came from fossil-fuel sources, while low-
emission hydrogen only accounted for 1 Mt (0.7%), from which almost all was fossil
fuels with carbon capture, utilization and storage (CCUS), with only 35 kilotons (kt) of
hydrogen produced from electricity via water electrolysis. The reason why hydrogen is
almost entirely produced using fossil fuels is because it is extremely cheap. According to
a publication made by S&P [18], the actual cost of fossil fuel-based hydrogen is around
1.80$/kg. Blue hydrogen produced with SMR and CCUS is around 2.40 $/kg. By the
other hand, green hydrogen produced with renewable resources costs between 3 $/kg and
6.55 $/kg, according to the European Commission hydrogen strategy [19].

Consequently, the main challenge for water electrolysis is to reduce its levelized cost of
production (LCOH). Three key factors are necessary in order to produce cheap green
hydrogen: reduce the capital expenditure (CAPEX) of the electrolyzer plant, improve
electrolyzer stack efficiency and most importantly to use renewable energy with low lev-
elized cost of electricity (LCOE).

The great need for sustainable energy storage solutions to replace high energy density fossil
fuels is another factor that drives a lot of attention. Green hydrogen is an energy vector
with very desirable properties such as high energy density, versatility and abundance.

There are several ways of producing hydrogen. The most popular methods have already
been mentioned, but there are two other technologies to produce sustainable hydrogen
that are worth mentioning since they could help expand hydrogen infrastructure: Plasma
Enhanced Gasification [20] and High Temperature Gas-cooled Reactor [21].

Today’s total energy supply according to World Energy Outlook 2022 developed by IEA
[13] is 624.2 EJ. This value is expected to increase by 18.55% by 2050 following Stated
Policies Scenario (STEPS) scenario. From this increase, the International Renewable
Energy Agency (IRENA) estimates that in order to keep a 1.5°C scenario in 2050, 10%
of the final energy demand must be supplied by hydrogen accounting for 74 EJ per year
[22], equivalent to 600 Mt of hydrogen per annum, 6 times todays’ hydrogen supply. This
escalation is already being pushed forward by the energy crisis 2021-2023, which promises
to become a historic turning point towards a cleaner and more secure energy system [23].

2
 1. Introduction

The main contributors for hydrogen escalation are growth in policy support boosted by
Europe’s post–COVID-19 recovery plan, and increase in gas natural prices boosted by
Russia’s invasion of Ukraine [10]. These 2 main factors provide economic, climate and
security arguments to invest in cost-competitive and sustainable technologies, creating a
positive outcome for humanity and the environment.

At today’s date the largest electrolyzer hydrogen production site is being constructed in
China by Sinopec with a total production capacity of 300 MW, connected to a photovoltaic
power plant. This plant will be able to produce 20 kt of H2 per year after its entry in
operation in June 2023 [24]. By contrast the biggest single stack electrolyzer constructed
so far is commissioned by HydrogenPro in Herøya, Norway, capable of producing 850 tons
of H2 per year [25], with an equivalent power input of 5 MW.

1.1 Aim
The main objective of this master thesis is to design and model a GW-scale electrolyzer
station using a single electrolyzer technology as well as perform steady-state and dynamic
simulations to ensure adequate, safe, reliable and uninterruptible hydrogen supply.

The electrochemical principals behind hydrogen electrolyzers are reviewed in order to

develop a solid virtual twin of the device. The basic load of the station is a 1.9 MW
electrolyzer, which is connected into modules, forming bigger loads and enabling a GW-
scale plant. As any other electrified industry, the station is connected to the transmission
grid, so grid code compliance is considered in the design.

1.2 Problem Description

Green hydrogen production anticipates to grow considerably in the coming years. The
objective of covering 12% of 2050 total energy supply with sustainable hydrogen stated
by IRENA [22] implies the installed capacity of electrolyzers to increase from 510 MW
today to 720 GW in 2030 and 3670 GW in 2050 according to IEA [13]. This expansion
is expected to be accompanied with its corresponding enlargement in renewable energy
installed capacity [22].

If these projections were accomplished, it means a rise of annual manufacturing capacity

production to at least 60 GW by 2030. Based on these assessments, GW-scale electrolyzer
stations can become a reality in the coming years. Thus, it is of interest to get a deeper
understanding of large-scale electrolyzer stations and analyze their technical feasibility.

Besides, these power stations must meet grid code requirements while at the same time
be competitive from an economical point of view. Hence the purpose of this master
thesis is to verify the first statement. Thanks to cheap renewable energy and technology
development in 2050, it is assumed that the second statement will also be fulfilled, since
the main operating cost of hydrogen is electricity cost [26].

3
1. Introduction

1.3 Scope
The thesis begins by examining the current state of electrolyzer technology. It provides a
technology classification based on the operating temperature and electrolyte, comparing
materials used, performance, and technological maturity. Notably, the focus of the thesis
is on AWE and PEM electrolyzers, which are the most established and developed in terms
of technology maturity and commercial development. Therefore, these two types are the
main subjects of further benchmarking and analysis.

From the outset, it is important to clearly define the boundaries and the complexity of
the model. An electrolyzer stack needs a broad infrastructure to operate: electrical power
supply, AC/DC converter, water and gas treatment systems, compressor, and hydrogen
storage. Nevertheless, the thesis scope covers only the first two components. Besides, an
accurate model of the electrolyzer demands an electrical and thermal model. However,
due to the project’s focus on the electrical aspects, the thermal dynamics are omitted in
this project, and only operation at constant temperature is considered.

A power electronics study identified the most relevant converter topologies for electrolyzer
applications. After selecting the 12-pulse thyristor rectifier and studying its working
principle, the converter’s Electromagnetic Transient (EMT) analysis was left out of the
scope. Afterward, the controller implementation is carried out through a system analysis
of the converter-electrolyzer interaction using only RMS simulation and average value
calculations on the DC side.

Station design for the GW-scale electrolyzer plant included evaluating different station
proposals available in two main projects. These single-line diagrams (SLD) are all con-
nected to the High Voltage (HV) transmission grid through a 400 kV Point of Common
Coupling (PCC). These diagrams differ in the bus voltage levels inside the station, the
transformers’ power rating, and the parallel connection of the HV transformers. Finally,
only one candidate is proposed for further analysis and simulation.

A subsection states the grid code requirements that an industry must comply with so the
Transmission System Operator (Svenska Kraftnät) allows its connection to the HV grid.
Several documentation sources, like the Demand Connection Code (DCC), IEEE, and IEC
standards for harmonic distortion, were considered. The covered grid code compliance
categories are harmonics and distortions, power factor correction, and voltage regulation.

Station design included selecting and rating HV components such as transformers, filters,
converters, and reactive power compensation devices. However, this project excluded
switchgear selection, so no short-circuit analysis or protection coordination studies were
carried out. This also led to the implementation of a single busbar for Point of Common
Coupling (PCC) instead of more complex substation schemes like sectionalized single
busbars or double busbars.

Afterward, the station operation is tested in multiple scenarios. For example, its design is
validated in a steady state through a load flow analysis, its dynamic operation is achieved
through a ramp-up simulation, and its voltage regulation capabilities are demonstrated
through load rejection events. Voltage variations at the PCC were excluded to reduce the
number of simulations. Lastly, the economic evaluation of this project is left outside of
the scope, mainly focused on the technical feasibility.

4
 1. Introduction

1.4 Societal, ethical and ecological aspects

The environmental benefit of substituting “grey and black” hydrogen with clean green
hydrogen is clear: it creatively addresses hard-to-electrify sectors in the climate crisis.
Nevertheless, it is necessary to consider its societal consequences and ethical dilemmas.

First, a direct consequence of the clean energy transition is an employment shift. Indus-
tries in power generation, renewables, power grid, energy efficiency, e-fuels, and biogas
expect unparalleled growth. Contrarily, Big Oil & Gas corporations face a loss of profits
that echos into a loss of employment unless a purpose shift is implemented. However,
some of these emerging sectors including hydrogen production offer a good opportunity
for job relocation of those dedicated to the fossil-fuel industry today [27].

An example at national perspective of how this could be achieved is Norway’s vision

for decarbonization. In 2022, Norwegian petroleum export revenues were roughly 50%,
supplying 25% of Europe’s oil and natural gas demand. Norway intends to cut emissions
to 80-95% by 2050 [28], with hydrogen playing a key role [29].

Secondly, clean water consumption is another factor to consider, since it is a globally

valued resource and is very scarce in some countries. Increase of clean water demand due
to massive hydrogen electrolyzers adoption might cause inequality and unfair distribution
within semiarid-desertic countries where fresh-water availability is an issue.

According to UN [30], as of today, more than 2 billion people live in countries exper-
iencing high water stress that might exacerbate with population growth and climate
change effects. To ensure reliable water supply in the coming years, these countries
expect to expand their water management infrastructure, especially sea water purific-
ation and wastewater management [31]. This might ensure water availability, but de-
preciate hydrogen supply. Nonetheless, the solution could be the use of solar radiation
in semiarid-desertic countries where PV and solar-thermal energy are cheap sources of
electricity. Chile, for example, could be a good proof of concept since it has the driest
nonpolar desert in the world (Atacama Desert) and water scarcity issues. Despite this,
its government presented in 2020 its ambitious National Green Hydrogen Strategy [32]
which aims to produce world’s cheapest green hydrogen by 2030.

Thirdly, some fuel cells and electrolyzers technologies require rare-earth metals. Iridium
and platinum electrodes are widely used in commercial electrolyzers which poses an un-
sustainable practice in the long term since they are considered as “endangered” species
among metals [33]. Naturally, new catalysts might appear in the market as technology
develops. For example, a 2022 study [34] by RIKEN CSRS in Japan found a rare metal-
free catalyst based on Co2 MnO4 , lasting around 1500 hours with a current density of 0.2
A/cm2 . This performance is superior to other non-rare metal catalysts; although it still
stays behind iridium catalysts with longer lifetimes.

Lastly, safety guidelines and protocols are crucial to consider when using hydrogen due to
its flammability and explosiveness, though it is non-toxic compound that rapidly dissipates
in case of leak [35]. The Hydrogen tools report [36] discusses potential incident scenarios
and appropriate prevention and mitigation measures.

5
1. Introduction

1.5 Method
The benchmarking between AWE and PEM begins with an individual technology de-
scription, emphasizing their fundamental differences. All key performance indicators are
summarized and tabulated, and a decision is made to implement one of the technologies
for the GW-scale station.

Creating an accurate model was a meticulous process that required a deep understanding
of the electrical and thermal basics of electrolysis. This thoroughness translated into
a comprehensive scientific literature review covering thermodynamics, electrochemistry,
and energy systems modeling. These topics are important because they impact the stack
efficiency and current-voltage characteristics.

The technical specifications for a commercial MW-scale electrolyzer, such as current-

voltage characteristics, were unavailable. Therefore, the MW-scale electrolyzer model was
derived from the performance of the PHOEBUS plant kW-scale electrolyzer. Adjustments
were made to design parameters such as electrode effective area and number of cells in
series to match the hydrogen production rate and power input requirements of the Nel
Hydrogen A485-stack while ensuring compliance with the maximum rating limitations.
Subsequent power upgrades like larger modular units (A3880) provided by Nel Hydrogen,
and load agglomeration enabled the simulation of larger equivalent loads.

The current controller design follows the Internal Model Control (IMC) tuning method,
and the system analysis includes a temperature test and ramp-up sequence test with dif-
ferent tuned control parameters. The controller is designed per unit, ensuring its universal
implementation independently of the electrolyzer ratings, thereby providing reassurance
about its adaptability.

All simulations are conducted using the power system software PowerFactory (DIgSI-
LENT). It is considered a solid power system analysis software similar to PSS/E and
PSCAD. Proficient in RMS simulations with an extensive and comprehensive library of
components while combining various power system analysis tools within one platform. In
contrast, PSS/E presents a steeper learning curve and less flexibility than PowerFactory,
making it a less attractive option. PSCAD, while excellent for detailed electromagnetic
transient (EMT) simulations, is not a good choice for RMS simulations and larger-scale
stability studies such as PowerFactory.

The electrical component rating process involved two stages: an initial rating determined
through simplified calculations, followed by validation through load flow calculations and
dynamic simulations such as ramp-up sequences and load rejection studies.

For grid code compliance, every category is validated through specific studies and the
implementation of necessary equipment. Harmonic distortion is evaluated through a har-
monic analysis and fulfilled through implementing passive filters. Power Factor Correction
is evaluated through a load flow analysis and ramp-up sequence, where a combination of
capacitor banks and MVAr compensators maintain a null exchange of reactive power with
the grid. Finally, voltage regulation requirements are accomplished by implementing a
STATCOM at PCC, and its performance is reviewed through load rejection studies.

6
 2
State of the art

Water electrolysis is a chemical reaction occurring within an insulating container filled

with water. This aqueous solution must increase its electrical conductivity in order to
allow a free transfer of ions, so a soluble salt or ionic compound, such as potassium,
sodium, or bicarbonate, is added. This combination with water forms an electrolyte,
creating an alkaline medium that serves as a conductive environment for ions. Then two
electrodes are introduced into the container and submerged in this aqueous phase. Upon
applying a voltage difference across both electrodes, a redox reaction occurs, leading to
the splitting of water into its fundamental elements: hydrogen and oxygen.

H2 O (l) −−→ H2 (g) + 21 O2 (g)

Electrolysis concept is a straightforward procedure, but the chemical (electrodes, elec-

trolyte) and operating (temperature and pressure) conditions in which it happens vary
considerably depending on the technology. It can be classified according to their electro-
lyte, charge carrier, operating temperature etc. The number of experimental technologies
is extensive as well, however, the attention will be focused only on the most mature. In
Figure 2.1, it is shown a popular classification of electrolyzer technologies according to
their electrolyte type and operating temperature.

High-temperature electrolyzers (HTE) present big advantages against low-temperature

technologies in terms of efficiency. Since electrolysis occurs at high temperatures, water
is present in steam phase, which decreases the reaction net exchange enthalpy and Gibbs-
free energy. This results in a lower energy input into the reaction which results in higher
efficiency. Nevertheless, this high temperature operating range is impractical for universal
application, which results only interesting for intensive-energy industries like iron&steel
or traditional thermal plants like geothermal, nuclear or CHP. Besides its technology is
still in R&D phase compared to low-temperature electrolyzers.

A comparison of the efficiencies, materials used, technology maturity and expected number
of hours of operation is described in Table 2.1. In terms of technology maturity and
commercial development, only AWE and PEM electrolyzers are fully established and
developed, so only these will be subject of further discussion and analysis.

7
2. State of the art

Figure 2.1: Electrolyzer classification according to electrolyte and operating temperature.

Table 2.1: Water Electrolysis Technology Description [37],[38],[39].

Parameter AWE PEM AEM SOE

Charge carrier OH- H+ OH- O2-
Electrolyte 20–40 wt% Solid polymer DVB poly- Yttria stabil-
KOH /NaOH electrolyte mer with 1M ized Zirconia
(PFSA) KOH/NaOH (YSZ)
Anode Nickel coated IrO2 , RuO2 NiFeCo alloys Perovskites
stainless steel (LSCF, LSM)
Cathode Nickel coated Platinum car- Nickel Ni/YSZ
stainless steel bon
Separator Asbestos/Zirfon Nafion Fumatech Solid electro-
lyte YSZ
Gas Diffusion layer Nickel mesh Titanium mesh Nickel foam Nickel mesh
/carbon cloth /carbon cloth /foam
Bipolar Plates Ni-coated Pt/Au-coated Ni-coated Co-coated
stainless steel titanium stainless steel stainless steel
Current density 0.2–0.8 A cm2 1–2 A cm2 0.2–2 A cm2 0.3–1 A cm2
Cell Area <3 m2 <0.15 m2 <0.03 m2 <0.02 m2
Voltage efficiency 50-78% (LHV) 50-83% (LHV) 50–59% (LHV) 89% (LHV)
Specific stack en- 4.2–4.8 4.4–5.0 - 3.4 kWh/Nm3
ergy consumption kWh/Nm3 kWh/Nm3
Temperature limits 70–90 o C 50–80 o C 40–60 o C 700–850 o C
Gas pressure 1–10 bar 15–30 bar <30 bar 1 bar
Lifetime stack (60-90) 000 h (50-80) 000 h 30 000 h 20 000 h
Lifetime system 20–30 years 10–20 years - -
Development Mature Commercialized R&D R&D
status

8
 2. State of the art

2.1 Alkaline Water Electrolyzer (AWE)

Alkaline electrolyzers are considered the most mature option in the market today. Its
large development began in the 1920s and 1930s with the expansion of the chemical in-
dustry [40], providing several decades of refinement and optimization, contributing to their
established reputation as a reliable and well-understood choice for hydrogen production.

They are characterized by using an alkaline electrolyte, typically, potassium hydroxide

with a concentration between 25-30% where it reaches its maximum conductivity at 80o C
[41]. This salt dissolved in water increases its pH forming a basic aqueous phase with
free hydroxide ions. To separate both electrodes typically a 252-463 µm Zirfon diaphragm
soaked with KOH solution is employed [38],[42]. Then, these are pressed onto the sep-
arator following a zero-gap cell design [43], reducing the surface-specific cell resistance.
These electrodes are typically stainless steel plates with nickel alloy coating. Nickel (Ni)
is frequently used due to its catalytic activity, corrosion resistance, conductivity, but most
importantly because of its availability, low cost and proven stability in high pH mediums.

Figure 2.2: AWE cell working principle [37].

In Figure 2.3, it is shown the typical system infrastructure required for the AWE operation
[44]. The electrolyte circuit (grey) consists of two gas separators, where the excess fluid is
recovered and reintroduced into the circuit. In the H2 piping network (purple), a scrubber,
deoxidizer and dryer are used to remove traces of electrolyte and water, while improving
the gas purity. Besides, since most AWE work at low pressure, a compressor is usually
added before the process or a storage tank ahead to increase the delivery pressure.

Figure 2.3: Typical AWE system architecture [44].

9
2. State of the art

2.2 Proton Exchange Membrane Electrolyzer (PEM)

PEM technology emerged as an alternative to AWE for hydrogen production in mid-
20th century. A breakthrough in ion-conducting membranes occurred in the late 1960s
with Dr. Walther G. Grot’s development of Nafion, a perfluorosulfonic acid (PFSA)-based
copolymer membrane. This innovation paved the way for PEM fuel cells and electrolyzers,
whose main driver was the possibility to generate oxygen with high differential pressure,
which was an indispensable tool for aerospace and submarine life support applications[40].

Overall, PEM’s key advantage lies in utilizing a solid polymer electrolyte (SPE). The
most used SPE in the market today is still the Nafion ionomer family [45] with PFSA-
membrane. Despite the corrosive environment, this technology employs durable cata-
lysts, typically platinum (Pt) nanoparticles in the cathode and iridium oxide (IrO2 ) nano-
particles in the anode. These particles are then integrated into carbon paper and sintered
titanium plates forming the electrodes.

Figure 2.4: PEM cell working principle [37].

The resulting membrane electrode assembly (MEA) enables a significant advantage, since
it has enough mechanical strength to allow differential pressure across the electrodes. In
Figure 2.5, Nel Hydrogen provides schematics for PEM electrolyzer system infrastructure
[44]. Here, hydrogen gas is generated at elevated pressures up to 30 barg which corresponds
to the process delivery pressure. Oxygen gas is produced at pressures close to ambient.

Figure 2.5: Typical PEM electrolyser system architecture [44].

10
 2. State of the art

2.3 Economic and Performance Evaluation

The most mature technologies available in the market have already been presented and its
characteristics stated. Therefore, it is necessary to compare both technologies, understand
their differences and also to ponder their advantages and disadvantages against their
compatibility with the electric power source typology, the transmission connection either
AC or DC and the scale of hydrogen production.

2.3.1 Investment costs

The first driver for building a station with such big dimensions as GW-scale is investment
cost, long term stability and the possibility to implement large scale systems which should
reduce the station’s CAPEX.

Currently, the investment cost for AWE stands around 500–1000 USD/kW [37], [38] and
the system’s lifespan reaches between 60 000 - 90 000 hours [46]. In comparison, PEM
boasts a stack lifetime of 50 000 - 80 000 hours with minimal performance decline [38].
The hydrogen production cost for PEM is between 700 and 1400 USD/kW [37], [38].

