No - Ntnu Inspera 79786156 64576396

Mohammad Heidari
Control Models for Providing Virtual

Inertia to the
Master’s thesis
Electrical Grid of Offshore Platforms

With Large
Contribution of Offshore Wind
Master’s thesis in Renewable Energies in the Marine Environment

Supervisor: Prof. Elisabetta Tedeschi
NTNU
Norwegian University of Science and Technology
Faculty of Information Technology and Electrical Engineering
Department of Electric Power Engineering
Co-supervisor: Salvatore D'Arco

June 2021
Mohammad Heidari
Control Models for Providing Virtual

Inertia to the
Electrical Grid of Offshore Platforms
With Large
Master’s thesis in Renewable Energies in the Marine Environment

Supervisor: Prof. Elisabetta Tedeschi
Co-supervisor: Salvatore D'Arco
June 2021
Norwegian University of Science and Technology

Faculty of Information Technology and Electrical Engineering
Department of Electric Power Engineering
Control Models for Providing Virtual Inertia to the
Electrical Grid of Offshore Platforms With Large
Mohammad Heidari
June 2021
Summary
Norway as one of the leading countries in the oil and gas and (floating) off-
shore wind energy sectors, plans to reduce its carbon emission according to the
European commission strategy to become carbon-neutral by 2050. One of the
considerable contributors to the carbon emission in Norway are oil and gas plat-
forms. Supplying these platforms thorough offshore wind can accelerate realizing
carbon-neutrality.
A well-built grid can withstand a contingency event and have an acceptable

rate of change of frequency due to high inertial characteristics. The reduction of
the grid inertia is one of the main issues with the paced integration of renewable
energy sources into the electricity grid and replacing the conventional generators.
To overcome this issue, multiple mathematical methods have been developed to
ensure grid stability. These methods are based on an additional energy source to
help the system stability throughout the event of a contingency.
In this project, a few of the methods to provide virtual inertia for renewable
energy sources are reviewed and compared against each other. Based on the ideas
inspired by these methods, an algorithm is proposed to assist the inertial response
of the grid of oil and gas platforms.
The proposed algorithm is implemented on a voltage source inverter that ex-

tracts energy from energy storage and injects it into the grid of the platform along-
side gas and wind turbines. The method is validated by a step-change in the load
of the platform and the interactions of the grid components are presented. Com-
pared to a scenario in which there is no inertial support provided, the proposed
method proves operational in reducing the rate of change of frequency and there-
fore, stabilizing the grid during the contingency.
Key words: Electrification of oil and gas platforms, Offshore wind turbine,
Virtual inertia, vector control of converter, Back to back converter control, DFIG,
gas turbine, Energy storage
iii
Acknowledgements
I have been studying abroad as part of the Erasmus Mundus joint master pro-
gram REM (Renewable energies in the marine environment) and I would like to
extend my gratitude towards the REM joint board of coordinators from the uni-
versities of Strathclyde, UPV/EHU, and NTNU. This work could not have been
possible without the support and guidance of my coordinators, lecturers, and su-
pervisors. I would also like to thank Ph. D. candidate Daniel Mota for his continu-
ous help for the past six months. In the end, I want to thank my friends and family
for their emotional support during the two years of my studies abroad.
Mohammad Heidari
June 2021
v
Contents
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Objectives and Research Questions . . . . . . . . . . . . . . . . . . . . 2
1.4 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Oil and Gas Platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Source of Inertia in the Power Systems . . . . . . . . . . . . . . . . . . 6
2.3.1 SG Model and Operation Principles . . . . . . . . . . . . . . . 6
2.3.2 Traditional Power Plants . . . . . . . . . . . . . . . . . . . . . . 8
2.3.3 DFIG model and operation principles . . . . . . . . . . . . . . 9
2.3.4 Inertia for Wind Turbines . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Converter Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.1 Types of grid connection for wind turbines . . . . . . . . . . . 15
2.4.2 Voltage source vs current source converters . . . . . . . . . . 17
2.4.3 Grid forming vs grid following converters . . . . . . . . . . . . 18
2.4.4 PLL units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.5 Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.6 Vector transformation . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.7 Modulation techniques . . . . . . . . . . . . . . . . . . . . . . . 22
3 Existing Literature for Methods to Emulate Virtual Inertia . . . . . . . 31
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 SG Model Based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2.1 Synchronverters . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2.2 Institute of Electrical Power Engineering (IEPE) Lab Topology 35
3.2.3 Kawasaki Heavy Industries (KHI) Lab’s Topology . . . . . . . 39
vii
viii M. Heidari: Control Models for Providing Virtual Inertia
3.3 Swing Equation Based Methods . . . . . . . . . . . . . . . . . . . . . . 41