A key factor contributing to the higher cost of PEM compared to AWE is the elevated
expense of cell components. PEM electrolyzers employ noble metals like IrO2 in the anode
and Pt in the cathode, which are more expensive than their counterpart Ni predominantly
used in AWE technology. Not to mention the bipolar plates made of Pt/gold-coated
titanium, which may account for 25% of total system capital cost [38]. To illustrate, a
1.9 MW PEM water electrolyzer operating at 2 A cm-2 , assuming an Ir loading of 5 mg
cm-2 and Pt loading of 2 mg cm-2 [38], requires approximately 2.47 kg of Ir and 0.95 kg
of Pt. The anticipated cost of the catalysts goes around 435 000 USD (147.5 USD/gram
and 30 USD/gram respectively as of August 2023 [47]), which means 206.75 USD/kW
and a 15% portion of CAPEX. In contrast, the cost of Ni fluctuates between 0.02 and
0.04 USD/gram as of 2020 to 2023 [47].

Finally there is the scale factor. According to Nel Hydrogen catalogue [44], it is offered
stacks MC500 (PEM) and A485 (AWE), both units capable of generating 490 Nm3 /h.
Afterwards, stacks can be mounted in series or parallel groups to reach higher production
rates. In PEM technology two of the biggest systems available in the market are Silyzer
300 (Siemens) with 2 000 Nm3 /h and MC5000 (Nel Hydrogen) with 4 920 Nm3 /h [37]. On
the AWE side, Nel Hydrogen offers an electrolyzer system A4000 with 3 880 Nm3 /h. After
this comparison, it is logical to conclude that both technologies have similar economies of
scale.

2.3.2 Operating conditions

If the operating conditions for both AWE and PEM is compared in Table 2.1, it is clear
that both technologies do not operate at the same range. The great advantage of PEM
technology is that it can operate at higher pressures than AWE, in the range of 30-70 barg
as well as it enables differential pressure between anode and cathode. For example, in Nel
Hydrogen catalogue [44], their example PEM Hydrogen plant works at 30 barg in the H2
side and at almost atmospheric pressure in the O2 side. This, however, may reduce the
efficiency since it will require a thicker membrane to improve the mechanical stability [38].

11
2. State of the art

In terms of temperature, both technologies work in similar ranges despite the difference of
pressure. The low operating pressure for AWE usually requires a compressor downstream
that increases the delivery pressure till even 200 barg [44].

Another detail to highlight is that PEM water re-circulation circuit requires only distilled
and de-ionized water to operate, while AWE needs a separate circuit to re-mix the lye
(potassium hydroxide) into the feed water supply, observe Figure 2.3. Nevertheless, both
electrolyzers are facing corrosive environments (alkaline and acidic respectively) and are
sensitive to water impurities.

2.3.3 Performance
It is also important to compare the performance of both technologies. The range of
efficiencies (LHV) vary, but it usually comprises between 50-78% for AWE and 50-83%
for PEM. Nonetheless, this efficiency comparison found in the literature [37], [38] is not
a definitive or case-closed comparison, since it comprises a large spectrum of sources. It
is also important to turn to commercial manufacturers for specification request like Nel
Hydrogen who reports efficiencies at stack around 3.8-4.4 kWh Nm-3 (AWE) and 4.5 kWh
Nm-3 (PEM) [44], equivalent to 67-77% (LHV).

On the other hand, AWE and PEM work in different ranges of current density as it
was shown in Table 2.1 and these are well settled. If the curves of AWE and PEM are
plotted operating in the same range, it is obtained the results that Schalenbach et al. [42]
performed in 2016 as displayed in Figure 2.6.

Figure 2.6: Overvoltage for AWE and PEM in the same current density range [42].

12
 2. State of the art

The red and blue dotted lines correspond to acidic cells with Nafion membranes of different
thicknesses (210 µm and 57 µm) and the dotted black line for an alkaline cell with a
diaphragm thickness of 463 µm. The objective of this study was to compare the voltage
curves for two equivalent cells with the same thickness (57 µm) for both, membrane and
diaphragm, according to their model [42].

AWE electrolyzers working principle consists on using hydroxil ions OH- as charge carrier
with a KOH solution permeating through the porous structure of the diaphragm. This
method carries the risk of gas intermixing (H2 and O2 ) between both chambers known
as crossover. To reduce this risk thicker diaphragms and spacers are used, however, this
increases the total ohmic resistance of the cell harming the efficiency [38]. Since it is not
viable to reduce the diaphragm thickness to comparable dimensions to Nafion membranes,
then the alternative is to work at lower current densities with the same voltage range and
efficiency than PEM. In order to increase the power to comparable levels then the electrode
area is increased. This logic leads to a relatively bulky design for alkaline stacks and quite
compact for PEM, but since the cost of components and materials for AWE is fairly more
economical than PEM, this is an adequate approach.

This relatively high crossover in AWE through the diaphragm limits its operating pressure,
as well as the main reason why AWE may not perform differential pressure between
cathode and anode. This problem also relates to the lower H2 purity in AWE (99.5-
99.99%) compared to PEM’s (99.9-99.99%) [37].

Lastly, it is important to compare the dynamic performance of both technologies.

Table 2.2: Performance comparison [38], [37].

Parameter description AWE PEM

Load range of nominal load 15-100% 5-120%
Start up time 1-10 min 1 s-5 min
Shutdown time 1-10 min Seconds
Ramp-up speed 0.2-20%/s 100%/s
Ramp-down speed 0.2-20%/s 100%/s

This table shows a superior performance of PEM respect to AWE, in terms of dynamic
response to changes in H2 demand. If the nature of renewable sources is analyzed, it
is characterized for intermittent production and operating at minimum and maximum
capacity during short periods of time. PEM electrolyzers seem to be better suited to
meet these requirements given their energy efficient stand-by operation, their short cold-
start times and wide loading range (5-120%) [46].

The Puertollano green hydrogen plant is a good proof of concept for this logic. This
Iberdrola project commissioned in 2022 enabled the connection of 100 MW photovoltaic
solar plant with a 20 MWh, 5 MW lithium-ion battery system and 20 MW capacity PEM
electrolytic hydrogen production system [48].

13
2. State of the art

2.4 Summary
In conclusion, since the aim of this thesis is to propose and model a hypothetical GigaWatt
electrolytic station connected to the High Voltage AC transmission grid, the economies
of scale for a plant of this size require the solution with lowest investment cost, more
scalability and longer lifetime.

Applying the maximum CAPEX costs stated in Table 2.3 and considering 1 GW of
electrolyzer capacity, implementing AWE can lead to a reduction of 380 million USD
of investment cost over PEM. Another argument is the non-intermittent condition of
supply which does not require any high performance response.

These factors push the balance for selection of AWE as the candidate for model develop-
ment in the substation. Nonetheless, as it is seen in subsequent sections, this model is
not limited by its technology and it is exchangeable for PEM electrolyzers.

All the information stated in this chapter is summarized in the table below, highlighting
the key differences between both technologies.

Table 2.3: Performance comparison [38], [37].

Parameter description AWE PEM

Temperature 70-90°C 50-80°C
Pressure <30 bar <70 bar
Nominal current density 0.2-0.8 A/cm2 1-2 A/cm2
Efficiency (LHV) 50-78% 50-83%
Gas purity >99.5% 99.99%
CAPEX 500-1000 USD/kW 700-1400 USD/kW
Operating stack lifetime 60 000 - 90 000 h 50 000-80 000 h
Load range of nominal load 15-100% 5-120%
Start up time 1-10 min 1 s-5 min
Shutdown time 1-10 min Seconds
Ramp-up speed 0.2-20%/s 100%/s
Ramp-down speed 0.2-20%/s 100%/s

Although the alkaline electrolyzer technology is an old concept as it was first introduced
by Troostwijk and Diemann in the year 1789 [40], there is still research and great break-
throughs coming out for this technology.

The latest achievement is the alkaline capillary-fed electrolysis technology [49], originally
conceived in 2021 at University of Wollongong, Australia [50]. Professor Gerry Swiegers
and Gordon Wallace, lead researchers of the team, later founded Hysata, a spin-out com-
pany of the university, to commercialize this new technology.

Its unique concept of water electrolysis comes from the introduction of capillary-induced
transport along a porous inter-electrode separator, creating a bubble-free operation at the
electrodes. This leads to a breakthrough in terms of efficiency, with a cell voltage of 1.51
V at 0.5 A cm-2 , 85o C and a stack energy consumption of 40.4 kWh/kg compared to 47.5
kWh/kg in commercial electrolysis cells [49].

14
 3
Thermodynamic and Electrochemical
Analyses

An electrolyzer is an electrochemical device and its performance depends on many factors

beyond the electrical, such as thermodynamics, chemical kinetics and equilibrium, acid-
base theory and fluid mechanics. Even so, this section will focus only on analyzing
the electrical and thermodynamic principles behind its behavior. Most of the theory
developed in this chapter is based on the book Comprehensive Energy Systems, Chapter
4.25 Electrolyzers by Ibrahim Dincer and Abdullah A. AlZahrani [51]. This analysis is
valid for both AWE and PEM electrolyzers equally.

3.1 Thermodynamics applied to chemical reactions

To analyze the thermodynamic behavior of the electrolyzer, boundaries are defined by
placing a control volume (CV) over it to track the amount of substance entering and
exiting, as depicted in Figure 3.1. There is one inlet for water and two separate outlets
for the gas products.

Figure 3.1: Control volume of the electrolyzer [51].

Assuming the conservation of mass and a steady-flow process, it is possible to apply the
First and Second Law of Thermodynamics expressed respectively as

15
3. Thermodynamic and Electrochemical Analyses

X X
Q̇in + Ẇin + ṁ ∗ h = Q̇out + Ẇout + ṁ ∗ h (3.1)
in out

X X Q̇in − Q̇out
Ṡgen = ṁ ∗ s − ṁ ∗ s − (3.2)
out in
T

where Q is the heat flow, W the work, Sgen the entropy generation inside CV, T the
CV temperature, ṁ the mass flow, h the specific enthalpy and s the specific entropy.
Subscript in and out represent the flow of either heat, work or substances going inwards
and outwards the CV respectively. These expressions may be further simplified defining
the net flow with subscript CV resulting in

X X
Q̇CV + ẆCV = ṁ ∗ h − ṁ ∗ h (3.3)
out in

X X Q̇CV
Ṡgen = ṁ ∗ s − ṁ ∗ s − (3.4)
out in
T

In order to have a better understanding of what it means to have net positive and negative
flow of heat and work in the CV, the sign criteria is mentioned in Table 3.1.

Table 3.1: Sign criteria for heat transfer and work.

Q̇CV = Q̇in − Q̇out Q̇CV > 0 : Endothermic process Q̇CV < 0 : Exothermic process
ẆCV = Ẇin − Ẇout ẆCV > 0 : Consumes work ẆCV < 0 : Produces work

Expressions (3.3) and (3.4) need to be adapted for chemical reactions environments. In
addition to sensible and latent energy, which are associated with temperature and phase
changes, the substances’ molecules also possess chemical energy that varies by modifying
their molecular composition. For the electrolysis reaction, First and Second Laws of
Thermodynamics can be reformulated by substituting the mass flows ṁ for molar flows
ṅ and expressing the specific enthalpy terms based on their enthalpy of formation hof .

Q̇CV + ẆCV = ṅH2 (hof + h − ho )H2 + ṅO2 (hof + h − ho )O2 − ṅH2 O (hof + h − ho )H2 O (3.5)

Q̇CV
Ṡgen = (ṅsH2 ) + (ṅsO2 ) − (ṅsH2 O ) − (3.6)
T

Standard enthalpy of formation hof may be understood as the enthalpy of a substance at

a specified state due to its chemical composition. All elements in their natural states like
hydrogen gas, oxygen gas and solid carbon in the form of graphite, have a hof of zero at

16
 3. Thermodynamic and Electrochemical Analyses

standard conditions of pressure and temperature (25o C and 1 atm). Furthermore, expres-
sion (3.5) can be expressed per mol of hydrogen produced, which gives the stoichiometric
coefficients and all properties expressed on specific terms (per mol terms).

1
qCV + wCV = (hof + h − ho )H2 + (hof + h − ho )O2 − (hof + h − ho )H2 O = ∆h (3.7)
2
1
qCV = T (sH2 + sO2 − sH2 O ) − T sgen = T (∆s − sgen ) (3.8)
2

All variables defined in the previous and in the following equations are listed in Table 3.2:

Table 3.2: Thermodynamic variables.

Parameter description Parameter Unit

Net specific enthalpy ∆h kJ mol-1
Net specific Gibbs energy ∆g kJ mol-1
Net specific entropy ∆s J (mol K)-1
Standard enthalpy of formation hof kJ mol-1
Specific enthalpy (T , P ) h kJ mol-1
Standard specific enthalpy ho kJ mol-1
Specific entropy (T , P ) s J (mol K)-1
Specific generated entropy sgen J (mol K)-1
Specific heat transfer qCV kJ mol-1
Specific work wCV kJ mol-1

Applying the different species properties of Table 3.3, it is possible to calculate the net
enthalpy ∆h and entropy ∆s of electrolysis at standard conditions using equation (3.7),
(3.8):

1
∆ho = hof,H2 + hof,O2 − hof,H2 O (3.9)
2
1
∆so = sH2 + sO2 − sH2 O (3.10)
2

which results in a ∆ho of 285.826 kJ mol-1 and ∆so = 0.163 kJ (mol K)-1 . Since electrolysis
∆ho is positive, the reaction is an endothermic reaction in standard conditions, which
means that it needs a heat flow or work of 285.826 kJ into the system per mol of hydrogen
produced in order to keep the system’s temperature constant.

Table 3.3: Thermodynamic properties for liquid water, hydrogen and oxygen under standard
conditions [51].

Substance Enthalpy of formation Gibbs free energy Entropy

hf [kJ mol-1 ] gf [kJ mol-1 ] s [J (mol K)-1 ]
Water(l) - 285.826 - 237.180 69.92
Hydrogen 0 0 130.68
Oxygen 0 0 205.14

17
3. Thermodynamic and Electrochemical Analyses

Two different scenarios are compared: one where the process occurs in an adiabatic system
(no heat flow is allowed) and another where the process is reversible (no generated entropy
or irreversibilities). In both scenarios, equations (3.7) and (3.6) are applied.

Table 3.4: Impact of an adiabatic system and reversible process on electrolysis.

Adiabatic system qCV = 0 Reversible process sgen = 0

wCV = ∆h = 285.826 kJ mol-1 qCV = T ∆s = 48.697 kJ mol-1
sgen = ∆s = 0.163 kJ (mol K)-1 ẇCV = ∆g = 237.13 kJ mol-1

From these results, it is possible to get three conclusions:

• The necessary work wCV for electrolysis is greater in an adiabatic process than in a
reversible process.

• In a reversible process, work wCV equals to the free Gibbs energy of reaction ∆g.

• Electrolysis free Gibbs energy ∆g is positive, which means that at standard condi-
tions, it is a non-spontaneous reaction.

3.2 Reversible and Thermoneutral Potentials

After evaluating the thermodynamics of the electrolyzer, the next step is analyzing its
electrical correlation. By merging equation (3.7), (3.8) and adding the definition of Gibbs
free energy ∆g, it is possible to obtain the relationship between this last variable and the
work applied to the electrolytic cell wCV .

wCV = ∆h − T ∆s + T sgen = ∆g + T sgen (3.11)

In principle, electrochemical cells convert chemical energy to electrical energy and vice
versa, so it is possible to correlate the necessary work of the reaction wCV with the
potential given by the cell Ecell . This results in equation (3.12), where wCV is equal to
the product of Ecell and the total charge transferred during the reaction zF , being z the
number of electron moles pero mol of hydrogen produced and F the Faraday constant.

wCV = zF Ecell (3.12)

By amalgamating equation (3.11) and (3.12), it is possible to define an equation that es-
tablishes a relationship between the net Gibbs free energy ∆g, and the reversible voltage
Er . The overvoltage term Eloss accounts for the reaction’s irreversibilities and Er repres-
ents the minimum voltage required to start the electrolysis reaction.

∆g T sgen
Ecell = + = Er + Eloss (3.13)
zF zF
Yet, this operating point does not provide any information about the thermal behavior
of the electrolyzer. Either if electrolysis is an endothermic or exothermic process depends

18
 3. Thermodynamic and Electrochemical Analyses

on the thermoneutral voltage Eth . Furthermore, it is possible to modify the cell potential
definition with Eth as the main input as shown in equation (3.14).

∆h T ∆s T sgen T ∆s
Ecell = − + = Eth − + Eloss (3.14)
zF zF zF zF

When the electrolytic cell is operated above Eth , the reaction becomes exothermic, and
heat must be removed from the cell for maintaining constant temperature operation which
is known as "isothermal operation". On the other hand, operating under Eth demands
absorption of heat to keep isothermal operation [52]. Applying standard conditions of
pressure and temperature, the following potentials are calculated using the values provided
in Table 3.4, with Ero equal to 1.23 V and Etho
to 1.48 V. This principle is displayed in
Figure 3.2.

Figure 3.2: Reversible and thermoneutral potential versus temperature at atmospheric pressure.

Although it is important to consider the thermoneutral voltage Eth when defining the
thermal model, working under endothermic operation is not a concern since the heat
demand is mitigated by the irreversibilities of the reaction Eloss . To operate above the
thermoneutral voltage, manufacturers usually apply a minimum loading limit which for
AWE is around 15% [44]. This ensures that the electrolyzer always functions as an
exothermic and refrigerated device.

Both potentials Er and Eth are not constant. They depend on the electrolyzer operating
pressure and temperature. Therefore, it is necessary to study this effect, for which task
the equations stated below will be used. Those are derived in the reference book [51].

∂Er 1 ∂∆g ∆s
= ∗ =− (3.15)
∂T zF ∂T zF
∂Er 1 ∂∆g ∆v ∆N RT 1
= ∗ = = (3.16)
∂P zF ∂P zF zF P

19
3. Thermodynamic and Electrochemical Analyses

Some conclusions may be obtained from both equations. First, equation (3.15) results into
a constant, which is -0.0008464 V/K at standard conditions using the data from Table
3.3, and gives Er a linear dependence respect to temperature. On the other side, it shows
a logarithmic behavior with pressure according to equation (3.16). If both expressions are
integrated between the standard point of operation and the objective temperature and
pressure, equation (3.15) and (3.16) are derived:

Z To
∂Er ∆s
Er (T, Po ) = dT = Ero − ∗ (T − To ) (3.17)
T ∂T zF
Z Po
∂Er ∆N RT P
Er (T, P ) = dP = Er (T, Po ) + ln (3.18)
P ∂P zF Po

(a) At atmospheric pressure (1 bar). (b) At ambient temperature (25o C).

Figure 3.3: Effects of operating pressure and temperature on the reversible voltage.

A conclusion gained from Figure 3.3b is why in PEM electrolyzers the minimum loading
limit is closer to 5% unlike 15% in AWE. Since PEM electrolyzers usually work at pressures
around 30 bar or more, it increases its reversible voltage, which gets Er closer to Eth than
its counterpart AWE working at atmospheric pressure.

Besides the operating temperature and pressure, there is one last factor that may modify
the reversible cell potential: the reactant concentration. This effect is calculated through
the Nernst equation, which allows to calculate Er in case of a differential pressure between
the electrolyzer’s cathode and anode as may happen in PEM electrolyzers.

1
RT pH2 p022
Er = Ero + ln (3.19)
zF pH 2 O

The results obtained from these equations agree with the ones published by Hammoudi et
al. [53] and Adibi et al. [54]. In both studies the reversible voltage increases accordingly
with the pressure and decreases with temperature as shown in Figure 3.4.

20
 3. Thermodynamic and Electrochemical Analyses

Figure 3.4: Pressure and temperature effect on reversible and thermoneutral potentials [53].

A simple analysis of thermoneutral potential Eth dependence with temperature at at-

mospheric pressure can be done by reformulating the enthalpies in equation (3.7) into
equation (3.20). The results are obtained by applying the enthalpy of formation of elec-
trolysis, the enthalpy of components at standard conditions and at operating conditions
(NIST thermophysical experimental data [55]).

1
∆h = ∆hf + (h − ho )H2 + (h − ho )O2 − (h − ho )H2 O (3.20)
2

Table 3.5: Enthalpy for water, hydrogen and oxygen at atmospheric pressure versus temperature.

Temperature [K] 290 300 310 320 330 340 350 360
hH2 (g) [J mol-1 ] 7691.2 7979.4 8268.3 8557.7 8847.7 9138.1 9428.8 9719.8
hO2 (g) [J mol-1 ] 8432.4 8726.5 9021.1 9316.2 9611.9 9908.2 10205 10503
hH2 O (l) [J mol-1 ] 1275.9 2029.5 2782.5 3535.5 4288.8 5042.9 5798 6554.3
∆h [kJ mol-1 ] 286.09 285.77 285.45 285.13 284.82 284.5 284.19 283.87
Eth [V] 1.483 1.481 1.479 1.478 1.476 1.474 1.473 1.471

These values obtained from tabulated experimental data can be compared with the results
acquired from the empirical equation deduced by Hammoudi et al. in [53]. When both
methods are compared, the difference in each temperature step is lower than 1 mV.

To study the thermoneutral potential dependence with pressure requires a more advanced
model as presented in articles [53] and [54]. Nevertheless, the results of that analysis can
be observed in Figure 3.4.

21
3. Thermodynamic and Electrochemical Analyses

3.3 Polarizations and efficiency

The main purpose of this chapter is to develop the overvoltage term Eloss introduced
in equation (3.13). This accounts for the reaction’s irreversibilities and the efficiency
decrease of the cell. It can be classified into three main types of polarizations:

3.3.1 Activation losses

The activation polarization is attributed to the charge transfer that happens at the elec-
trode surface. An analogy for this overvoltage is a resistance in the electrochemical re-
action occurring at the electrode–electrolyte interface in both the anode and cathode.
The real reason of this overvoltage is due to the fact that there are elementary reactions
happening at the molecular level, in contrast to the simplistic cathodic and anodic reac-
tions shown in the schematics of section 2. Its value is determined by the Butler-Volmer
equation (3.21) and can be separated into cathode and anode overpotentials known as
Tafel equations.

zF zF
Jcell = Jo [exp ( αan ηact,an ) − exp (− αca ηact,ca )] (3.21)
RT RT

RT Jcell RT Jcell
ηact,an = ln ( ) ηact,ca = ln ( )
αan zF Jo,an αca zF Jo,ca

3.3.2 Ohmic losses

The ohmic polarization is more intuitive than the rest since it presents a linear behavior
respect to the current density and fulfills Ohm’s Law. The ions moving through the elec-
trolyte and membrane/diaphragm find ionic resistance in its path as well as the electrons
through the electrodes. The sum of all these components, since they are all in series, gives
the total electric resistance.

X δi X δi
ηohm = Icell Rtot = Icell = Jcell (3.22)
i
σi A i
σi

In the case of AWE, the ohmic resistance Rtot mainly depends on the electrolyte conduct-
ivity σ which depends on the molarity and temperature [41]. In PEM electrolyzers it is
similar but on the thickness δ and conductivity of the membrane σ instead [56].

3.3.3 Concentration losses

The concentration polarization is an overvoltage attributed to the lack of reactants at the
electrode-electrolyte interface. In electrolysis, the overvoltage is typically negligible due
to the abundance of reactants in the electrolyzer: the electrodes are fully immersed in
water, ensuring a wide supply of reactants. Nevertheless, this effect becomes noticeable
at high loadings due to the accumulation of gas bubbles in the membrane, resulting in an
exponential voltage increase.

22
 3. Thermodynamic and Electrochemical Analyses

The sum of the reversible cell potential and the overvoltages stated above gives the cell
potential. Dividing the reversible potential by the cell potential gives the voltage efficiency
ηV , also known as low heating value (LHV) efficiency ηLHV of the electrolyzer.

Er Er
ηV = = (3.23)
Ecell Er + Eloss

Another type of efficiency measured against the thermoneutral voltage Eth instead of Er
is known as high heating value (HHV) efficiency ηHHV , which considers the necessary
heat at voltages lower than Eth to maintain isothermal operation. This method is not as
popular as LHV since it gives efficiencies greater than 100% at voltages lower than Eth .

An analysis of temperature and pressure on the reversible and thermoneutral voltage has
already been made, but these can impact the polarization terms as well. As shown in
Figure 3.5a, the cell voltage decreases with temperature since all potentials decrease with
it accordingly. In the case of pressure, it increases mainly due to the reversible voltage
and it is displayed in Figure 3.5b.

(a) Temperature effect on cell voltage. (b) Pressure effect on cell voltage.

Figure 3.5: Cell voltage dependence on temperature and pressure [54].

3.4 Faraday’s Law of Electrolysis

Besides his great contributions to electromagnetism with his fundamental law of induction
formulated in 1831, Michael Faraday also made important discoveries in electrochemistry.
The famous English scientist formulated two laws for electrolysis that combined (3.24)
allows to calculate the ideal production of hydrogen ṁideal,H2 with only the cell current
Icell as input, being MH2 the molecular weight of hydrogen and nseries the number of cells
in series.

MH2 nseries
ṁideal,H2 = Icell (3.24)
zF

23
3. Thermodynamic and Electrochemical Analyses

Although most of the times this expression can be directly implemented to calculate
the hydrogen production, this law needs a correction factor that accounts for physical
phenomena that reduces hydrogen production. This factor is popularly named Faraday
or current efficiency ηF , which has a value of 92-98% in most of the operating range.

This factor is measured empirically using a mass flow meter, either Coriolis [57] or thermal
bypass, and it is divided by the theoretical mass flow ṁideal,H2 given by equation (3.24).
Both kinds of flow meter technologies are found and explained in the Bronkhorst cata-
log [58]. Other measuring set-ups can be implemented using volume flow meters, like
differential pressure or vortex, in combination with thermal and pressure sensors.

ṁH2
ηF = ∗ 100 (3.25)
ṁideal,H2

The physical phenomena causing this reduction in hydrogen production are parasitic cur-
rents, gas crossover and gas leakage. Gas leakage is generally negligible and gas crossover
is very important in PEM electrolyzers, where differential pressure between the anode and
cathode may be used. According to Hug et al. [59], the crossover flow can be determined
doing a simple test where the anode side is operated under normal conditions, but the
technician closes the cathode side using a valve as displayed in Figure 3.6.

Figure 3.6: Test schematics, applied currents and cathode’s pressure [59].

Although measuring the hydrogen crossover presents the risk of contamination and cre-
ation of an explosive environment in the the oxygen stream, it is not the main cause
of hydrogen leak in regular operation. The problem comes from the parasitic or shunt
currents, which go through the side channels causing an unintended discharge and con-
sequent reduction of the effective current going through the cell and producing hydrogen.
An analogy has been done with an AWE with bipolar arrangement as presented in Figure
3.7a, which is more common than the monopolar configuration. The resistances shown
in the inlet channels (bottom side of Figure 3.7a,3.7b) and the outlet channels (top side)
represent the ionic resistance of the KOH solution, not the channels themselves, which
are made of electrically insulating materials.

24
 3. Thermodynamic and Electrochemical Analyses

(a) Equivalent Circuit [51].

(b) Electrical Analogue [60].

Figure 3.7: Lumped Parameter Circuit.

25
3. Thermodynamic and Electrochemical Analyses

The electrical analogue of Figure 3.7b comes from an article published by Divisek et al.
[60], who created a lumped parameter model of an alkaline bipolar 30-cell stack based on
its geometrical dimensions as shown in Figure 3.7a. Solving this model with Kirchhoff’s
laws, it is obtained the electrolyzer’s shunt current distribution shown in Figure 3.8.

(a) Current density of 4000 A m-2 . (b) Current density of 50 A m-2 .

Figure 3.8: Parasitic current distribution and individual cell voltage in a thirty-cell stack [60].

From these profiles it can be observed that the shunt currents in the outlet or gas channels
present a susceptible behavior to current density variations, unlike in the electrolyte feed
channels. A second observation is that the parasitic currents in the gas channels only
become important for very low current densities. A possible theory behind this behavior
is the emergence of hydrogen and oxygen bubbles in the gas channels when there is high
current density, which increases significantly the ohmic resistivity of this path.

In a similar article [61] to the paper published by Divisek, Mergel and Schmitzto [60],
Hug et al. developed a lumped model of their electrolyzer, confirmed said model with
measurements and created a chart for the current efficiency contrasting both model and
empirical results. The results validate the conclusions reached in the previous paragraph
since the efficiency drops heavily for low density values. An additional insight is the drop
of efficiency at higher temperatures due to lower resistivity of the KOH solution.

26
 4
Alkaline Water Electrolyzer Modeling
in PowerFactory

Two articles were used as core references to develop the AWE model. The first paper
[43], published by Øystein Ulleberg in 2003, divides the model implementation in three
sections: I-U curves, thermal model and hydrogen production. Subsequently, an IEEE
team of Aalborg University published an article [62] in 2013 using Ulleberg’s outline
to develop their own model in PowerFactory. This last article served as benchmark to
design and validate the electrolyzer in DIgSILENT PowerFactory, for which they deserve
acknowledgement.

Given the lack of technical specifications for a commercial MW-scale electrolyzer beyond
their catalog performance characteristics like hydrogen production, operating temper-
ature, pressure and footprint, the implemented procedure was to base the model on the
electrolyzer installed in the PHOEBUS demonstration plant, situated in Jülich, Germany.

The PHOEBUS electrolyzer under consideration boasts a robust operational history span-
ning almost a decade, from 1994 to 2003, and its performance monitored by Ghosh et al.
[63]. Moreover, it has recurrently served as a benchmark in scientific literature, facil-
itating comparisons between theoretical models and empirical findings, as evidenced by
references such as [43] and [53]. Lastly, its operation ratings and design specifications are
shown in Table 4.1.
Table 4.1: PHOEBUS AWE operation ratings and design specifications [53].

Parameter Value Parameter Value

Electrolyzer rated power 26 kW Electrode area 0.25 m2
Number of cells in stack 21 Electrolyte KOH (30-40 wt.%)
Max. block voltage 35-37 V Anode (O2 side) Ni/Co3 O4 /Fe on Ni-plate
Max. current density [43] 3500 A m-2 Cathode (H2 side) CePt on performed Ni-plate
Max. temperature 80o C Diaphragm Ni-mesh supported NiO
Max. operation pressure 7 bar Design Zero spacing

4.1 Model definition

As mentioned before, Øystein Ulleberg [43] divided his model implementation in three
sections: I-U curves, thermal model and hydrogen production. The thermal model is out
of the thesis scope but it would have been beneficial for the general project scope, since

27
4. Alkaline Water Electrolyzer Modeling in PowerFactory

it enables to simulate thermal transients like cold-start procedure, as well as considering

additional electrical consumption such as a water pump, which extracts heat but reduces
the system’s efficiency. Nonetheless, an experimental method would be required to update
the thermal model parameters like the electrolyzer’s thermal resistance and capacitance
when scaling-up the size and electrical consumption of the device from kWs to MWs.

Table 4.1 presents the operating limits and design parameters of the electrolyzer. However,
this data does not provide information regarding cell polarizations or Faraday efficiency,
both necessary to analyze the AWE performance at a specific point of operation. The
calculation of both properties is possible using equations (4.1), (4.2) and the parameters
listed in Table 4.2 introduced by Øystein Ulleberg and the Aalborg University team.

t2 t3
Ecell = Er (T, P ) + (r1 + r2 T )Jcell + s log((t1 + + 2 )Jcell + 1) (4.1)
T T
a2 + a3 T a4 + a5 T
ηF = a1 exp ( + ) (4.2)
Jcell Jcell 2

Table 4.2: PHOEBUS AWE I-U curves and Faraday Efficiency parameters [43],[62].

Parameter Parameter description Value Unit

r1 Ohmic parameter 1 8.05 x 10-5 Ω m2
r2 Ohmic parameter 2 -2.5 x 10-7 Ω m2 /o C
s Coefficient of electrode overvoltage 0.185 V
t1 Overvoltage coefficient 1 -0.1 m2 /A
t2 Overvoltage coefficient 2 8.424 m2 o C/A
t3 Overvoltage coefficient 3 247.3 m2 o C2 /A
a1 Faraday efficiency coefficient 1 99.5 %
a2 Faraday efficiency coefficient 2 -9.578 A/m2
a3 Faraday efficiency coefficient 3 -0.055 A (m2 o C)-1
a4 Faraday efficiency coefficient 4 1502.708 A2 /m4
a5 Faraday efficiency coefficient 5 -70.8 A2 (m4 o C)-1

The first step to calculate the cell potential Ecell in equation (4.1) is introducing the
reversible voltage Er at operation conditions, which is calculated as described in section
3.2. The second term of the equation reflects the ohmic polarization, which shows a linear
behavior with temperature, while the third term accounts for the activation overvoltage
which presents a logarithmic behavior. It is noticeable that the operating pressure affects
the reversible potential, but not the polarization terms.

In both figures 4.1 and 4.2, it is performed a temperature sweep from 20 to 80o C and the
voltage and current efficiencies are calculated. This temperature range depicts the nom-
inal behavior of the electrolyzer from ambient till operating temperature at atmospheric
pressure. There is no value in obtaining curves above this point since this is the highest
operating temperature for industrial AWEs. Besides, working close to 100o C carries the
risk of reaching the boiling point of water.

28
 4. Alkaline Water Electrolyzer Modeling in PowerFactory

Figure 4.1: Characteristic IU curves and Voltage Efficiency.

In [62], it is proposed the empirical equation (4.2) to model Faraday efficiency, which
is shown in Figure 4.2. This expression reflects the concepts explained in section 3.4,
accounting for the parasitic currents and gass crossover.

Figure 4.2: Faraday Efficiency.

29
4. Alkaline Water Electrolyzer Modeling in PowerFactory

4.2 Power increase from kW electrolyzer to MW-level

To scale up the electrolyzer, it was necessary to find a suitable candidate among different
manufacturer catalogues. The chosen manufacturer was Nel Hydrogen given its domin-
ant position in the market together with Cummins and Siemens Energy [64], as well as
their products similar operation characteristics with the 26 kW PHOEBUS electrolyzer
as shown in Table 4.3. Another criteria was that this company offers modular products
like module A3880 which consists of 8 stacks-A485.

Table 4.3: Series A Nel Hydrogen Atmospheric Alkaline Water Electrolyser [44].

Specifications A485 A1000 A3880 Units

Net Production Rate ṅH2 300-485 600-970 2400-3880 Nm3 /h
ṁH2 647-1046 1295-2094 5180-8374 kg/24 h
Production Capacity Range 15-100 15-100 15-100 %
Efficiency at Stack η 3.8-4.4 3.8-4.4 3.8-4.4 kWh/Nm3
42.22-48.88 42.22-48.88 42.22-48.88 kWh/kg
H2 Purity 99.99 99.99 99.9 %
Delivery Pressure 1–200 1–200 1–200 barg
Dimensions 225 350 770 m2
Ambient Temperature 2-35 2-35 2-35 o
C
Electrolyte 25 25 25 % KOH
Operating Temperature 80 80 80 o
C
H2 outlet pressure 200–400 200–400 200–400 mm WG
after electrolyzer 0.02-0.04 0.02-0.04 0.02-0.04 barg

Since it was not possible to get more accurate data like the number of cells nseries or their
voltage-current curves, some assumptions and deductions have been made to build the
model. The logical derivation starts with the definition of power consumption Pdc , which
can be expressed as

Pdc = Vdc Idc = nseries Ecell Acell Jcell (4.3)

where Vdc and Idc are the stack voltage and current respectively, each decomposed into
their elemental components. In Table 4.3, it is given the hydrogen flow capacity ṁH2
which determines the product of number of cells times maximum current (nseries ∗ Imax )
using Faraday’s Law (3.24). Lastly, it is plausible to consider that all Series A modules
are agglomerations of A485, which can be induced by comparing their production rates
as indicated in Table 4.4.

Table 4.4: Deductions of Series A.

Module No stacks ṁH2 [kg/s] nseries ∗ Imax [kA]

A485 1 0.012 1168.541
A1000 2 0.024 2339.761
A3880 8 0.097 9351.004

30
 4. Alkaline Water Electrolyzer Modeling in PowerFactory

Same as in Figure 4.1, it is selected an individual cell voltage Ecell of 1.8 V at operating
temperature and maximum current density Jmax of 3500 A m-2 . Lastly, it is assumed 500
cells for nseries and an electrode cell area Acell of 0.6 m2 . For stack A485, this results in
the ratings described in Table 4.5.

Table 4.5: A485 assumptions and operation ratings.

Assumptions Results
Temperature, T 80 C
o
Nominal voltage, Udc,nom (80o C) 900 V
Max. current density, Jmax 3500 A m-2 Max. voltage, Udc,max (25o C) 1050 V
Cells in series, nseries 500 Max. current Istack,max 2100 A
Area, Acell 0.6 m2 Nominal power, Pdc,nom (80o C) 1.9 MW
Cell voltage, Ecell (80o C) 1.8 V Max. power,Pdc,max (25o C) 2.2 MW

In Figure 4.3, a visual comparison is made between the 26 kW electrolyzer used in the
PHOEBUS demonstration plant and the 1.9 MW Nel Hydrogen A485 electrolyzer.

(b) Nel Hydrogen Alkaline Electrolyzer A485 [44].

(a) 26 kW electrolyzer used in the

PHOEBUS demonstration plant [65].

Figure 4.3: Scale-up from kW to MW level electrolyzer.

4.3 Dynamic Model Implementation in PowerFactory

Once the specifications for the model were defined, it could be developed in PowerFactory.
The aforementioned article, published by an IEEE team of Aalborg University in 2013
[62], served as guideline for the implementation of AWE in DIgSILENT PowerFactory.
The book "Modelling and Simulation of Power Electronic Converter Dominated Power
Systems in PowerFactory" [66] written by Francisco M. Gonzalez-Longatt and José Luis
Rueda Torres was also used as literature reference for developing models in DIgSILENT.

31
4. Alkaline Water Electrolyzer Modeling in PowerFactory

The model definition prompted the need for substantial troubleshooting efforts, given the
operational challenges of the software. Also many approaches were taken, but finally only
two designs were considered for its implementation:

• General load (Model 1): simulating the whole system as a general AC load where it
is possible to tune the active P and reactive power Q independently to control the
power flow. This solution was based on article [62].

• Converter + voltage source (Model 2): locating a current ipu controller in the con-
verter and implementing the IU curves in the voltage source block.

Model 1 allows voltage dependence of the load when performing load flow calculations
since the whole system (rectifier and electrolyzer) is a "black-box" or PQ load. This
setup allows dynamic simulation known in PowerFactory as RMS simulation, but not
instantaneous analysis of waveforms inside the rectifier or EMT simulation. In their
article [62], Iker Diaz de Cerio Mendaza, Birgitte Bak-Jensen and Zhe Chen implemented
the model assuming that the sole demand of reactive power is the one consumed by the
compressor. This is not general, since it depends heavily on the converter topology and
varies depending on the loading as described in Chapter 5. The power quality definition
of the load can be introduced to change the converter type. Overall, this model enables
good and fast analysis of the load.

On the other hand, Model 2 enables a more thorough and comprehensive understanding
of the system by segregating both the power electronics and the load. It is very suited
for dynamic simulation (RMS) and enables the possibility of analyzing the instantaneous
spectrum of the rectifier (EMT) and to change the harmonics contents produced by differ-
ent types of rectifier topologies (6-pulse thyristor, 12-pulse thyristor, PWM with IGBT).
However, it supposes an independent handling of the load flow analysis and the dynamic
behavior in RMS/EMT simulation.

Both models are recommended for implementation in future projects, but this thesis
will be based on the second option. System modelling for stability and electromagnetic
transient analysis (RMS/EMT) is a critical issue in the power system analysis. Given
this level of complexity in time-domain simulations, PowerFactory modelling philosophy
presents a strictly hierarchical system modelling approach, which combines both graphical
and script-based modelling methods. The basis for this modelling approach is presented
in the following points, from the highest level to the lowest one:

• In the highest level it is found the "composite models". These objects are defined
by a "composite frame", which combine and interconnect "built-in models" and
"common models". In simpler words, the composite frames are a type-file which
enables to reuse a basic structure for a composite model.

• "Built-in models" are the transient PowerFactory models for standard equipment,
i.e generators, motors, loads. The "common models" are based on "DSL block
definitions" and are the front-end of the user-defined transient models.

• The "DSL block definitions", based on the "DIgSILENT Simulation Language"

(DSL), form the basic building blocks to represent transfer functions and differential
equations for the control systems.