3.3.1 Ise Lab’s Topology . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.2 Synchronous Power Controller (SPC) Topology . . . . . . . . 43
3.4 Frequency Power Based Methods . . . . . . . . . . . . . . . . . . . . . 44
3.4.1 VSYNC Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4.2 Virtual SG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Uncategorized Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5.1 Droop Based Approach . . . . . . . . . . . . . . . . . . . . . . . 46
3.5.2 Rate Limiter Approach . . . . . . . . . . . . . . . . . . . . . . . 46
3.5.3 Other methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4 Proposed Methodology and Case Studies . . . . . . . . . . . . . . . . . . 49
4.1 Comparison Between the Existing Methods . . . . . . . . . . . . . . . 49
4.2 Selected Approach to Emulate Inertia . . . . . . . . . . . . . . . . . . . 49
4.3 Case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3.1 Case a . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3.2 Case b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3.3 Case c . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3.4 Case d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3.5 Case e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3.6 Case f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3.7 Case g . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Platform grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2.1 Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2.2 Gas turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Wind turbine and related converters . . . . . . . . . . . . . . . . . . . 57
5.3.1 Wind model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.3.2 Wind turbine, drive train and pitch controller (Wind turbine
package) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3.3 Simplified wind turbine model . . . . . . . . . . . . . . . . . . 67
5.3.4 Doubly Fed Induction Generator (DFIG) . . . . . . . . . . . . 68
5.3.5 Rotor Side converter (RSC) . . . . . . . . . . . . . . . . . . . . 70
5.3.6 Grid Side Converter (GSC) . . . . . . . . . . . . . . . . . . . . . 75
5.4 Virtual inertia provision unit . . . . . . . . . . . . . . . . . . . . . . . . 77
5.4.1 Case e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.4.2 Case f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.4.3 Case g . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.1 Case a: Wind turbine drive train and pitch controller (Wind turbine
package) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.2 Case b: DFIG only with RSC . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.2.1 Case b1: PI controller and wind turbine package . . . . . . . 84
6.2.2 Case b2: MPPT and Simplified wind turbine model . . . . . . 89
Contents ix
6.3 Case c: DFIG with RSC and GSC . . . . . . . . . . . . . . . . . . . . . . 93

6.3.1 Case c1: PI controller with variable wind speed and wind
turbine package . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.3.2 Case c2: MPPT in variable wind speeds and simplified wind
turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.3.3 Case c3: MPPT in constant wind speed and simplified wind
turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.4 Case d: Gas turbine and wind turbine . . . . . . . . . . . . . . . . . . . 112
6.5 Case e: Gas turbine, ESS with P controller for Virtual inertia and
wind turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.6 Case f: Gas turbine, ESS with PI controller for virtual inertia and
wind turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.7 Case g: Gas turbine, ESS with PI controller and dead zone for vir-
tual inertia and wind turbine . . . . . . . . . . . . . . . . . . . . . . . . 122
6.8 Comparison between cases with virtual inertia emulation . . . . . . 125
7 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Figures
2.1 LEOGO platform main components and mass/energy flow [8] . . . 6