32
 4. Alkaline Water Electrolyzer Modeling in PowerFactory

For terminology purposes, AWE or electrolyzer system will be referring to the whole
module: converter and electrolyzer. In Figure 4.4, it is shown the AWE system composite
frame, which includes all necessary blocks for building the model:

• Measurement devices like phase-locked loop or PLL (.ElmPhi), current measurement

(.StaImea) and, since this project does not count with a thermal model, a reference
block for pressure and temperature (.ElmDsl).

• Model Definitions like the current controller (.ElmDsl), the AWE IU curves (.Elm-
Dsl) and the hydrogen production block (.ElmDsl).

• Two built-in models to represent the voltage source (.ElmDcu) and the rectifier
(.ElmRecmono).

Figure 4.4: AWE system composite frame (BlkDef. Test Voltage Source).

Diving into the AWE IU curves block definition: this module gets the operating temper-
ature T and pressure P as inputs which allows to determine the operating IU curve, the
reversible Er and thermoneutral potential Eth . With the last two properties it is possible
to calculate the instantaneous efficiency of the electrolyzer ηLHV and ηHHV . However, the
IU curves are not implemented using the equations of subsection 4.1 since the objective
of this script is to be repeatable for other electrolyzers with different curves. Instead a
look-up table is used with temperature T and current density Jcell as inputs. All these
schematics are illustrated in Figure 4.5.

33
4. Alkaline Water Electrolyzer Modeling in PowerFactory

Figure 4.5: AWE IU curves block (BlkDef. IU curves).

An important point to highlight is that the current measurement ipu is defined per unit,
but the look up table for single cell IU curves is not, so per unit conversion is required.
Afterwards, ipu is converted to its absolute value due to the DC voltage source (.ElmDcu)
sign criteria. Conversion back to per unit is done with the voltage Ustack at the output.
At the composite frame Figure 4.4, this value uset is introduced into the "Voltage Load
Slot", where this value updates the voltage source element.

Finally, the current density Jcell and cell current Icell measured in standard units are taken
as input for the hydrogen production block (BlkDef. Faraday Law). As observed in Figure
4.6, the left block applies Faraday’s law with only current as input and gives the mass
flow ṁH2 in kg/s. The look-up table below gives the Faraday efficiency using the curves
of Figure 4.2. Finally, the mass ṁH2 and volume ṅH2 flow, expressed in kg/h and Nm3 /h
respectively, are obtained applying hydrogen density at atmospheric pressure.

Figure 4.6: Hydrogen production block (BlkDef. Faraday Law).

Lastly, the block definition of the current controller is presented in the following sections.

34
 5
Converter-load interaction and control

The AWE IU curves and H2 production model definitions developed in section 4.3 are valid
independently of the converter topology. This fact, besides the dependence of the con-
verter technology on the semiconductor topology, makes it necessary to dedicate this part
a separated section. Before delving into the controller design, a concise preliminary study
has been conducted. Electrolyzers require high current rectifier topologies to operate: at
least 2.1 kA DC for the model at hand, so only some technologies are evaluated.

5.1 Prestudy

Figure 5.1: Active Components Ratings in Power Electronics [67].

It is possible to classify rectifier topologies into three categories depending on its semi-
conductor technology: diode, thyristor and IGBT based.

Diode-based rectifiers are typically overlooked since they are non-controllable devices.
Nontheless, there is a viable approach involving diode rectifiers in conjunction with a

35
5. Converter-load interaction and control

DC/DC converter to achieve voltage regulation [68]. Such systems, referred to as rectifier-
DC choppers, find extensive use in applications with power ratings in the kilowatt range
but prove inadequate for megawatt-level tasks. This leaves only rectifiers based on thyris-
tors and IGBTs for further examination. Figure 5.1 depicts the limits of voltage, current
and frequency across various semiconductor technologies employed in power electronics.

The thyristor stands out as the dominant component in managing high electrical loads. It
excels in handling voltages exceeding 8 kV and currents surpassing 2.2 kA. Nevertheless,
they are slow devices with switching frequencies no exceeding 100 Hz. Semi-controllable
in nature, the thyristor can be activated at will but does not have turn-off capabilities.

In contrast, the IGBT emerges as an active component without the drawbacks of thyris-
tors. However, its proficiency in handling voltage and current is comparatively restricted.
For applications demanding high voltage, the solution often entails stacking several IGBTs
in series within modular converters.

Currently, thyristor-based rectifiers dominate the market including industrial and Power-
to-Gas applications. They are mature technologies that offer high efficiency, reliability
and good control of the current, but unfortunately, they suffer from degrading the power
quality of the AC grid by injecting low-frequency harmonics, consuming undesired reactive
power and having high DC current ripple. Its standard converter is the 6 pulse thyristor-
based rectifier (6-PTR).

There are multiple configurations to decrease the thyristor line RMS current as well as the
low-frequency harmonics and output current ripple: 12-pulse, 24-pulse, 48-pulse, double-
star rectifier [68].

Noteworthy progress has been achieved in the realm of IGBTs for medium-voltage and
power applications, leading to their integration into PWM current source rectifiers (PWM-
CSR). However, this configuration proves unsuitable for high-power applications, mainly
due to the substantial voltage and current demands that elevate conduction losses, ulti-
mately undermining energy efficiency.

Figure 5.2: PWM-CSRs [68].

To compensate these disadvantages usually multilevel converters are implemented. An

example is the three-level Neutral Point Clamping (NPC) converter shown in Figure

36
 5. Converter-load interaction and control

5.3. These converters utilize multiple voltage levels to achieve efficient power conversion.
However, they suffer from increased high voltage and current stress, which leads to higher
conduction losses and reduced energy efficiency compared to thyristor-based rectifiers.
Other options include the Modular Multilevel Converter (MMC) with half-bridge or H-
bridge modules [67].

Figure 5.3: Three-level Neutral Point Clamping converter.

In essence, there are two primary configurations available for rectifiers: those based on
thyristors and those utilizing IGBT power electronics. In modern applications, where
power quality and harmonic mitigation are critical, more advanced converter topologies
such as active front-end converters or multilevel converters are preferred.

It is anticipated that IGBT converters will exhibit competitive costs compared to thyristor-
based solutions in the future since they are exempt of needing harmonic-tuned filters to
comply with emissions regulation and reactive power compensation devices. Nevertheless,
thyristor-based rectifiers remain the preferred option for MW-level electrolyzer applica-
tions due to their superior efficiency and reliability resulting from the utilization of fewer
components and its higher ratings [69],[70].

5.2 Power converters in PowerFactory

As it was commented in section 4.3, there are two main possibilities to model the whole
electrolyzer system: decoupling the power electronics and DC behavior or modeling the
whole as a general load with a predetermined level of harmonics. Since it was decided to
follow the first alternative, next step in the modeling process is to implement the converter
and the controller behind it.

There are various alternatives available to implement converters in PowerFactory, as de-

picted in Figure 5.4. The highlighted elements displayed in the image below enable the
modeling of different types of converter topologies, unlike the unmarked ones, which rep-
resent specific topologies of their own, such as HVDC, Softstarters, TCSC and SSSC.

• Rectifier/Inverter (.ElmRec, .ElmRecmono)

• PWM Converter (.ElmVsc, .ElmVscmono)

• DC-DC Converter (.ElmDcdc, .ElmDcdcbi)

37
5. Converter-load interaction and control

• DC Valve (.ElmValve)

Figure 5.4: Power Electronic Devices available in PowerFactory.

The converter implemented in this thesis is AC-DC type, thus excluding the study of
the DC-DC Converters. As the intention is to utilize a prebuilt version of the converter,
the examination of individual DC valve elements such as diodes, thyristors and IGBTs is
omitted. Therefore, only two options follow:

• The Rectifier/Inverter can be modeled either as a full-bridge diode rectifier or a

6-pulse thyristor converter (selection as rectifier or inverter orientation).

• The PWM Converter model which can be configured to represent both a self-
commutated, voltage source AC/DC two-level converter or modular multilevel con-
verter.

Model Definition type .ElmRec allows three kinds of model for rectifier implementation:

• Three-phase diode rectifier: this option is disregarded since it is an uncontrolled

rectifier topology, which gives no regulation over the hydrogen production.

• Three-phase line-commutated (LCC) rectifier/inverter: this model is the preferred

choice since it is the standard topology for 6-PTR and the most robustly designed
and controlled.

• Three-phase capacitor-commutated (CCC) rectifier/inverter: this configuration is

the same as the LCC but with series capacitors between thyristors and the com-
mutation reactances. This topology allows to advance the commutating voltage over
the thyristors, resulting in a lower firing angle and therefore drawing less reactive
power of the grid. However, it has a difficult design and resonance phenomena to
consider, not the best model for a prototype.

5.3 6 and 12 pulse three phase thyristor-based rectifiers

The "6-pulse" designation refers to the arrangement of six thyristors in the rectifier cir-
cuit. These thyristors are semi-controlled switches that allow current to flow in only one
direction, effectively rectifying the AC input into a smoother DC output. The arrange-
ment consists of three pairs of thyristors, forming a three-phase full-wave bridge rectifier
configuration. This converter corresponds to the LCC rectifier described before and shown
in Figure 5.5.

38
 5. Converter-load interaction and control

Figure 5.5: PowerFactory three-phase LCC rectifier model.

This converter is often connected to grid through a step-down transformer and its leakage
inductance usually represents most of the commutation inductance Lc . The rectifier is
connected to a DC load through the inductance Ld intended to reduce the current ripple.

Assuming the case for an uncontrolled 6-PTR which would correspond to a three phase
full-wave bridge diode rectifier, it is possible to define the "ideal no-load DC voltage" Vd0
in equation (5.1) being VLL the rms line to line voltage.

√
3 2
Vd0 = VLL ≈ 1.35VLL (5.1)
π

The thyristors are turned on and off at specific times to control the flow of current through
the circuit. However, current must reach zero by itself for turning-off the switch which
disables the use of pulse width modulation (PWM). Instead, the current flowing through
each thyristor can be delayed using the ignition angle α which determines the DC voltage.

Vdα = Vd0 cos α (5.2)

Nonetheless, to reflect more realistic converters the voltage drop ∆Vd caused by current
commutation from one valve to the next one must be contemplated.

√ √
3 2 3 2 3ωLc
Vd = VLL cos α − ∆Vd = VLL cos α − Id (5.3)
π π π

Notice that if the ignition angle becomes bigger than 90o , the DC voltage turns negative
where it enters into the inverter mode. Since for our application it is only desired the
rectifier mode, α will always operate in the range [0, 90o ] providing maximum active power
at low firing angles and minimum active power at high firing angles.

39
5. Converter-load interaction and control

Figure 5.6: Normalized DC voltage Vd /Vd0 as function of the firing angle α assuming no com-
mutation inductance Lc [71].

In Figure 5.7, the instantaneous AC voltage, AC current, DC voltage, firing angle α and
commutation angle µ corresponding to the operation of the converter are presented. It
can be appreciated the transition of the currents with the commutation angle and its
effect on the DC waveform and its average Vd .

Figure 5.7: Phase voltages, currents and DC voltage of a 6-PTR working with a firing angle
α=30o and a commutation angle µ=20o .

40
 5. Converter-load interaction and control

The current waveforms exhibit an ideal rectangular shape when disregarding the com-
mutation angle µ. Assuming his idealization, a Fourier decomposition of the currents into
their harmonic components will be done. In the context of a single-phase rectifier, the
rectangular waveform is broken down into its odd harmonics. However, in the case of a
three-phase system, all third-multiple (3h with h ∈ N) or zero sequence current harmonics
vanish. Consequently, what remains are the non-triple odd harmonics (6h±1 with h ∈
N). Each of these possesses a value inversely proportional to its harmonic order and is
characterized by its fundamental value as described in equation (5.4) for phase A.

√
6 IˆA(1)
IˆA(1) = Id IˆA(h) = (5.4)
π h

Implementing a 6-PTR for a high current load, like an AWE system, becomes impractical
given the magnitude of its harmonics [71]. Fortunately, there is a simple cancellation
technique for some of these low frequency harmonics which consists into parallel two 6-
PTR with a 30o phase shifted three-winding transformer. √The transformer must have a
Yy0 and Y△11 winding configuration with a turn ratio 1: 3:1 as shown in Figure 5.8.

Figure 5.8: 12-pulse thyristor rectifier [70].

The implementation of a 12-PTR allows to cancel those harmonic orders 6h±1 with h =
1, 3, 5 ... This is shown in Figure 5.9 and it is also compared the ideal harmonic distortion
distribution in 6 and 12-PTR. The non-cancelled harmonic orders are 12h±1 with h ∈ N,
which are typically managed with specific harmonic-tuned filters.

In Figure 5.9, the applied harmonic distortion distribution presents a lower value than
the ideal scenario for the non-cancelled harmonic orders (11th , 13th , 23rd , 25th ...), but
not a complete cancellation for the rest. The main explanation is that the current does
not present a perfect square-step waveform caused by the commutation angle µ and the
parasitic components of the thyristors. Another factor to consider in the 12 PTR topology
is that both bridges may not switch exactly as intended in the controller, which introduces
a relative small time error that also affects the current waveform. Overall, the applied
Total Harmonic Disorder (THD) results in a lower level than the ideal, as indicated in
Table 5.2. The THDs presented in this table are calculated until the 50th harmonic order.

41
5. Converter-load interaction and control

Figure 5.9: Harmonic distortion distribution for 6 and 12-PTR topologies.

Since the categorical difference between a 6 and 12-PTR is the harmonic distortion dis-
tribution of the current, the 12-PTR can be simulated in PowerFactory as an equivalent
6-PTR but manipulating the harmonic contents. The table implemented in the model is
described in Table 5.1.

Table 5.1: Implemented Harmonic Distortion in 12-PTR [72].

Harmonic HD [%] Angle [º] Harmonic HD [%] Angle [º]

5 2.3 0 29 0.5 0
7 2 0 31 0.5 0
11 7 -4.4 35 1.4 215.6
13 6 -8.3 37 1.2 201.7
17 1 0 41 0.4 0
19 0.9 0 43 0.4 0
23 1.5 285.8 47 0.4 488.5
25 1.4 269.3 49 0.4 477.5

Table 5.2: Total Harmonic Disorder (THD) for 6 and 12 PTR topologies.

Rectifier THD
6-PTR Ideal 30.02%
12-PTR Ideal 14.17%
12-PTR Applied 10.24%

The topology and the rectifier operating principles have now been explained, so in the rest
of the sections, the system is studied with rms values to develop the current controller.
The electromagnetic transient (EMT) operation of the converter is left out of the scope
from a control point of view.

42
 5. Converter-load interaction and control

5.4 Controller Implementation

Since the primary objective of the plant is to control the hydrogen production ṁH2 , this is
the reference input of the controller. Besides, given its direct proportionality with the DC
current Idc , both variables have the same values in per unit. In fact, the desire of having
the same controller independently of how many stacks are put in parallel or series makes
a per unit controller the ideal solution. In the next subsections, a proportional-integral
(PI) controller for the hydrogen flow is designed using the Internal Model Control (IMC)
tuning method [73]. This process has 3 design steps:

1. Process model

• Describe the process with equations

• Estimate the process with measurements

• Verify the process model

2. Controller design

• RL circuit control

• Feed-forward and anti-windup introduction

3. System analysis

• Is the system behaving as expected? Testing and validation

5.4.1 Process model

This part entails to describe the behavior of the process model, in this case, the electro-
lyzer. This device is far from a simple electric circuit model as described in a previous
section, but to design the hydrogen controller some simplifications are made. The elec-
trolyzer process model is considered as a simple resistance in series with a voltage source,
although this assumption ignores the activation losses. Two linear process models are
proposed as candidates to design the controller with different slopes and y-intercepts.

(a) Comparison. (b) Equivalent circuit.

Figure 5.10: Analogy of electrolyzer with resistance and voltage source.

43
5. Converter-load interaction and control

5.4.2 Controller design

As mentioned before, the reference input for the controller is the objective DC current
in per unit iref,pu . This value corresponds also to the hydrogen production vH2 ,ref,pu
established by Faraday’s Law as well as the electrolyzer’s loading.

This controller works based on average and DC values, ignoring the electrical transient
nature of a rectifier with instantaneous variations like current ripples. In Figure 5.11, it
is shown the PI controller for a RL circuit represented by the transfer function Gc (S).

Figure 5.11: Current controller for RL circuit.

Nevertheless, the electrolyzer Gc (S) is not comparable to a RL circuit, it also has a

reversible voltage Er that acts like a disturbance D. In order to control the current as it
would be for the RL circuit, a feed-forward D̂ is introduced in the controller to remove
the disturbance D. No active damping is implemented. This way, the closed-loop system
Gcl (S) is designed to behave like a first order system.

Figure 5.12: Introduction of feed-forward D̂ in the current controller.

Another necessary implementation in the controller is the anti-windup. The converter

is not an ideal source of DC voltage. It has a limit in the maximum deliverable output
voltage Vdc which depends on the AC voltage of the grid. The conditional integration anti-
windup method switches off the integral part of the controller to avoid windup effects when
y ≥ ymax and y ≤ ymin . As reference, it is implemented the PI block with an anti-windup
limiter in accordance with the IEEE Standard 421.5-2016 [74] as described below

If y ≥ ymax : yo = ymax and ẋ = 0

If y ≤ ymin : yo = ymin and ẋ = 0
Otherwise : yo = y = Kp yi + x and ẋ = Ki yi

44
 5. Converter-load interaction and control

5.4.3 System analysis

Upon completion of the controller design, the subsequent step involves studying and
validating the response of the closed-loop system Gcl (S), but this time integrating the
actual AWE system. Both the simplified schematic and real controller implemented in
PowerFactory are illustrated in Figure 5.13 and 5.14, respectively.

Figure 5.13: Proportional-integral current controller with feed-forward and anti-windup limiter.

Figure 5.14: Current controller block definition (BlkDef. Current Control).

In Figure 5.14, it is shown that the current controller receives two inputs: the DC current
measurement ir and the operating temperature T . The controller provides two outputs:
the rectifier’s firing angle α and the maximum voltage in per unit Emax,pu .

Temperature T is used for calculating the reversible Er and maximum Emax potentials at
that given temperature. These potentials are then multiplied by the number of cells in
series nseries and divided by the base voltage Vbase to obtain per unit values.

The current reference iref,pu gets subtracted by the DC current measurement ir giving the
total error. This value gets inside the PI controller with anti-windup, where boundaries
are defined so Vpu,target , after feed-forward, gives a maximum voltage of 1 p.u. Afterwards,
this voltage is transformed into the necessary firing angle α to obtain Vpu,target .

45
5. Converter-load interaction and control

The IMC method is a robust approach for tuning controller parameters [73]. In this
case, the proportional KP and integral KI constants can be tuned by considering the
desired rise time trise for the closed-loop system Gcl (S) and the circuit’s resistance and
inductance. Here, trise represents the time required to reach from 10 to 90% of iref,pu and
can be calculated using equation (5.5), where αc denotes the system’s bandwidth.

ln(9)
trise = (5.5)
αc

Two sets of control parameters may be implemented considering the two linear process
models defined in Section 5.4.1. Both KP and KI can be calculated through equations 5.6
and 5.7 with lload and rload representing the inductance and resistance of the DC load in
per unit. Since these variables are calculated in per unit, they must be divided by Rbase
which is 0.5 Ω, calculated with the electrolyzer maximum ratings.

KP = lload ∗ αc (5.6)
KI = rload ∗ αc (5.7)

For this electrolyzer model, the considered DC inductance Lload is null, which makes the
proportional constant KP zero for Case 1 and 2. This inductance must not be confused
by Ld , DC inductance connected at each output of the thyristor rectifiers. The purpose
of Ld is to reduce the voltage ripple output in an EMT environment, while Lload is the
inherent inductance of the DC load, in this case the electrolyzer.

Additionally, a third case is suggested imposing a proportional constant KP to compare

the performance of the system despite the absence of inherent DC inductance. The control
parameter sets are tabulated in Table 5.3.

Table 5.3: Candidates for final controller.

Parameter Case 1 Case 2 Case 3

Rload 0.147 Ω 0.085 Ω 0.085 Ω
rload 0.294 p.u. 0.171 p.u. 0.171 p.u.
trise 2s 2s <2 s
KP 0 0 0.1
KI 0.32365 0.18756 0.18756
αc 1.1 rad/s 1.1 rad/s <1.1 rad/s

Applying these PI configurations and performing a ramp-up analysis, it is obtained three

ramp-up scenarios, which represent the voltage and current at the DC side of the electro-
lyzer. All cases are performed with an operating temperature of 80o C.