2.2 SG components and stator voltages [9] . . . . . . . . . . . . . . . . . . 7
2.3 Simplified SG model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Low and high inertia concepts . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 The inertial response of an SG with H = 1 (red) and H = 3 (green)
to a contingency event [12] . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.6 DFIG connection to the grid [15] . . . . . . . . . . . . . . . . . . . . . 10
2.7 DFIG circuit [15] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.8 DFIG modes of operation [15] . . . . . . . . . . . . . . . . . . . . . . . 12
2.9 Inertia constant for different wind turbine technologies [17] . . . . 14
2.10 A 3 phase back to back converter . . . . . . . . . . . . . . . . . . . . . 15
2.11 Different wind turbine types [20] . . . . . . . . . . . . . . . . . . . . . 16
2.12 (a) CSC, (b) VSC topologies [21] . . . . . . . . . . . . . . . . . . . . . 18
2.13 A simple Grid forming (down) and grid following (up) algorithm
for a P V power plant [23] . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.14 PLL unit diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.15 harmonics definition and waveform . . . . . . . . . . . . . . . . . . . . 21
2.16 abc to αβ to dq transformation . . . . . . . . . . . . . . . . . . . . . . . 23
2.17 (a) Bipolar, (b) Unipolar PWM . . . . . . . . . . . . . . . . . . . . . . . 25
2.18 Three phase inverter with switching states . . . . . . . . . . . . . . . . 25
2.19 Graphical representation of the voltage vectors in each state . . . . 28
2.20 Third harmonic injection (dotted lines are the resultant signals) . . 29
3.1 Methods to provide virtual inertia . . . . . . . . . . . . . . . . . . . . . 33

3.2 Methods to provide virtual inertia [30] . . . . . . . . . . . . . . . . . . 34
3.3 Synchronverter control model . . . . . . . . . . . . . . . . . . . . . . . 35
3.4 Self-synchronised synchronverter model . . . . . . . . . . . . . . . . . 36
3.5 VISMA model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.6 VISMA (a) hysteresis controller, (b) PLL, (c) phase generation, (d)
amplitude generation, (e) grid synthesizing, (f) Distortion com-
pensation subsystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.7 Updated IEPE Controller topology for SG . . . . . . . . . . . . . . . . 39
3.8 KHI SG topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.9 Angular deviation in the KHI PLL . . . . . . . . . . . . . . . . . . . . . 41
xi
xii M. Heidari: Control Models for Providing Virtual Inertia
3.10 KHI topology (a) Governer, (b) AVR and (c) PLL circuits . . . . . . . 42
3.11 Ise topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.12 SPC topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.13 VSYNC topology and its reference current block . . . . . . . . . . . . 45
3.14 VSG topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.15 Droop-based topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.16 Rate limiter topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.17 Inducverter control topology . . . . . . . . . . . . . . . . . . . . . . . . 48
4.1 Project study case concept . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.1 A typcial AVR model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

5.2 Governor model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.3 Wind speed model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.4 kaimal model for wind . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.5 kaimal filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.6 Kaimal coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.7 Wind speed profile of the site . . . . . . . . . . . . . . . . . . . . . . . . 62
5.8 Wind turbine, drive train and pitch controller (wind turbine package) 62
5.9 Wind turbine block parameters . . . . . . . . . . . . . . . . . . . . . . . 63
5.10 Wind turbine power characteristics for default MATLAB values . . . 64
5.11 Wind turbine power characteristics for the modeled wind turbine . 64
5.12 Wind turbine model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.13 Two mass drive train model . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.14 Pitch controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.15 Simplified wind turbine model . . . . . . . . . . . . . . . . . . . . . . . 68
5.16 (left) Torque coefficient-λ, (right) Power-Vwind . . . . . . . . . . . . . 69
5.17 Vector control of the RSC . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.18 PLL circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.19 PLL data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.20 MPPT graph on (a) P − ω, (b) T − ω charts for different wind speeds 74
5.21 MPPT algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.22 Vector control of the GSC . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.23 Vector control of the VIC . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.24 Case e- Virtual inertia provision with P controller . . . . . . . . . . . 79
5.25 Case f- Virtual inertia provision with PI controller . . . . . . . . . . . 79
5.26 Case g- Virtual inertia provision with PI controller and dead zone . 79
6.1 Wind turbine package simulation model . . . . . . . . . . . . . . . . . 81

6.2 DFIG rotor speed reference (Wind turbine package simulation model) 82
6.3 Wind profile, (Wind turbine package simulation) . . . . . . . . . . . 82
6.4 Pitch angle, (Wind turbine package simulation) . . . . . . . . . . . . 82
6.5 Wind turbine speed, (Wind turbine package simulation) . . . . . . . 82
6.6 Shaft input and output torque, (Wind turbine package simulation) 83
6.7 Case b1 simulation model . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Figures xiii
6.8 Case b1- Wind speed (m/s) . . . . . . . . . . . . . . . . . . . . . . . . . 85

6.9 Case b1- Wind turbine speed (pu) . . . . . . . . . . . . . . . . . . . . . 85
6.10 Case b1- Pitch angle (deg) . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.11 Case b1- shaft torque (pu) . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.12 Case b1- Mechanical input torque of generator (N.m) . . . . . . . . . 86
6.13 Case b1- Rotor speed (rad/s) . . . . . . . . . . . . . . . . . . . . . . . . 87
6.14 Case b1- q axis RSC current (A) . . . . . . . . . . . . . . . . . . . . . . 87
6.15 Case b1- d axis RSC current (A) . . . . . . . . . . . . . . . . . . . . . . 87
6.16 Case b1- stator current (A) . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.17 Case b1- rotor current (A) . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.18 Case b1- Electromagentic torque (N.m) . . . . . . . . . . . . . . . . . 89
6.19 Case b2 simulation model . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.20 Case b2- Rotor speed (rad/s) . . . . . . . . . . . . . . . . . . . . . . . . 90
6.21 Case b2- q axis RSC current (A) . . . . . . . . . . . . . . . . . . . . . . 90
6.22 Case b2- d axis RSC current (A) . . . . . . . . . . . . . . . . . . . . . . 90
6.23 Case b2- stator current (A) . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.24 Case b2- rotor current (A) . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.25 Case b2- Electromagentic torque (N.m) . . . . . . . . . . . . . . . . . 91
6.26 Case c1 simulation model . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.27 Case c1- Wind speed (m/s) . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.28 Case c1- Wind turbine speed (pu) . . . . . . . . . . . . . . . . . . . . . 94
6.29 Case c1- Pitch angle (deg) . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.30 Case c1- shaft torque (pu) . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.31 Case c1- Mechanical input torque of generator (N.m) . . . . . . . . . 95
6.32 Case c1- Rotor speed (rad/s) . . . . . . . . . . . . . . . . . . . . . . . . 95
6.33 Case c1- q axis RSC current (A) . . . . . . . . . . . . . . . . . . . . . . 95
6.34 Case c1- d axis RSC current (A) . . . . . . . . . . . . . . . . . . . . . . 96
6.35 Case c1- Electromagentic torque (N.m) . . . . . . . . . . . . . . . . . . 96
6.36 Case c1- stator current (A) . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.37 Case c1- rotor current (A) . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.38 Case c1- DC link voltage (V) . . . . . . . . . . . . . . . . . . . . . . . . 97
6.39 Case c1- q axis GSC current (A) . . . . . . . . . . . . . . . . . . . . . . 97
6.40 Case c1- d axis GSC current (A) . . . . . . . . . . . . . . . . . . . . . . 98
6.41 Case c1- q axis GSC voltage (V) . . . . . . . . . . . . . . . . . . . . . . 98
6.42 Case c1- d axis GSC voltage (V) . . . . . . . . . . . . . . . . . . . . . . 98
6.43 Case c1- rotor (purple), stator (red), and grid (yellow) voltage fre-
quency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.44 Case c1: power delivered to the load (W) . . . . . . . . . . . . . . . . 99
6.46 Case c2- Rotor speed (rad/s) . . . . . . . . . . . . . . . . . . . . . . . . 101
6.49 Case c2- q axis RSC voltage (V) . . . . . . . . . . . . . . . . . . . . . . 102
6.50 Case c2- d axis RSC voltage (V) . . . . . . . . . . . . . . . . . . . . . . 103
xiv M. Heidari: Control Models for Providing Virtual Inertia