In Figure 5.15 it is appreciated that all controllers, independently of their configuration,

have a big time-delay in the first 0.2 p.u. step, and none of them fulfill the theoretical
rise time trise of 2 seconds. This is due to the simplification of the load which ignores
the logarithmic component of the voltage equation that represents the activation losses.
Besides that, all controllers work adequately.

46
 5. Converter-load interaction and control

Figure 5.15: Time-domain response of the system with different PI parameters.

Case 2 uses the most simple first order response while adjusting to the real electrolyzer
behavior, given the similarity of Process Model 2 to the electrolyzer, shown in the IU
curves. These arguments lead to the design of an integral controller based on Case 2
parameters.

Another factor to consider in the system analysis is the temperature influence on the load.
Two ramp-up simulations at ambient and nominal temperature are compared in Figure
5.16. As expected, operating at ambient temperature implies working at higher voltages
to obtain the same desired hydrogen flow. The most noticeable contrast is the first 0.2
p.u. step, where the difference between activation potentials is the highest.

Figure 5.16: Time-domain response of the system with different operating temperatures.

47
5. Converter-load interaction and control

The empirical rise time for each step using this controller is illustrated in Table 5.4. Due
to the higher voltage requirements at 25o C, the integration of the current error takes
longer to reach steady state compared to operating it at 80o C. This leads to a higher
delay in the rise time at 25o C.

Table 5.4: Rise time at ambient and operating temperature.

iref trise (25o C) trise (80o C)

0-20% 4s 3.13 s
20-40% 1.57 s 1.4 s
40-60% 1.23 s 1.13 s
60-80% 1.09 s 0.98 s
80-100% 1s 0.92 s

The result of applying operating temperatures of 25o C and 80o C in the electrolyzer gives
different firing angles α, as depicted in Figure 5.17. Operating at higher voltages implies
lower firing angles due to the cosine function, introduced in equation (5.2).

Figure 5.17: Stack voltage Vpu and firing angle α at different temperatures.

5.4.4 Power consumption analysis

Once the DC current and voltage are calculated and a ramp-up analysis has been made,
the next step was to analyze the AC behavior of the AWE system. Three scenarios are
simulated using the single line diagram (SLD) shown in Figure 5.18, where the active
PAC and reactive QAC power consumption of the rectifier is mapped by increasing the
hydrogen production rate or loading.

48
 5. Converter-load interaction and control

Figure 5.18: Grid single bus.

First and second scenario assume an ideal external grid with negligible leakage reactance
in the step-down transformers and, therefore, no voltage drop across them. First scenario
(dashed lines) assumes an operating temperature of 25o C while second scenario (bold line
- blue) assumes 80o C. Finally, the third scenario (bold line - green) assumes a leakage
reactance of 0.1 p.u. on each transformer and 80o C operating temperature. Second and
third scenario share the same active power curve (bold line - red).

Figure 5.19: Active PAC and reactive QAC power consumption as function of loading level.

Since switching and conduction losses are ignored for the converter, the consumption of
active power in AC and DC sides is the same. It may be observed in Figure 5.19 that active
power consumption PAC reaches 2.2 MW and 1.9 MW for 25o C and 80o C, respectively.
This is logical since the voltage at maximum loading in the IU curve for a single cell is
2.1 V at 25o C and 1.8 V at 80o C.

In terms of reactive power, the root cause of consumption in the rectifier is the firing angle
α. As it can be observed in Figure 5.20, the regulation of the firing angle intrinsically
delays the AC current which decreases the power factor P F . Additionally, in Figure 5.17
it can be observed that the higher the loading, the lower is the required firing angle. The
reactive power consumption is the result of the multiplication of both factors, the AC
current and sine function of the firing angle in the loading range.

49
5. Converter-load interaction and control

Figure 5.20: Line current wave-forms in a 6-PTR as function of firing angle α assuming no
commutation reactance [71].

As already mentioned, the main difference between scenario 2 and 3 is the introduction
of a reactance before the converter. It may be observed that QAC for scenario 3 is lower
than for scenario 2. This is due to the voltage drop across the transformers reactances,
which decrease the AC voltage at the rectifier bus. Maintaining the same current range,
it decreases the reactive power consumption, but not the active power demand.

Lastly, a visualization of the stack efficiency in terms of ηHHV and ηLHV is shown below.
Here it is visible the limitation of 15% as minimum operating limit for the electrolyzer due
to the fact that, below this current density, the electrolyzer works at a lower voltage than
the thermoneutral potential Eth which makes the electrolyzer an endothermic device. At
this point, the electrolyzer is 100% efficient from a thermal analysis.

Figure 5.21: Efficiency Chart.

50
 6
Structure Design for Electrolyzer
stations at GW-scale

The technical design of the plant is based on two reference projects: Electric infrastructure
for electrolyser systems thesis [56] and Hydrohub GigaWatt-Scale Electrolyzer project [69].

The first one presents the electric infrastructure of a station with state of the art techno-
logy using 3 voltages levels 400/150/(66 & 33) kV between the Point of Common Coupling
(PCC) and a transformer-thyristor based rectifier module. On the other hand, the Hy-
drohub project includes a design with only 2 voltage levels 380kV/66kV between PCC
and a transformer-IGBT based rectifier module.

Hydrohub project presents a global vision for building a GigaWatt green hydrogen plant
considering not only the electrical installation, but process units, civil, structural works,
piping, process automation and utilities. It also includes an economic evaluation of the
project implementing both electrolyzer technologies, AWE and PEM, and providing an-
ticipated investment cost estimations.

Nevertheless, the Hydrohub project is based on the assumption of having an "Advanced

Design for 2030", which presumes technological advancements like IGBT enhanced charac-
teristics, higher current densities and efficiencies for the electrolyzers enabling a reduction
in the stack size and, therefore, in the station footprint. This thesis presents a compen-
dium between both projects, as well as a more detailed plant design.

6.1 Station Design Criteria

The conventional power station is typically linked to a HVAC grid, specifically designed to
accommodate extremely high loads like this case study. As a general rule, higher demand
requires higher voltage to efficiently manage the current, thereby reducing losses.

Since electrolyzers are relatively-low voltage devices compared to the PCC voltage mag-
nitude, it is necessary to provide a step-down infrastructure to connect the loads to the
grid. For terminology purposes, this infrastructure is called distribution grid. They can
be classified in three categories depending on the nature of the HV line and the connection
of DC loads. These distributions grids are described in Figure 6.1.

51
6. Structure Design for Electrolyzer stations at GW-scale

(a) AC distribution grid connected to HVAC grid through step-down transformers.

(b) DC distribution grid connected to HVAC grid through a large rectifier.

(c) DC distribution grid connected to HVDC grid through a large DC/DC converter.

Figure 6.1: Distribution grids for an electrolyzer station comprised of system modules [56].

The scope of the thesis was to analyze the technical feasibility of a GW-scale station
connected to the traditional grid, therefore, the third distribution system has been disreg-
arded. However, it may prove convenient to analyze in a future project since it unlocks
the possibility to connect directly a renewable power plant to the substation skipping
the intermediate AC/DC conversion infrastructure. The selection of 12-PTR as converter
module for the electrolyzers leaves the AC distribution grid as the only option.

In Figure 6.2, it is presented an example of a SLD for a GW-scale electrolyzer station

developed in Electric infrastructure for electrolyser systems thesis. This station is imple-
mented in four different scenarios. All cases have a common structure of 3 voltage levels
400/150/Vbus kV with a Static Synchronous Compensator (STATCOM) connected to the
PCC busbar, three 400/150 kV transformers and the electrolyzer systems connected to
Vbus busbar.

52
 6. Structure Design for Electrolyzer stations at GW-scale

Figure 6.2: Electrolyzer station base line diagram [56].

The implementation of 66 kV or 33 kV at the electrolyzer system terminal, and the

parallel connection or not of the 400kV/150KV transformers impacts the specifications
of the station and the load shedding consequences in case of fault. The specification
differences are summarized in Table 6.1.

Table 6.1: Structure comparison for cases 1-4 [56].

Parameter Case 1 Case 2 Case 3 Case 4

Vbus 33 kV 66kV 33 kV 66kV
In,bus 3.0 kA 2.66 kA 3.0 kA 2.66 kA
ISCC 27 kA 22 kA 21 kA 15 kA
Sn,T (400/150 kV) 570 MVA 570 MVA 400 MVA 400 MVA
Sn,T (150/Vbus kV) 172 MVA 300 MVA 172 MVA 300 MVA
No transformers 7 4 7 4

While case 1 and 2 have the 400/150 kV transformers connected in parallel, case 3 and 4
have them separated. The first approach is better in terms of redundancy and efficiency,
since it avoids load shedding in case of failure of one transformer and sharing out the
load between the transformers reduces the overall copper losses. However, this approach
is not convenient in terms of protection. For example, a fault in the 150 kV busbar
would interrupt the operation of the whole plant whereas in case 3 and 4 only a third

53
6. Structure Design for Electrolyzer stations at GW-scale

of the station would be offline. Separating the 400/150 kV transformers also reduces
the short-circuit current at Vbus busbar in case 3 and 4 compared to their counterparts,
since the impedance of one 400/150 kV transformer is higher than the equivalent for 3
parallel-connected transformers.

A relevant conclusion is to highlight the difference between using 33 or 66 kV as Vbus .

Observing Table 6.1, the specifications of case 4 seem the most beneficial since it allows the
lowest nominal bus current, short-circuit current and number of transformers. However,
the application in industry of Medium Voltage (MV) Transformer Rectifiers led to the
selection of 33kV as voltage level [72].

Case 3 could have been the selected design for the further analysis of the plant, but the
Hydrohub project brought in the possibility of omitting the intermediate 150 kV busbar,
linking directly the 400 kV and 33 kV busbars through 4 state-of-the-art three-winding
transformers [69]. This approach is possible because there are no loads connected directly
to the 150 kV busbar in the station SLD and it is technologically viable to manufacture
these 400 kV/33kV transformers [72]. Besides, this scheme reduces the total number
of transformers, busbars and switchgear systems necessary inside the station. The final
structure design is illustrated in Figure 6.3.

Figure 6.3: Selected SLD for station design.

54
 6. Structure Design for Electrolyzer stations at GW-scale

6.2 Model implementation of different structure designs

The first step to implement the GW-scale structure design is to step-up the 1.9 MW
electrolyzer system (A485-stack) to higher power ratings. As explained in section 4.2,
Nel Hydrogen’s catalogue [44] includes the modules A1000 and A3880, each using 2 and
8 A485-stacks respectively. Since A3880 enables a larger hydrogen production and power
rating, this was the preferred choice.

Figure 6.4: A485 and A3880 modules [44].

Since the connection settings are not defined in the specifications [44], it is possible to
connect the stacks in 4 different schemes: all parallel (1s8p), 2-series in 4-parallel (2s4p),
4-series in 2-parallel (4s2p) and all series (8s1p). Some of these proposals would be imprac-
tical in real application, for example, problems like the self-discharge of parallel-connected
batteries. Assuming all connection schemes are technically possible, the preferred choice
for a 12-PTR would be the 4s2p module as shown in Figure 6.5.

Figure 6.5: A3880 4s2p electrolyzer system.

55
6. Structure Design for Electrolyzer stations at GW-scale

This configuration provides a good performance since arranging 4 stacks in series in each
leg allows to quadruple the DC voltage to 4.2 kV and 2 parallel legs to maintain the DC
current to 2.1 kA per 6-pulse bridge. In other words, the current per thyristor reaches the
maximum allowable of 2.1 kA and the blocking voltage per thyristor increases to 4.2kV,
both values under the thyristor maximum ratings shown in Figure 5.1.

It is relevant to determine how many electrolyzer systems must be placed under each 33 kV
busbar. The total number of electrolyzer systems in the one-GigaWatt plant is calculated
dividing the total station active power capacity Pn,station by the individual demand of each
electrolyzer system Pn,sys . Assuming an A3880 module as individual load (15.2 MW at
80o C), equation (6.1) would dictate 66 electrolyzer systems. Since 64 systems are better
distributed in multiples of 2, this was the implemented number in the substation.

Pn,station
Nsys,total = (6.1)
Pn,sys

The typical nominal current for a 33kV busbar is around 3 kA, which gives a maximum
capacity per busbar of 172 MVA. Assuming a power factor of 0.86, each A3880 module
consumes an apparent power Sn,sys of 17.66 MVA at full load. Introducing this value in
equation (6.2) gives a upper limit of 9 systems per branch and a subsequent minimum of
7 busbars for 33kV.

Sn,branch
Nsys,branch = (6.2)
Sn,sys

The proposed number of transformers between 400kV and 33kV (Figure 6.3) is 4 with
a three-winding configuration, which gives a total of eight 33kV busbars. Following the
same logic as distributing the systems in multiples of 2, the total number of systems per
busbar is 8, which fulfills the minimum number of busbars.

Lastly, in order to reduce the simulated number of elements, the 8 electrolyzer systems
connected to each busbar have been grouped into 2 equivalent modules of 70.6 MVA each
with a power factor of 0.86 as shown in Figure 6.6.

Figure 6.6: Grouping electrolyzer systems into electrolyzer modules.

56
 6. Structure Design for Electrolyzer stations at GW-scale

6.3 Demand Connection Code

Grid code compliance is a critical aspect of connecting an industrial plant to the power
grid at a high voltage level. Electrical codes and standards ensures the safety, reliability
and interoperability of the connection.

In general, grid code requirements are not defined by a single entity, but by multiple
applicable regulations and standards in each country. These may include national codes
dictated by the transmission grid operator (Svenska kraftnät in Sweden), international
standards (e.g. IEC), and local regulations for electrical installations and operations.

Three main documents are considered for grid code compliance in the plant’s design and
operation: the Demand Connection Code (DCC) [75], the IEEE Recommended Practice
and Requirements for Harmonic Control in Electric Power Systems and the IEC assess-
ment for waveform distortion limits.

Voltage regulation and power quality standards are the main aspects to consider when
connecting an industrial facility to the power grid at HV level. These standards ensure a
stable operation, prevent disruptions and mitigate adverse effects on the grid and connec-
ted equipment. These are summarized in 3 main categories: harmonics and distortion,
power factor correction and voltage regulation.

6.3.1 Harmonics and Distortion

Compliance with harmonic standards is specifically relevant for the present study since
the base load of the whole plant is a non-linear load that generates significant current
distortions. This may cause electrical and mechanical stress in connected equipment,
overheating and most importantly a distortion in the PCC voltage waveform, which can
potentially affect other customers located nearby in the grid.

Since in DCC article 20 [75] it is not specified any harmonic limits, it is used the IEEE
Standard 519-2014 [76] and IEC TR 61000-3-6:2008 Standard [77] as reference.

Table 6.2: Current distortion limits for systems rated above 161 kV IEEE Std 519-2014 [76].

Individual harmonic order (odd harmonics) [%]

ISC /Imax 3≤h<11 11≤h<17 17≤h<23 23≤h<35 35≤h<50 TDD
<25 1.0 0.5 0.38 0.15 0.1 1.5
25<50 2.0 1.0 0.75 0.3 0.15 2.5
≥ 50 3.0 1.5 1.15 0.45 0.22 3.75

Table 6.3: Individual harmonic order (odd harmonics) IEC TR 61000-3-6 [77].

Harmonic order Harmonic Voltage [%]

h MV HV-EHV
5 5 2
7 4 2
11 3 1.5
13 2.5 1.5
17 ≤ h ≤ 49 1.9 h − 0.2
17
1.2 17
h

57
6. Structure Design for Electrolyzer stations at GW-scale

The Total Harmonic Distortion (THD) is a measurement that provides insight of a signal
distortion. A low THD indicates a signal close to a pure sinusoidal wave. It is defined as
the ratio of the RMS amplitude of the higher harmonic frequencies to the RMS amplitude
of the fundamental frequency, as shown in equation (6.3).

pP∞
Vn2
T HDv [%] = n=2
∗ 100 (6.3)
V1

In Table 6.2, the Institute of Electrical and Electronics Engineers (IEEE) defines the cur-
rent harmonic limits for users connected to transmission systems with a rated voltage at
PCC greater than 161 kV. However, all tabulated values are expressed as a percentage
of the maximum demand current, Imax , which transforms THD into Total Demand Dis-
tortion (TDD), calculated relative to the full load demand. This distinction is important
because, at low currents, the harmonic components can be of similar magnitude to the
fundamental. Both variables are correlated through equation (6.4), equal only at full load.

I1
T DDi [%] = T HDi [%] ∗ (6.4)
Imax

In Table 6.3, the International Electrotechnical Commission (IEC) specifies voltage distor-
tion limits based on voltage levels: medium voltage (MV) under 132 kV, and high voltage
to extra high voltage (HV-EHV) above 132 kV. It is notable that the IEC imposes stricter
limits for industries connected to HV-EHV busbars than for MV, due to the greater risk
of jeopardizing the supply to a larger number of loads.

6.3.2 Power Factor Correction

Compliance with power factor standards is necessary to ensure an efficient use of elec-
trical energy and avoid penalties imposed by utilities for poor power factor. Article 15
reactive power requirements of the DCC [75] limits the actual reactive power range at 0.9
power factor either for import or export of active power, except in situations where either
technical or financial system benefits are demonstrated by the demand facility owner.

Nonetheless, this general guideline is a soft limit so in this thesis power factor is kept in
unity at PCC. This statement implies that all reactive power produced or consumed inside
of the plant should be compensated. This may be accomplished with passive components
like capacitor banks or with active reactive-power compensators, such as Synchronous
Condenser, Static Var Compensator (SVC) or Static Synchronous Compensator (STAT-
COM).

6.3.3 Voltage Regulation

High voltage industrial connections should include voltage regulation equipment to man-
age fluctuations and maintain a stable voltage supply within specified limits under ex-
traordinary scenarios such as load rejection. The use of Dynamic Voltage Restorers
(DVR), STATCOMs or similar devices are necessary to avoid or dampen voltage dips
and temporary overvoltages.

58
 6. Structure Design for Electrolyzer stations at GW-scale

6.4 Power Plant Electrical Components Design

6.4.1 Grid Characterization
For load flow calculations, the external grid is operated like a Slack Bus (SL) where the
main setpoint is keeping the bus voltage at 1 p.u. Equation (6.5) and (6.6) define the
′′ ′′
grid’s short-circuit capacity Sk as well as the short-circuit current Ik , according to the
settings defined in PowerFactory.

′′ √ ′′
Sk = 3Unom Ik (6.5)
′′ cf actor Unom
I¯k = (6.6)
Z̄1

Rewriting both equations and assuming a R/X short-circuit ratio, the grid’s positive se-
quence impedance Z̄1 can be derived. This impedance represents the Thevenin equivalent
impedance of the external grid and is calculated dividing the nominal voltage Unom of the
′′
grid by its short-circuit capacity Sk . The c-factor coefficient is a safety factor at the user’s
disposal that is multiplied by Unom to account for voltage variations at PCC. Splitting the
impedance into their resistive and reactive components gives equation (6.7) and (6.8).

2
cf actor Unom
X1 = q ′′ (6.7)
R 2 Sk
1 + (X )
R
R1 = X1 (6.8)
X

As this section does not involve unbalanced load flow calculations or unbalanced faults,
negative and zero sequence impedance calculations are beyond the scope of this study.
Consequently, short-circuit current calculations for unbalanced faults are also excluded.

Table 6.4: Grid Characterization.

Parameter description Parameter Value Units

′′
Short-circuit power Sk 10000 MVA
Nominal voltage Unom 400 kV
c-factor coefficient cf actor 1.1 -
Positive sequence R to X short-circuit ratio R/X 0.1 -
′′
Short-circuit current Ik 14.43 kA
Positive sequence resistance R1 1.75 Ω
Positive sequence reactance X1 17.51 Ω

59
6. Structure Design for Electrolyzer stations at GW-scale

6.4.2 Transformers Rating Estimation

As it is shown in Figure 6.3, there are 2 main voltage levels in the station (400 and 33
kV). Additionally, it is necessary to consider the 3.11 kV terminal inside the electrolyzer
module to rate the equivalent 2-winding transformer. In a real-case substation, each
A3880 module would have its own 3-winding transformer with 18 MVA of capacity given
the nature of the 12-PTR topology, but a 2-winding is used to simplify calculations.

The maximum active power Pn,sys demanded by a A3880 system is 15.2 MW with a
power factor of 0.86, assuming operating temperature at 80o C. Since 4 AWE systems
constitute an equivalent module, the Sn,module becomes 70.8 MVA. There are 2 modules
per busbar, therefore the power rating of the transformer between 3.11 kV and 33 kV
terminals becomes 141.6 MVA according to (6.9).

Sn,2T R = Nsys,branch ∗ Sn,sys (6.9)

The total apparent power of the 400/33kV three-winding transformers Sn,3TR doubles the
33/3.11kV transformer rating Sn,2TR , which gives an estimated capacity of 283.2 MVA.
This may be assumed since both medium (MV) and low (LV) voltage windings are sym-
metric.

This is a simplistic method to estimate the transformer power rating disregards two im-
portant factors. The first one is the reactive power consumed by the leakage impedance,
which increases the power rating of the transformer. Secondly, the electrolyzer module
reactive power consumption is a quadratically proportional to the AC bus voltage as it
was shown in Figure 5.19.

These two effects are accounted for in the load flow and ramp-up simulations in a latter
section. These preliminary calculations serve only in the first instance, since additional
components are connected in latter stages. These modify the load flow through the trans-
formers, which makes these calculations only provisional. Both ratings and parameters
are printed in Table 6.5.

Table 6.5: Transformers Rating Sheet.

Parameter TR2 TR3 Units

Power Demand 141.6 283.2 MVA
Rated Power 150 300 MVA
HV side 33 400 kV
MV/LV side 3.11 33/33 kV
Positive sequence leakage impedance 3 14/14 %
Zero sequence leakage impedance 3 11/11 %
Vector Group YNyn0 YNyn0yn0 -

60
 6. Structure Design for Electrolyzer stations at GW-scale

6.4.3 Harmonic Analysis

The 12-PTR line current waveform exhibits a stepped square shape, which makes the
whole electrolyzer system a non-linear load. It is important to eliminate these distortions
before they get transmitted to the grid to avoid any effect on facilities nearby the station
and unnecessary losses. In the absence of any filtering components, the current harmonic
distortion at PCC and 33 kV busbar at maximum loading is depicted in Figure 6.7. The
T HDi is 10.24%, way above the limit of 1.5% set by IEEE Std 519-2014.

Figure 6.7: Current harmonic distortion distribution at 33kV and 400kV busbar without filtering.

Since the rectifier topology is thyristor-based, the resulting harmonic distortion is pre-
dominantly of low frequency. Various methods can be employed to mitigate harmonic
distortions in electrical power systems, including the previously mentioned use of Phase-
Shifting Transformers in Multi-Pulse Converters. However, two primary methods stand
out as particularly effective, as illustrated in Figure 6.8.

• Active filtering. Through dynamic power converters such as DVRs and STATCOMs
[78], the control scheme is designed to produce the measured harmonics in same mag-
nitude but opposite phase. This way the harmonics get cancelled out seen from the
power system side. This system is very effective, but requires high investment costs.

• Passive filtering. Using band-pass tuned filters for specific harmonic orders is a
more economic and widespread solution. Designed to provide low impedance paths
for harmonic currents, shunt filters divert harmonic currents away from the power

61
6. Structure Design for Electrolyzer stations at GW-scale

system. They are robust, but have the risk of overloading if an external distortion
is injected from the grid, surpassing their rating.

(a) Active Filtering. (b) Passive Filtering.

Figure 6.8: Harmonic Filtering [79].

Given the advantages and disadvantages outline above, it was decided to proceed with
passive filters. However, it would be beneficial to analyze the potential implementation
of active filtering in a future project.

The working principle of a passive filter is illustrated in Figure 6.9. If the non-linear load
is modeled as a current source Ih , the filter is designed to have zero impedance at the
frequency of the target harmonic order. Therefore, the filter offers the path of least energy
for the harmonic component of the current If h , as depicted in Figure 6.9b.

(a) Norton equivalent circuit of grid and fil-

ter. (b) Frequency sweep of filter current.

Figure 6.9: Single-tuned filter [80].

To design a band-pass passive filter, it is required the definition of 4 main parameters:

rated reactive power (Qf ), rated voltage (Vf ), resonant frequency (fr ) and quality factor
(q). With these four design parameters, the corresponding Lf , Cf , Rf values can be
calculated using equations (6.10), (6.11) and (6.12).

62
 6. Structure Design for Electrolyzer stations at GW-scale

Vf 2
Xf = X L − XC = (6.10)
Qf

1 1
Lf Cf = 2
= (6.11)
ωr (2πfr )2

ωr Lf 1
q= = (6.12)
R ωr Rf Cf

The selection of the inductance Lf and capacitance Cf is straightforward when imple-

menting the corresponding resonance frequency fr for the selected harmonic order and
the desired reactive power Qf . However, there is no specific rule to decide which quality
factor q to implement, which in turn determines the filter resistance Rf . The quality factor
significantly affects the steady-state losses and smooths the transition from capacitive to
inductive impedance at the resonance frequency.

When studying the harmonic distribution of Figure 6.7, it can be observed that the most
conflicting harmonics are the 11th , 13th , 23rd and 25th . In a first instance, it was considered
to implement a band-pass filter tuned for each harmonic. Opportunely, it resulted to be
enough with 2 band-pass filters, each tuned for the 11th and 23rd respectively.

There are two alternatives to connect the passive filters and mitigate the harmonics. The
first option is to connect them to the 400kV busbar. It represents a cost-effective solution
and requires a low footprint since it agglomerates all filtering equipment in a single bank.
However, this option was quickly discarded since it allows the voltage distortion of each
33kV busbar, polluting the power supply of all converters connected to the terminal.
Besides, the harmonics going through the transformers would cause unnecessary losses,
but also strain on equipment and shortening of its life span [81].

The second option consisted on connecting filters to each 33 kV busbar, which avoids the
intrinsic problem of the previous configuration, but increases the station footprint and
cost. With this station design, the reactive power flow through the transformer is reduced
considerably, which alleviates the voltage drop on its windings. The specifications for
both filters is described in Table 6.6. If a frequency sweep is executed, it is obtained the
filter impedance response shown in Figure 6.10.

Table 6.6: Specifications for 11th and 23rd tuned band-pass filters at 33kV busbar.

Design Parameter Value Parameter Value

Rated Reactive Power 25 MVAr Rf 0.2 Ω
Rated Voltage 33 kV Lf 1.155 mH
Resonant Frequency 550 Hz Cf 72.47 uF
Quality Factor 20
Rated Reactive Power 25 MVAr Rf 0.1 Ω
Rated Voltage 33 kV Lf 0.26 mH
Resonant Frequency 1150 Hz Cf 72.94 uF
Quality Factor 20

63
6. Structure Design for Electrolyzer stations at GW-scale

Figure 6.10: Frequency sweep of filter impedance.

As it can be observed Figure 6.11, the current distortion is handled so all components
are maintained under limits and the T DDc is 1.133% below the limit of 1.5 % stated by
IEEE Std 519 (2014). Similarly, the harmonic distortion distribution of the voltage at
PCC with the selected filtering is shown in Figure 6.12. Once more, it is checked that the
magnitude for each harmonic order is below limits and the T HDv is 0.923% below the
demanded 3% by IEC TR 61000-3-6 (2008).

Figure 6.11: Current harmonic distortion distribution at 33kV and 400kV busbar with filtering.

64
 6. Structure Design for Electrolyzer stations at GW-scale

Figure 6.12: Voltage harmonic distortion distribution at PCC with filtering.

6.4.4 Reactive Power Compensation

As mentioned before, the passive filters at 33 kV busbar have an intrinsic positive effect:
they produce reactive power, which partially compensates the electrolyzer systems reactive
power demand. Therefore, it also decreases the voltage drop on the transformer and avoids
the need of tap changers at the LV side by keeping the voltage within 0.95 and 1.05 p.u.

Nevertheless, an instantaneous reactive power compensation method would not be possible

using only passive filters. Although capacitor banks are scalable and can operate in steps,
these are fixed and present a complex operation for real-time power factor adjustments
[82].

A possible alternative could be using another reactive power compensator (synchronous

condenser, SVC) at the 33 kV busbar in parallel with the filters, which would solve the
power factor adjustment. However, since there are already capacitor banks implemented
at each 33 kV busbar, it would be more economical to increase their power rating instead
of adding new equipment. Besides, filters present advantages against their competitors
like low maintenance costs and low power losses [79].

Considering all theses facts, the proposed solution is a hybrid combination of these passive
filters at 33kV and a STATCOM connected to the 400kV terminal imitating the solution
proposed by J. Solanki, N. Fröhleke and J. Böcker [83] as illustrated in Figure 6.13.

65
6. Structure Design for Electrolyzer stations at GW-scale

Figure 6.13: Hybrid system composed of STATCOM, passive filter, 12-PTR and electrolyzer
[83].

The strategy of this system is to generate the bulk of reactive power demand with the pass-
ive filters, and use the STATCOM to provide the variable part. When the station is not
demanding power, the injection of reactive power by the filters needs to be compensated
by the STATCOM to avoid overvoltage. On the other hand, when the station is at full
load, the STATCOM changes to capacitive mode, therefore merging the contribution of
both devices.

The dilema is how to distribute the capacity between the two devices. According to J.
Solanki et al. [83], the capital cost of a STATCOM is much higher than its equivalent
capacitor bank of the same rating. This leads to an optimization problem of the rating
of the STATCOM to minimize costs and increase the system performance. The proposed
distribution of capacity between the STATCOM (Qst ) and filters (Qf ) suggested in [83] is
as follows:

Qmax + Qmin Q3T R,max + Q3T R,min

Qf = = N3T R (6.13)
2 2
Qmax − Qmin Q3T R,max − Q3T R,min
Qst = = N3T R (6.14)
2 2

where Qmax and Qmin symbolize the maximum and minimum reactive power demanded

66
 6. Structure Design for Electrolyzer stations at GW-scale

by the station. In this case, the maximum and minimum scenarios will correspond to the
station working at 100% and 20% loading respectively. Assuming no filters or STATCOM
present in the SLD, the station load flow is presented in Figure 6.14 according to the
calculations made in section 6.4.2.

Figure 6.14: Preliminary load flow calculation.

Substituting the reactive power flows of Figure 6.14 on equation (6.13) and (6.14) gives
367.83 MVAr and 210.23 MVAr respectively, being Qf the total capacity of all passive
filters in the station. To obtain the individual power rating of each filter it is necessary to
divide Qf by the number of 33kV busbars, resulting in 46 MVAr, which evenly distributed
between the 11th and 23rd tuned filters gives a power rating of 23 MVAr each, below
the filter rating calculated in section 6.4.3. In Table 6.7, the STATCOM and filters
specifications are presented with Qst recalculated for the new Qf :

Table 6.7: Preliminary specifications for STATCOM and filters.

STATCOM 11th filter 23rd filter

Rated power 250 MVAr 25 MVAr 25 MVAr
Line-line voltage 33 kV 33 kV 33 kV
Continuous voltage limit 1.1 / 0.9 p.u. 1.1 / 0.9 p.u. 1.1 / 0.9 p.u.

This configuration stands out for its dynamic advantages and being a more cost-effective
solution. While the STATCOM’s main purpose is to regulate the reactive power com-
pensation, it can be used to fulfill other grid code regulations for the station. In section
6.3, it was stated 3 regulations that the station must comply to be able to connect to the
grid. Harmonics and power factor correction have already been addressed, so only voltage
regulation equipment remains up for discussion.

67
6. Structure Design for Electrolyzer stations at GW-scale

6.4.5 STATCOM Design

The simplest way to simultaneously achieve voltage regulation and power factor correction
is by implementing a control strategy in the STATCOM, which allows for short-term
voltage control at PCC and long-term compensation of reactive power. In some ways,
this behavior resembles that of a synchronous machine [84]. This control strategy is
implemented in PowerFactory by introducing an External Station Controller set up to
have null reactive power flow exchange with the grid in load flow analysis, and a DC &
AC voltage control block diagram for dynamic operation during RMS simulation.

The most suitable configuration for the station’s STATCOM is a multilevel converter
(MMC) with cascaded H-bridge valves [72], based on the current and voltage ratings
depicted in Table 6.7. However, since the purpose of this model is only to demonstrate
the behavior of the STATCOM, the Voltage Source Converter (VSC) is modeled as a
two-level VSC with PWM using Grid Following control.

In Figure 6.15, it is illustrated the STATCOM system SLD and composite frame control
schematic diagram. It receives three main inputs, DC-link voltage measurement Vdc , PCC
AC voltage measurement Vac and the current at VSC terminal iac . The voltage control
block calculates the reference current iref in dq coordinates required to obtain a reference
voltage of 1 p.u. at PCC. For the synchronization of the STATCOM, a Phase-locked Loop
(PLL) estimates the PCC phase angle θ. Both variables, iref and θ, introduced inside the
PowerFactory Converter Block (.ElmVsc) gives the VSC switching pattern.

Figure 6.15: STATCOM Single Line Diagram and Composite Frame.

The purpose of implementing Grid Following (GFL) instead of Grid Forming (GFM) is
due to its straightforward implementation, simplicity in control design, and suitability
for stable grids with well-defined voltage profiles [85]. However, GFL converters may en-
counter instability issues in weak grids and do not respond effectively to deviations in grid
frequency. Due to these problems, it is of interest to implement a GFM control structure
in a future project, particularly where the external grid connected to the PCC presents
low a short-circuit capacity and low inertia due to a high concentration of renewables.

Figure 6.16 presents the STATCOM voltage control diagram, divided into DC and AC
voltage control, each with a PI controller for the respective voltage error. The STATCOM

68
 6. Structure Design for Electrolyzer stations at GW-scale

reference frame is synchronized with the grid via the PLL, allowing direct control of active
power exchange with the grid through the d component of the current id . Similarly,
reactive power is controlled through the q component of the current iq . Since active
power is directly related to the discharge of the DC capacitors, there is a relationship
between the DC voltage error ∆Vdc and id . The same logic applies to the AC voltage and
the reactive power, linking the AC voltage error ∆Vac and iq . All variables shown in the
control block diagram are expressed in per unit, and the control parameters are listed in
Table 6.8.

Figure 6.16: STATCOM Voltage Control Block Diagram.

Table 6.8: STATCOM controller parameter definition.

Parameter description Parameter Value

Active Power Control Gain KP,st 10
Active Power Control Time Constant TP,st 0.002 s
Voltage Control Gain Kv,st 4
Voltage Control Time Constant Tv,st 0.002 s
Droop Droop 0
DC Voltage Control Time Delay Tr,dc 0.001 s
AC Voltage Control Time Delay Tr,ac 0.01 s
Maximum current imax 1 p.u.
Minimum current imin -1 p.u.

Table 6.9: PLL parameter definition.

Parameter description Value

Proportional Gain 5
Integration Gain 1
Upper frequency limit 1.2 p.u.
Lower frequency limit 0.8 p.u.

69
6. Structure Design for Electrolyzer stations at GW-scale

70
 7
Simulation cases

In order to verify the station design and the electrical components ratings described in
section 6.4, it is necessary to carry out several simulations, including both steady-state
and dynamic operation of the plant.

There is a key difference to consider between steady-state and dynamic operation. In

load flow analysis the external gird is modeled as a Slack Bus, an ideal voltage source of
magnitude 1 p.u., whereas in dynamic simulation, it considers the Thevenin equivalent
model of the external grid.

1. Load flow analysis

2. Ramp-up simulation

3. Load Rejection

In load flow calculation, results are presented for two different scenarios: "Service Off"
and "Service On". While the first scenario keeps the station working without any reactive
power compensation devices, the second scenario includes both STATCOM and filters in
operation. This separation allows for a comparative analysis of both performances and
ensures that the station systems were correctly rated during the design phase. Ramp-up
simulation and load rejection are executed only for "Service On" scenario due to non-grid
code compliance in "Service Off" scenario.

7.1 Load flow analysis

In this section, load flow is calculated for the station working at three different load-
ing rates: 20%, 50% and 100%. To apply these ratings, all electrolyzers are uniformly
configured with a specific capacity, thus the notation "20% Loading Service Off" implies
that each electrolyzer within the facility is functioning at 20% of its hydrogen production
capacity, with the STATCOM and filters deactivated. Same logic applies for the rest.

In Table 7.1, external grid load flow results are displayed. The first observation is the
correct performance of the STATCOM and filters, which are capable of compensating
the reactive power throughout the full loading of the plant in steady state. Meanwhile
in "Service Off" scenario, power factor values are identical to those described for A485
stack, which correlates with the fact that the station is a scale-up of that individual stack.
Small differences of active power between both scenarios is caused by the filters losses.

71
7. Simulation cases

Table 7.1: External Grid load flow with 400 kV nominal voltage.

Case Active Power Reactive Power Apparent Power Power

MW MVAr MVA Factor
20% Loading Service Off 162.0 157.3 225.8 0.717
50% Loading Service Off 442.6 350.4 564.5 0.784
100% Loading Service Off 972.1 574.0 1129.0 0.861
20% Loading Service On 163.2 0.0 163.2 1.000
50% Loading Service On 443.7 0.0 443.7 1.000
100% Loading Service On 973.2 0.0 973.2 1.000

In Table 7.2, load flow results for the three-winding transformers are shown. An important
observation is that load flow analysis in "Service Off" scenario agrees with the theoretical
calculations made in Sections 6.4.2 and 6.4.4.

Another noticeable finding from these results is the reversal of reactive power flow in the
transformer for the "20% Loading Service On" case, where the introduction of filters at
33 kV busbars creates an injection of reactive power greater than the one demanded by
the electrolyzers. This causes a net increase of apparent power flow from 18.8% to 24.9%
at 20% loading, but with the desired benefit of reducing the maximum apparent power
flow from 94.1% to 85% at 100% loading.

Table 7.2: Three-winding transformers load flow results at 400 kV terminal.

Case Max. Loading Tot. Active Power Tot. Reactive Power

% HV-Side [MW] HV-Side [MVAr]
20% Loading Service Off 18.8 40.5 39.3
50% Loading Service Off 47.0 110.6 87.6
100% Loading Service Off 94.1 243.0 143.5
20% Loading Service On 24.9 40.8 -62.4
50% Loading Service On 37.0 110.9 -3.0
100% Loading Service On 85.0 243.3 76.5

In Table 7.3 one can notice the proportional relation between filters’ reactive power and
33kV terminal voltage, due to the shunt impedance connection. The load flow is identical
for both 25 MVAr BP filters. The net difference of reactive power between electrolyzer
system consumption and filter generation becomes capacitive below 50% loading, boosting
the voltage at 33 kV, and inductive above 50%, droping the voltage. These results prove
correct the optimization method principle described in section 6.4.4.

Table 7.3: 11th and 23rd harmonic filters.

Case Reactive Power u, Magnitude

MVAr p.u.
20% Loading Service On -27.4 1.047
50% Loading Service On -25.3 1.006
100% Loading Service On -23.0 0.960

72
 7. Simulation cases

Finally, STATCOM load flow results are displayed in Table 7.4. The most noticeable
observation is the overloading of the STATCOM system, which demands a rating increase
to at least 350 MVAr. This underestimation is mainly caused by two factors: the increase
of electrolyzer reactive power consumption due to voltage boosting of the AC terminal
and the dependence of capacitor banks reactive power injection with the 33kV busbar
operating voltage, both factors not accounted for in the theoretical calculations. Lastly,
it is seen how the STATCOM capacitive and inductive mode of operation determines the
STATCOM LV bus voltage, decreasing it at 20% loading when working as a reactor and
increasing it at 100% loading as a capacitor.

Table 7.4: STATCOM reactive power flow and 33kV busbar voltage.

Case Reactive Power Reactive Power u, Magnitude

LV-Side [MVAr] HV-Side [MVAr] LV-Side [p.u.]
20% Loading Service On 231.9 249.7 0.929
50% Loading Service On 10.7 10.7 0.997
100% Loading Service On -332.7 -305.9 1.087

7.2 Ramp-up simulation

In this section, the aim is to study the dynamic operation of the plant, in particular the
power consumption ramp-up sequence from 20% till full load; how the station systems
interact with each other, following the strategy discussed in Section 6.4.4; and how the
loading variation in one side of the station affects the rest of components.

Since the thermal model of the electrolyzers is out of the thesis scope, the simulation
starts with all devices already operating at 80o C. The real start sequence of the plant
would have a preheating stage with the electrolyzers operating at a loading greater than
15% until reaching the desired operating temperature. This criteria is applied by starting
the ramp-up with the station operating in steady state, all electrolyzer systems at 80o C
and a station total loading of 20%.

An important feature of the start sequence is the operation of the STATCOM and the
capacitor banks. It is assumed that the station never operates in a no load scenario: the
electrolyzers with no loading and the capacitors banks connected. The capacitor banks
circuit breakers close only when all electrolyzer modules under a 33 kV busbar reach 20%
loading. Until that moment, the STATCOM takes all responsibility for reactive power
compensation. This procedure avoids an unnecessary over-sizing of the STATCOM, since
in no load scenario, the non compensation of the capacitor bank reactive power injection
by the electrolyzers boosts the 33kV voltage to 1.1 p.u.

The station ramp-up sequence starts at 50 seconds and is done in steps of 50% loading for
each electrolyzer system module under each 33kV busbar every 40 seconds, while this is
alternated with the adjacent module every 20 seconds. This sequence details are described
in Table 7.5.

73
7. Simulation cases

Table 7.5: Ramp-up sequence.

Station loading Loading step Time Electrolyzer module number

27.5% 50% Loading at 50s 0, 4, 8, 12
35% 50% Loading at 60s 2, 6, 10, 14
42.5% 50% Loading at 70s 1, 5, 9, 13
50% 50% Loading at 80s 3, 7, 11, 15
62.5% 100% Loading at 90s 0, 4, 8, 12
75% 100% Loading at 100s 2, 6, 10, 14
87.5% 100% Loading at 110s 1, 5, 9, 13
100% 100% Loading at 120s 3, 7, 11, 15

The major difference with the load flow analysis is the exchange of reactive power with
the grid as illustrated in Figure 7.1. While the active power remains identical to the load
flow, reaching 163.5 MW at 20%, 442.5 MW at 50% and 973.5 MW at 100% loading,
the reactive power presents a progressive ramp-up curve as well. Instead of a complete
cancellation, it reaches a maximum of -137.4 MVAr and a steady state of -56.4 MVAr at
100% loading.

Figure 7.1: Ramp-up power flow in the External Grid.

This behavior is attributed to the STATCOM short-term voltage control, which becomes
evident looking at the voltage graph in Figure 7.2.

Figure 7.2: Ramp-up External Grid excitation voltage and PCC voltage.

74
 7. Simulation cases

As mentioned before, since this dynamic simulation considers the grid’s equivalent Thevenin
impedance, the STATCOM’s role to maintain the terminal voltage at 1 p.u. prompts it
to inject additional reactive power and compensate for the voltage drop.

Nevertheless, as shown in Figure 7.3 the STATCOM reaches the current limit at 110
seconds in the 87.5% loading step. The control action effectively maintains the 400 kV
busbar at 1 p.u. until this point, but drops to a voltage level of 0.985 p.u., which is
still under valid limits [86]. During the steady state, this gets compensated by the grid
increasing its excitation voltage from 0.995 to 1.015 p.u., as observed in the next section.

The STATCOM’s SVC has a power rating of 350 MVAr at nominal voltage. However, the
maximum deliverable reactive power comes determined by the SVC AC voltage limits,
from 0.9 to 1.1 p.u., and the maximum current rating Imax , in this case 6.123 kA. According
to equation (7.1), this gives an SVC power limit of 315 MVAr in inductive mode and 385
MVAr in capacitive mode, which agrees with the maximum reactive power reached at the
end of the ramp-up, 383 MVAr, illustrated in Figure 7.3.

√
Qst = 3VLL cf actor Imax (7.1)

Figure 7.3: Ramp-up STATCOM load flow, current and voltage in p.u.

In Figure 7.4, it is clearly reflected the incorporation of passive filters at the 33 kV busbars.
Same as in load flow analysis, the 400/33kV three-winding transformers present a negative
reactive power flow until the 50% loading step at 80 seconds, where it exhibits a null net

75
7. Simulation cases

exchange, and a positive reactive power flow until 100% loading. Comprehensively, the
active power flow remains consistent with the load flow "Service Off" scenario.

Figure 7.4: Ramp-up power flow in a three-winding 400/33kV transformer.

While Figure 7.4 presents the total power flow at the HV side of the three-winding trans-
formers, Figure 7.5 shows the load flow for the equivalent load under each 33 kV busbar.
The first observation is that the voltage at one 33kV busbar has 8 drops although the load
only increases 4 times through it. This behavior is caused by the three-winding nature of
the transformer, which makes the voltage of both MV windings mutually dependent.

Figure 7.5: Ramp-up power flow in adjacent electrolyzer modules and voltage at 33 kV busbar.

76
 7. Simulation cases

The main consequence of this phenomenon is that after completing one loading step in
one of the MV windings, the reactive power follows a small decrease due to the reasons
explained in Section 5.4.4. Contrarily, the active power remains constant despite the
voltage decrease, except for a transient-bump where the electrolyzer controller adjusts
the firing angle accordingly. This behavior is proven in Figure 7.6.

Figure 7.6: Ramp-up electrolyzer dynamic response.

7.3 Load Rejection

Load rejection in electric power systems is a dynamic event where a significant portion of
the electrical load is suddenly disconnected from the power grid. This situation can be
triggered for several reasons, such as faults in the distribution network, tripping of circuit
breakers and, the present case of study, sudden disconnection of large industrial loads.

Load rejection causes notable repercussions for both the grid and the station. From
the grid side, there is a great drop of active power consumption, which has important
effects in the frequency, whose magnitude is dependent on the plant’s portion of discon-
nection. From the station side, the load-shedding significantly impacts the STATCOM
mode of operation, which needs to adapt to the net reactive power flow changes. The
aim of performing this simulation is to validate the voltage control of the STATCOM, its
performance, and current limits rating.

These simulations are carried out with the plant operating at nominal conditions after a
long period. Therefore, steady state conditions are assumed: the grid has increased its
excitation voltage, from 0.996 to 1.015 p.u., to supply the station at full load. This leads
to a null net exchange of reactive power with the grid, unlike in the ramp-up sequence.

77
7. Simulation cases

7.3.1 Partial Load Rejection

This study case shows the global behavior of the power plant under the load rejection
of one station branch. In other words, when all MV breakers open simultaneously and
affect all components downstream in a 33kV busbar, except for the capacitor banks. This
load shedding represents a spontaneous loss of 121.5 MW and an injection into the grid
of 98.8 MVAr, result of the deprivation of electrolyzer systems power consumption, but
still production from the capacitor banks.

In Figure 7.7, one can notice that post-disturbance there is a slight positive generation
of reactive power in the external grid. In steady state, the boost of excitation voltage
to 1.015 p.u. compensated for the voltage drop in the grid’s Thevenin impedance at
full load without the need of STATCOM reactive power injection. Post-disturbance, the
PCC voltage would be greater than 1 p.u. without reactive power flow, so it requires an
additional voltage drop delivered by the external grid reactive power generation.

Figure 7.7: Power flow in the External Grid under partial load rejection

The STATCOM voltage control brings along the capacity to compensate disturbances
in the PCC terminal, i.e. to mitigate a temporary overvoltage and bring it down to
continuous operation limits (0.90-1.05 p.u.) [86] within few milliseconds. In this case, the
injection of almost 100 MVAr into the grid causes a temporary overvoltage of 1.015 p.u.
as illustrated in Figure 7.8.

Figure 7.8: External Grid excitation voltage and PCC voltage under partial load rejection.

78
 7. Simulation cases

The STATCOM generation in steady state before the disturbance is 305.9 MVAr, same
result as calculated in load flow analysis. To compensate for the generation surplus during
the disturbance, the STATCOM system reduces its generation 110.9 MVAr as displayed in
Figure 7.9. The STATCOM controller mitigates the disturbance within 50 ms manifesting
a second order under-damped behavior.

Figure 7.9: STATCOM load flow, current and voltage in p.u. under partial load rejection.

7.3.2 Full Load Rejection

A more severe case is the load rejection of the whole station. Two options were considered
to carry away this analysis: opening simultaneously all HV breakers in the LV-side of the
three-winding transformers, or opening all MV breakers downstream at 33kV busbars.
Since the first case involves shedding all the electrolyzer systems and filters and in the
second case only the electrolyzers, it was decided to proceed with option 2. This is the
worst-case scenario: in the first one the STATCOM would need to shift from generating
305.9 MVAr down to zero, but in the other, to consume 358.3 MVAr instead.

This scenario produces the same behavior as the partial load rejection but on a larger
scale. As presented in Figure 7.10, the peak of reactive power exchanged with the grid
reaches above 800 MVAr during 6 ms, which causes a temporary overvoltage of 1.1 p.u.
(440 kV). The disturbance is mitigated within 50 ms and the net reactive power flow
exchange with the grid is brought down to 97.5 MVAr.

79
7. Simulation cases

Figure 7.10: Power flow in the External Grid under total load rejection.

Unlike partial rejection, the system is unable to completely compensate for the reactive
power produced by the capacitor banks, leaving a post-disturbance voltage of 1.02 p.u.
(408 kV) illustrated in Figure 7.11. Therefore, some reactive power is injected into the
grid. This situation is only temporary since the capacitor banks are switched off minutes
after the event, as it corresponded to the no load scenario explained in the start-up
sequence strategy.

Figure 7.11: External Grid excitation voltage and PCC voltage under total load rejection.

In Figure 7.12, it is observed an immediate transition from capacitive to inductive mode in

the STATCOM. During the first 3 ms after the disturbance there is an increase of reactive
power production caused by the temporary overvoltage, reaching a maximum STATCOM
production of 366.3 MVAr. The SVC reaches the current limit at this point providing 395
MVAr, above the limit of 385 MVAr calculated in Section 7.2, caused by the STATCOM
LV-side increase to 1.2 p.u. The STATCOM inductive mode of operation reaches 358.3
MVAr post-disturbance, which translates into a SVC production of 320 MVAr, similar to
the simplified limit calculated in Section 7.2.

With this last simulation, the station has been checked for the worst-case scenario, a very
unlikely event given the nature of load rejection of only the electrolyzer systems, but not
the filters. With this last simulation, it has been ensured the correct sizing of the electrical
components inside the station.

80
 7. Simulation cases

Figure 7.12: STATCOM load flow, current and voltage in p.u. under total load rejection.

81
7. Simulation cases

82
 8
Conclusions

8.1 Results from present work

The aim of the master thesis was to design and model a GW-scale electrolyzer station
which resulted in a 972 MW station capable of producing 190 kt of H2 per annum at full
capacity. Benchmarking between AWE and PEM proved PEM superiority in terms of
performance and infrastructure. Nevertheless, AWE technology presented itself as a more
suitable candidate given its significantly lower CAPEX and longer service lifetime, which
led to its selection for model development.

The developed electrolyzer model holds a solid background in the scientific literature,
based on the PHOEBUS demonstration plant, and properly scaled-up accordingly to Nel
Energy Series A megawatt-level electrolyer stack. The plant design accounts for 64 A3880
modules and a reduction to 16 equivalent loads.

Electrolyzer model and converter controller were successfully designed to be easily rep-
licable in future projects using look-up tables and a per unit controller. The controller
design follows IMC methodology accordingly with the electrolyzer process model and val-
idated through ramp-up simulation, giving insight into key differences between expected
and simulated performance.

From different station designs, it was proposed a final candidate for further analysis.
Electrical equipment and components employed in the plant have been rated to endure
the station full loading scenario. This has been confirmed through steady state simulation
(load flow analysis), as well as dynamic simulation (ramp-up and load rejection).

Grid code requirements stated for station design criteria was fulfilled through the imple-
mentation of filters and STATCOM. The harmonic boundaries were established following
international electro-technical standards like IEEE Std 519 (2014) and IEC TR 61000-3-6,
and its limits fulfilled with a 1.133% TDD for current, below 1.5%, and a 0.912% THD
for voltage, below 3%.

An optimization method was proposed for reactive power distribution between the STAT-
COM and capacitor banks, reaching an optimum solution in terms of cost and perform-
ance. The STATCOM implementation enabled flexible reactive power supply and voltage
regulation, with the capacitor banks supplying the bulk of reactive power demand. This
strategy allowed the compliance in both power factor correction and voltage regulation
grid code requirements. An additional advantage of this design was the downsized rating
for the 400/33 kV transformers.

83
8. Conclusions

The ramp-up simulation demonstrated the effect of loading increase in neighborly electro-
lyzer systems due to the terminal voltage drop caused by increased reactive power con-
sumption, as well as, the voltage dependence of both MV windings in the three-winding
transformers. Finally, the introduction of the STATCOM enabled the management of
voltage fluctuations to ensure a stable voltage supply under partial and total load rejec-
tion events and a disturbance handling within a 50 ms time span.

Overall, these steady state and dynamic simulations ensured adequate, safe, reliable and
uninterruptible hydrogen supply at the same time grid code compliance was accomplished.

8.2 Future work

This thesis provides a solid base to evaluate large scale hydrogen plants. Nonetheless,
further work is required to give a more complete scope, electrolyzer model and, evaluate
the economic implications of electrical equipment selection. Further work can be classified
in the next items:

Electrolyzer modelling

To commence, first idea in line would be to improve the electrolyzer model by implement-
ing real data provided by the electrolyzer’s manufacturer, instead of making assumptions
about their catalogue products. Along this line, upgrading the model by integrating
the thermal behavior and analyzing the evolution of the operating temperature during
cold-start could proof interesting.

PowerFactory modelling

Analyzing the 12 PTR performing electromagnetic transient (EMT) simulation using

PowerFactory or PSCAD softwares would prove beneficial for model verification. This
study would be essential for the analysis on how the grid’s Thevenin reactance and trans-
formers leakage inductance affect the commutation angle in the 12-PTR as well as meas-
uring the DC current ripple.

Station components modelling

′′
Furthermore, a correct estimation of the short-circuit capacity Sk of the grid as well as
the leakage inductance of the transformers is key to have correct dynamic simulations and
short circuit analysis.

A valuable addition for plant design would be the introduction of magnetizing inductance
inside transformers since this would have an added-value for an energization study of
the station, station load flow calculation at low loading and harmonic analysis of the
magnetizing current 3rd harmonic introduction [87]. The implementation of Phase Shifting
Transformers [88] between 400kV and 33kV is a useful low-order harmonic cancellation
technique left unexplored.

Protection set-up

A noticeable study of the plant design left untouched is the protection coordination and
switchgear selection, indispensable aspect for the safety and long-term lifespan of the sta-
tion. Another interesting add to the station plant would be the incorporation of surge

84
 8. Conclusions

arresters at 33kV busbars to protect the transformers and capacitor banks against tem-
porary overvoltages.

Economic evaluation

Ancillary services like frequency reserve and interruptible service as demand management
tool for the TSO could add system security to the grid and economical benefits for the
hydrogen plant [89].

Evidently, an economical evaluation would have to be done of STATCOM along with

cost considerations for passive filters per MVAr to determine the optimal distribution of
reactive power in economic terms, while fulfilling the technical feasibility and complying
with overall requirements.

Medium voltage DC distribution grid

Finally, a plant model connected to a MVDC grid would also add valuable information,
as well as implementing PEM electrolyzers to harness wind curtailments and intermittent
power flows due to renewable power supply. This would be followed by a cross comparison
of HVAC and MVDC plant structure, their advantages respect to number of components
and an anticipated total investment cost evaluation for further analysis.

Grid code compliance

Lastly the station should be capable of operating within a PCC voltage range. Swedish
Energy Markets Inspectorate’s statute book [86] stipulates that consumption units shall be
capable of continuous operation within the voltage range 0.90-1.05 p.u. and operation for
at least 60 minutes in the voltage range 1.05-1.10 p.u. which could imply the introduction
of tap-changers in the 400/33kV transformers.

85
8. Conclusions

86
 Bibliography

[1] ‘Climate Change ‘Biggest Threat Modern Humans Have Ever Faced’, World-Renowned
Naturalist Tells Security Council, Calls for Greater Global Cooperation,’ United Na-
tions, Feb. 2021. [Online]. Available: https://press.un.org/en/2021/sc14445.
doc.htm (visited on 27/02/2023).
[2] ‘United Nations Conference on the human environment, Stockholm 1972,’ United
Nations, [Online]. Available: https://www.un.org/en/conferences/environment/
stockholm1972 (visited on 27/02/2023).
[3] Britannica, The Editors of Encyclopaedia, ‘Kyoto protocol,’ Encyclopedia Britan-
nica, Sep. 2022. [Online]. Available: https://www.britannica.com/event/Kyoto-
Protocol (visited on 27/02/2023).
[4] G. Rannard, ‘ExxonMobil: Oil giant predicted climate change in 1970s - scientists,’
BBC News, Jan. 2023. [Online]. Available: https://www.bbc.com/news/science-
environment-64241994 (visited on 27/02/2023).
[5] G. Supran and N. Oreskes, ‘Assessing ExxonMobil’s Climate Change Communica-
tions (1977–2014),’ Environmental Research Letters, vol. 12, no. 8, p. 084 019, 2017.
doi: 10.1088/1748-9326/aa815f.
[6] ‘The Paris Agreement,’ United Nations, [Online]. Available: https://www.un.org/
en/climatechange/paris-agreement (visited on 27/02/2023).
[7] M. McGrath, ‘Climate change: US formally withdraws from Paris Agreement,’ BBC
News, Nov. 2020. [Online]. Available: https : / / www . bbc . com / news / science -
environment-54797743 (visited on 27/02/2023).
[8] ‘The United States officially rejoins the Paris Agreement - United States Department
of State,’ U.S. Department of State, Feb. 2021. [Online]. Available: https://www.
state.gov/the-united-states-officially-rejoins-the-paris-agreement/
(visited on 27/02/2023).
[9] ‘COVID-19 caused only a temporary reduction in carbon emissions – UN report,’
UN Environment, Sep. 2021. [Online]. Available: https://www.unep.org/news-
and-stories/press-release/covid-19-caused-only-temporary-reduction-
carbon-emissions-un-report (visited on 27/02/2023).
[10] J. Tollefson, ‘What the war in Ukraine means for energy, climate and food,’ Nature,
vol. 604, no. 7905, pp. 232–233, 2022. doi: 10.1038/d41586-022-00969-9.
[11] K. Ng, ‘China’s population falls for first time since 1961,’ BBC News, Jan. 2023.
[Online]. Available: https://www.bbc.com/news/world-asia-china-64300190
(visited on 27/02/2023).

87
Bibliography

[12] L. Silver, C. Huang and L. Clancy, ‘Key facts as India surpasses China as the
world’s most populous country,’ Pew Research Center, Feb. 2023. [Online]. Avail-
able: https://www.pewresearch.org/fact-tank/2023/02/09/key-facts-as-
india-surpasses-china-as-the-worlds-most-populous-country/ (visited on
27/02/2023).
[13] IEA, ‘World Energy Outlook 2022,’ IEA, Paris, Tech. Rep. [Online]. Available:
https://www.iea.org/reports/world-energy-outlook-2022.
[14] IEA, ‘Global Energy-Related CO2 Emissions by Sector,’ IEA, Paris, Tech. Rep.
[Online]. Available: https : / / www . iea . org / data - and - statistics / charts /
global-energy-related-co2-emissions-by-sector.
[15] IEA, ‘Direct CO2 emissions from industry in the Net Zero Scenario, 2000-2030,’
IEA, Paris, Tech. Rep. [Online]. Available: https : / / www . iea . org / data - and -
statistics / charts / direct - co2 - emissions - from - industry - in - the - net -
zero-scenario-2000-2030.
[16] IEA, ‘Global hydrogen demand by sector in the Net Zero Scenario, 2019-2030,’
IEA, Paris, Tech. Rep. [Online]. Available: https : / / www . iea . org / data - and -
statistics/charts/global-hydrogen-demand-by-sector-in-the-net-zero-
scenario-2019-2030.
[17] IEA, ‘Global Hydrogen Review 2022,’ IEA, Paris, Tech. Rep. [Online]. Available:
https://www.iea.org/reports/global-hydrogen-review-2022.
[18] T. DiChristopher, ‘Experts explain why green hydrogen costs have fallen and will
keep falling,’ S&P Global Homepage, Mar. 2021. [Online]. Available: https : / /
www . spglobal . com / marketintelligence / en / news - insights / latest - news -
headlines/experts-explain-why-green-hydrogen-costs-have-fallen-and-
will-keep-falling-63037203.
[19] E. Commission, ‘A hydrogen strategy for a climate-neutral Europe,’ European Com-
mission, Brussels, Tech. Rep., Jul. 2020. [Online]. Available: https://eur- lex.
europa.eu/legal-content/EN/TXT/?uri=CELEX%5C%3A52020DC0301.
[20] J. Favas, E. Monteiro and A. Rouboa, ‘Hydrogen production using plasma gasi-
fication with steam injection,’ International Journal of Hydrogen Energy, vol. 42,
no. 16, pp. 10 997–11 005, 2017. doi: 10.1016/j.ijhydene.2017.03.109.
[21] Nuclear Energy Agency, ‘High-temperature Gas-cooled Reactors and Industrial
Heat Applications,’ OECD Publishing, Paris, Tech. Rep., 2022. [Online]. Available:
https://www.oecd-nea.org/jcms/pl_20497/high-temperature-gas-cooled-
reactors.
[22] IRENA, ‘Global hydrogen trade to meet the 1.5°C climate goal: Part I – Trade out-
look for 2050 and way forward,’ International Renewable Energy Agency, Abu Dh-
abi, Tech. Rep., 2022. [Online]. Available: https://prod-cd.irena.org/-/media/
Files/IRENA/Agency/Publication/2022/Jul/IRENA_Global_hydrogen_trade_
part_1_2022_.pdf?rev=f70cfbdcf3d34b40bc256383f54dbe73.
[23] H. Central, ‘Green Hydrogen: Technology Breakthroughs Mean Iridium Shortage
and High Prices will Ease in 2023 – Rystad Energy,’ Feb. 2023. [Online]. Available:
https://hydrogen-central.com/green-hydrogen-technology-breakthroughs-
mean - iridium - shortage - and - high - prices - will - ease - in - 2023 - rystad -
energy/.

88
 Bibliography

[24] IEEFA, ‘Sinopec building world’s largest green hydrogen plant in China,’ Balkan
Green Energy News, Aug. 2022. [Online]. Available: https://ieefa.org/articles/
sinopec-building-worlds-largest-green-hydrogen-plant-china.
[25] S. Matalucci, ‘The Hydrogen Stream: World’s largest electrolyzer to be deployed
in Norway,’ Pv Magazine International, Sep. 2022. [Online]. Available: https://
www.pv-magazine.com/2022/09/13/the-hydrogen-stream-worlds-largest-
electrolyzer-to-be-deployed-in-norway/.
[26] M. Minutillo, A. Perna, A. Forcina, S. Di Micco and E. Jannelli, ‘Analyzing the
levelized cost of hydrogen in refueling stations with on-site hydrogen production via
water electrolysis in the Italian scenario,’ International Journal of Hydrogen Energy,
vol. 46, no. 26, pp. 13 667–13 677, 2021. doi: 10.1016/j.ijhydene.2020.11.110.
[27] W. E. Forum, ‘How many jobs could the clean energy transition create?’ World
Economic Forum, [Online]. Available: https://www.weforum.org/agenda/2022/
03/the-clean-energy-employment-shift-by-2030/.
[28] Stiftelsen for industriell og teknisk forskning (SINTEF), ‘Norwegian Energy Road
Map 2050,’ SINTEF, [Online]. Available: https://www.sintef.no/en/projects/
2016/energymap/.
[29] K. Espegren, S. Damman, P. Pisciella, I. Graabak and A. Tomasgard, ‘The role
of hydrogen in the transition from a petroleum economy to a low-carbon society,’
International Journal of Hydrogen Energy, vol. 46, no. 45, pp. 23 125–23 138, 2021.
doi: 10.1016/j.ijhydene.2021.04.143.
[30] ‘Water,’ United Nations, [Online]. Available: https://www.un.org/en/global-
issues/water.
[31] ‘Water Resource Considerations for the Hydrogen Economy,’ JD Supra, Dec. 2020.
[Online]. Available: https : / / www . jdsupra . com / legalnews / water - resource -
considerations-for-the-84603/.
[32] Ministry of Energy, Government of Chile, ‘National Green Hydrogen Strategy,’
Tech. Rep., Nov. 2020. [Online]. Available: https : / / energia . gob . cl / sites /
default/files/national_green_hydrogen_strategy_-_chile.pdf.
[33] E. Phiddian, ‘Taking the precious metals out of hydrogen electrolysis,’ Cosmos, Feb.
2022. [Online]. Available: https://cosmosmagazine.com/science/chemistry/
hydrogen-electrolysis-precious-metals-catalyst/.
[34] A. Li et al., ‘Enhancing the stability of cobalt spinel oxide towards sustainable
oxygen evolution in acid,’ Nature Catalysis, vol. 5, no. 2, pp. 109–118, Feb. 2022.
doi: 10.1038/s41929-021-00732-9.
[35] U.S. Department of Energy, ‘Safe use of hydrogen,’ [Online]. Available: https :
//www.energy.gov/eere/fuelcells/safe-use-hydrogen.
[36] ‘Hydrogen incident examples,’ Hydrogen Tools, Mar. 2020. [Online]. Available: https:
//h2tools.org/lessons/hydrogen-incident-examples.
[37] S. S. Kumar and H. Lim, ‘An overview of water electrolysis technologies for green
hydrogen production,’ Energy reports, vol. 8, pp. 13 793–13 813, 2022. doi: https:
//doi.org/10.1016/j.egyr.2022.10.127.
[38] IRENA, ‘Green Hydrogen Cost Reduction: Scaling up Electrolysers to Meet the
1.5ºC Climate Goal,’ Abu Dhabi: International Renewable Energy Agency, 2020.

89
Bibliography

[39] O. Posdziech, ‘Production of renewable hydrogen and syngas via high-temperature

electrolysis,’ Sunfire GmbH, Mar. 2021. [Online]. Available: https://multiplhy-
project . eu / Documents / Workshop % 20on % 20Advanced % 20PtG % 20and % 20PtL %
20Technologies%20High-Temperature%20Electrolysis_Posdziech.pdf.
[40] T. Smolinka, H. Bergmann, J. Garche and M. Kusnezoff, ‘Chapter 4 - the his-
tory of water electrolysis from its beginnings to the present,’ in Electrochemical
power sources: fundamentals, systems, and applications, Elsevier, 2022, pp. 83–
164. doi: https : / / doi . org / 10 . 1016 / B978 - 0 - 12 - 819424 - 9 . 00010 - 0. [On-
line]. Available: https : / / www . sciencedirect . com / science / article / pii /
B9780128194249000100.
[41] R. Gilliam, J. Graydon, D. Kirk and S. Thorpe, ‘A review of specific conductivit-
ies of potassium hydroxide solutions for various concentrations and temperatures,’
International Journal of Hydrogen Energy, vol. 32, no. 3, pp. 359–364, 2007.
[42] M. Schalenbach, G. Tjarks, M. Carmo, W. Lueke, M. Mueller and D. Stolten, ‘Acidic
or alkaline? towards a new perspective on the efficiency of water electrolysis,’ Journal
of The Electrochemical Society, vol. 163, no. 11, F3197–F3208, 2016. doi: 10.1149/
2.0271611jes.
[43] O. Ulleberg, ‘Modeling of advanced alkaline electrolyzers: A system simulation ap-
proach,’ International Journal of Hydrogen Energy, vol. 28, no. 1, pp. 21–33, 2003.
doi: 10.1016/s0360-3199(02)00033-2.
[44] ‘Nel hydrogen electrolyzer,’ Nel Hydrogen, [Online]. Available: https://nelhydrogen.
com/water-electrolysers-hydrogen-generators/ (visited on 04/04/2023).
[45] ‘Nafion™ Sulfonic Membranes,’ [Online]. Available: https://www.nafion.com/en/
products/sulfonic-membranes (visited on 15/01/2024).
[46] O. Schmidt, A. Gambhir, I. Staffell, A. Hawkes, J. Nelson and S. Few, ‘Future cost
and performance of water electrolysis: An expert elicitation study,’ International
journal of hydrogen energy, vol. 42, no. 52, pp. 30 470–30 492, 2017. doi: https:
//doi.org/10.1016/j.ijhydene.2017.10.045.
[47] ‘Metal Spot Price Charts Iridium Price (USD / Gram) Chart for the Last Year,’
Daily Metal Prices, [Online]. Available: https : / / www . dailymetalprice . com /
metalpricecharts.php?c=ir&u=oz&d=10 (visited on 16/08/2023).
[48] ‘Iberdrola commissions the largest green hydrogen plant for industrial use in Europe,’
Iberdrola, [Online]. Available: https : / / www . iberdrola . com / about - us / what -
we - do / green - hydrogen / puertollano - green - hydrogen - plant (visited on
18/08/2023).
[49] A. Hodges et al., ‘A high-performance capillary-fed electrolysis cell promises more
cost-competitive renewable hydrogen,’ Nature communications, vol. 13, no. 1, p. 11,
Mar. 2022.
[50] ‘Breakthrough opens door to low-cost green hydrogen,’ University of Wollongong
(UOW), Mar. 2022. [Online]. Available: https://www.uow.edu.au/media/2022/
breakthrough-opens-door-to-low-cost-green-hydrogen.php.
[51] I. Dincer and A. A. AlZahrani, ‘4.25 Electrolyzers,’ in Comprehensive Energy Sys-
tems, I. Dincer, Ed., Oxford: Elsevier, 2018, pp. 985–1025. doi: https : / / doi .
org / 10 . 1016 / B978 - 0 - 12 - 809597 - 3 . 00442 - 9. [Online]. Available: https :
//www.sciencedirect.com/science/article/pii/B9780128095973004429.

90
 Bibliography

[52] M. Mori, T. Mržljak, B. Drobnič and M. Sekavčnik, ‘Integral Characteristics of

Hydrogen Production in Alkaline Electrolysers,’ Strojniški vestnik-Journal of Mech-
anical Engineering, vol. 59, no. 10, pp. 585–594, 2013.
[53] M. Hammoudi, C. Henao, K. Agbossou, Y. Dubé and M. Doumbia, ‘New multi-
physics approach for modelling and design of alkaline electrolyzers,’ International
Journal of Hydrogen Energy, vol. 37, no. 19, pp. 13 895–13 913, 2012.
[54] T. Adibi, A. Sojoudi and S. C. Saha, ‘Modeling of thermal performance of a com-
mercial alkaline electrolyzer supplied with various electrical currents,’ International
Journal of Thermofluids, vol. 13, pp. 100–126, 2022.
[55] ‘Thermophysical properties of fluid systems,’ NIST Standard Reference Database,
[Online]. Available: http://webbook.nist.gov/chemistry/fluid/ (visited on
27/03/2023).
[56] T. Dervišić and C. Engström, ‘Electric infrastructure for electrolyser systems. Design
proposals for AC and DC distribution systems,’ M.S. thesis, Chalmers University
of Technology, Göteborg, Sweden, 2021. [Online]. Available: https://hdl.handle.
net/20.500.12380/302627.
[57] J. Kotowicz, M. Jurczyk, D. Węcel and W. Ogulewicz, ‘Analysis of hydrogen pro-
duction in alkaline electrolyzers,’ Journal of Power Technologies, vol. 96, no. 3,
pp. 149–156, 2016. (visited on 12/05/2023).
[58] ‘Flow meters for hydrogen applications,’ Bronkhorst, [Online]. Available: https :
//www.bronkhorst.com/int/markets/renewable-energy/hydrogen/ (visited on
12/05/2023).
[59] B. Bensmann, R. Hanke-Rauschenbach and K. Sundmacher, ‘In-situ measurement
of hydrogen crossover in polymer electrolyte membrane water electrolysis,’ Interna-
tional journal of hydrogen energy, vol. 39, no. 1, pp. 49–53, 2014.
[60] J. Divisek, J. Mergel and H. Schmitz, ‘Advanced water electrolysis and catalyst
stability under discontinuous operation,’ International journal of hydrogen energy,
vol. 15, no. 2, pp. 105–114, 1990.
[61] W. Hug, J. Divisek, J. Mergel, W. Seeger and H. Steeb, ‘Highly efficient advanced
alkaline electrolyzer for solar operation,’ International journal of hydrogen energy,
vol. 17, no. 9, pp. 699–705, 1992.
[62] I. D. de Cerio Mendaza, B. Bak-Jensen and Z. Chen, ‘Alkaline electrolyzer and V2G
system DIgSILENT models for demand response analysis in future distribution net-
works,’ 2013 IEEE Grenoble Conference, 2013. doi: 10.1109/ptc.2013.6652429.
[63] P. Ghosh, B. Emonts, H. Janßen, J. Mergel and D. Stolten, ‘Ten years of operational
experience with a hydrogen-based renewable energy supply system,’ Solar Energy,
vol. 75, no. 6, pp. 469–478, 2003.
[64] ‘Electrolyzer market size, share and industry analysis, by type (traditional alkaline
electrolyzer, pem electrolyzer), by application (power plants, steel plant, electronics
and photovoltaics, industrial gases, energy storage or fueling for fcevs, power to gas,
others), and regional forecast, 2024-2032,’ Fortune Business Insights, 2024. [Online].
Available: https://www.fortunebusinessinsights.com/electrolyzer-market-
103919.

91
Bibliography

[65] M. K. Gustafson, ‘A computational approach to simulating the performance of a

24-hour solar-fuel cell-hydrogen electric power plant,’ M.S. thesis, Wright State Uni-
versity, May 2013.
[66] F. Gonzalez-Longatt and J. L. Rueda Torres, Modelling and simulation of power
electronic converter dominated power systems in PowerFactory. Springer Nature,
2021.
[67] M. Beza, ENM061 - Power Electronic Converters, Lecture Notes, Chalmers Univer-
sity of Technology, Dept. Elect. Eng., 2022.
[68] B. Yodwong, D. Guilbert, M. Phattanasak, W. Kaewmanee, M. Hinaje and G.
Vitale, ‘AC-DC Converters for Electrolyzer Applications: State of the Art and
Future Challenges,’ Electronics 2020, vol. 9, no. 6, May 2020. doi: 10 . 3390 /
electronics9060912. [Online]. Available: https://www.mdpi.com/2079-9292/9/
6/912.
[69] ‘A One-GigaWatt Green-Hydrogen Plant Advanced Design and Total Installed-
Capital Costs,’ Hydrohub Innovation Program, [Online]. Available: https://ispt.
eu/projects/hydrohub-gigawatt/ (visited on 04/10/2023).
[70] M. Chen, S.-F. Chou, F. Blaabjerg and P. Davari, ‘Overview of Power Electronic
Converter Topologies Enabling Large-Scale Hydrogen Production via Water Elec-
trolysis,’ Applied Sciences, vol. 12, p. 1906, Feb. 2022. doi: 10.3390/app12041906.
[71] N. Mohan, T. M. Undeland and W. P. Robbins, Power electronics: converters,
applications and design. John Wiley & Sons, 2003.
[72] Hitachi Energy, Private communication, Hitachi Energy, 2023.
[73] S. Lundberg, ENM076 - Electric Drive Systems, Lecture Notes, Chalmers University
of Technology, Dept. Elect. Eng., 2022.
[74] ‘IEEE Recommended Practice for Excitation System Models for Power System Sta-
bility Studies,’ IEEE Std 421.5-2016 (Revision of IEEE Std 421.5-2005), pp. 1–207,
2016. doi: 10.1109/IEEESTD.2016.7553421.
[75] ‘Commission Regulation (EU) 2016/1388 of 17 August 2016 Establishing a Network
Code on Demand Connection,’ [Online]. Available: https://eur-lex.europa.eu/
eli/reg/2016/1388/oj (visited on 05/10/2023).
[76] ‘IEEE Recommended Practice and Requirements for Harmonic Control in Electric
Power Systems,’ IEEE Std 519-2014 (Revision of IEEE Std 519-1992), pp. 1–29,
Jun. 2014. doi: 10.1109/IEEESTD.2014.6826459.
[77] ‘Electromagnetic compatibility (EMC) - Part 3-6: Limits - Assessment of emission
limits for the connection of distorting installations to MV, HV and EHV power
systems,’ International Electrotechnical Commission, Tech. Rep. IEC TR 61000-3-
6:2008, Feb. 2008.
[78] D. J. Meisner, B. Niemann, M. Shevchenko, E. Fombang, I. Khosravi and H. von
Geymüller, ‘STATCOM with Active Filter Using STATCOM as Active Filter, Im-
proving Power Quality and reducing Harmonics,’ pp. 1–5, 2020. doi: 10 . 1109 /
TD39804.2020.9299900.
[79] M. Bongiorno, ENM100 - Power Electronic Solutions for Power Systems, Lecture
Notes, Chalmers University of Technology, Dept. Elect. Eng., 2022.

92
 Bibliography

[80] L. Morán, J. Dixon and M. Torres, ‘41 - Active Power Filters,’ in Power Electron-
ics Handbook (Fourth Edition), M. H. Rashid, Ed., Fourth Edition, Butterworth-
Heinemann, 2018, pp. 1341–1379. doi: https : / / doi . org / 10 . 1016 / B978 - 0 -
12-811407-0.00046-5. [Online]. Available: https://www.sciencedirect.com/
science/article/pii/B9780128114070000465.
[81] ‘Harmonic filters charm,’ Hitachi Energy, [Online]. Available: https://www.hitachienergy.
com / products - and - solutions / capacitors - and - filters / high - voltage -
capacitors-and-filters/harmonic-filters (visited on 05/06/2023).
[82] ‘Metal enclosed capacitor banks ABBACUS,’ Hitachi Energy, [Online]. Available:
https://www.hitachienergy.com/products-and-solutions/capacitors-and-
filters/medium-voltage-capacitors-and-filters/capacitor-banks/metal-
enclosed-capacitor-banks-abbacus (visited on 20/08/2023).
[83] J. Solanki, N. Fröhleke and J. Böcker, ‘Implementation of Hybrid Filter for 12-
Pulse Thyristor Rectifier Supplying High-Current Variable-Voltage DC Load,’ IEEE
Transactions on Industrial Electronics, vol. 62, no. 8, pp. 4691–4701, 2015. doi:
10.1109/TIE.2015.2393833.
[84] V. Azbe and R. Mihalic, ‘Statcom control strategies in energy-function-based meth-
ods for the globally optimal control of renewable sources during transients,’ Interna-
tional Journal of Electrical Power & Energy Systems, vol. 141, pp. 108–145, 2022.
doi: https://doi.org/10.1016/j.ijepes.2022.108145. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S0142061522001843.
[85] R. Heydari et al., ‘Grid-Forming Control for STATCOMs – a Robust Solution for
Networks with a High Share of Inverter-Based Resources,’ CIGRE, pp. 1–11, 2022.
[Online]. Available: https : / / www . e - cigre . org / publications / detail / c4 -
10822-2022-grid-forming-control-for-statcoms-a-robust-solution-for-
networks-with-a-high-share-of-inverter-based-resources.html.
[86] S. Kraftnät, ‘Energimarknadsinspektionens föreskrifter om fastställande av gener-
ellt tillämpliga krav för anslutning av förbrukare,’ [Online]. Available: https : / /
www.ei.se/om- oss/publikationer/publikationer/foreskrifter- el/2019/
foreskrift-eifs-20196 (visited on 05/10/2023).
[87] L. Mari, ‘Magnetizing and Exciting Currents Waveshapes in Transformers,’ EEP-
ower, Jul. 2020. [Online]. Available: https://eepower.com/technical-articles/
magnetizing- and- exciting- currents- waveshapes- in- transformers/# (vis-
ited on 30/08/2023).
[88] ‘Phase-shifting transformers (PST),’ [Online]. Available: https://www.hitachienergy.
com/products- and- solutions/transformers/power- transformers/system-
intertie-transformers/phase-shifting-transformers (visited on 30/08/2023).
[89] ‘Interruptibility Service,’ Red Eléctrica, [Online]. Available: https://www.ree.es/
en/activities/operation-of-the-electricity-system/interruptibility-
service#:~:text=Interruptibility%20Service%20Interruptibility%20service%
20is%20a%20demand%20management,%28system%20security%29%20and%20economic%
20criteria%20%28reducing%20system%20costs%29. (visited on 30/08/2023).

93
DEPARTMENT OF ELECTRICAL ENGINEERING
CHALMERS UNIVERSITY OF TECHNOLOGY
Gothenburg, Sweden
www.chalmers.se