quency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.62 Case c3- Rotor speed (m/s) . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.65 Case c3- q axis RSC voltage (V) . . . . . . . . . . . . . . . . . . . . . . 108
6.66 Case c3- d axis RSC voltage (V) . . . . . . . . . . . . . . . . . . . . . . 108
quency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.77 Case d simulation model . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.78 Case d- active power of load [W] (yellow), output power of the gas
turbine (red) and output power of the ESS [W] (green) . . . . . . . 113
6.79 Case d- grid frequency [Hz] . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.80 Case d- gas turbine simulation results. output power [W] (top-
left), input mechanical power [pu] (mid-left), generator speed [pu]
(bottom-left), three phase ac output voltage [V] (top-right), and
three phase ac current output current [A] (mid-right) . . . . . . . . . 114
6.81 Case e simulation model . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.82 Case e- active load power [W] (yellow), gas turbine output power
[W] (red), and ESS output power [W] (green) . . . . . . . . . . . . . 116
Figures xv
6.83 Case e- ESS converter simulation results. d axis current and its ref-
erence [A] (top-left), q axis current and its reference [A] (mid-left),
Output active power and its reference [W] (bottom-left), Output re-
active power [Var] (top-mid), PLL angle [deg] (mid-mid), output
voltage in d and q reference [V] (bottom-mid), ac output voltage
of phase a [V] (top-right), ac output current of phase a [A] (mid-
right), and frequency [Hz] (bottom-right) . . . . . . . . . . . . . . . . 117
6.84 Case e- gas turbine simulation results. output power [W] (top-left),
input mechanical power [pu] (mid-left), generator speed [pu] (bottom-
left), three phase ac output voltage [V] (top-right), and three phase
ac current output current [A] (mid-right) . . . . . . . . . . . . . . . . 118
6.85 Case f- active load power [W] (yellow), gas turbine output power
6.86 Case f- ESS converter simulation results. d axis current and its ref-
6.87 Case f- gas turbine simulation results. output power [W] (top-left),
input mechanical power [pu] (mid-left), generator speed [pu] (bottom-
left), three phase ac output voltage [V] (top-right), and three phase
ac current output current [A] (mid-right) . . . . . . . . . . . . . . . . 121
6.88 Case g- active load power [W] (yellow), gas turbine output power
6.89 Case g- ESS converter simulation results. d axis current and its ref-
6.90 Case g- gas turbine simulation results. output power [W] (top-
left), input mechanical power [pu] (mid-left), generator speed [pu]
(bottom-left), three phase ac output voltage [V] (top-right), and
three phase ac current output current [A] (mid-right) . . . . . . . . . 124
6.91 Frequency response of the platform grid in case d (blue), e (red), f
(yellow), and g (purple) to the load change . . . . . . . . . . . . . . . 125
6.92 Virtual inertia block response in case d (blue), e (red), f (yellow),
and g (purple) to the load change . . . . . . . . . . . . . . . . . . . . . 125
6.93 Gas turbine speed response in case d (blue), e (red), f (yellow),
and g (purple) to the load change . . . . . . . . . . . . . . . . . . . . . 126
xvi M. Heidari: Control Models for Providing Virtual Inertia
6.94 Gas turbine output power in case d (blue), e (red), f (yellow), and
g (purple) to the load change . . . . . . . . . . . . . . . . . . . . . . . . 126
1 LEOGO platform electrical grid . . . . . . . . . . . . . . . . . . . . . . . 141

2 LEOGO platform-name of each component . . . . . . . . . . . . . . . 141
Tables
2.1 Switching states and line voltages . . . . . . . . . . . . . . . . . . . . . 27

2.2 Switching states and phase voltages . . . . . . . . . . . . . . . . . . . . 27
3.1 The operation modes of the self-synchronised synchronverter . . . . 35
5.1 Gas turbine generator data . . . . . . . . . . . . . . . . . . . . . . . . . 54

5.2 Gas turbine AVR data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3 Gas turbine governor data . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.4 Wind turbine data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.5 Default wind turbine data of Matlab . . . . . . . . . . . . . . . . . . . 63
5.6 DFIG data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.7 VIC data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
1 Field characteristics parameter values . . . . . . . . . . . . . . . . . . 140

2 Generator data for the LEOGO platform . . . . . . . . . . . . . . . . . 142
3 Transformer data for the LEOGO platform . . . . . . . . . . . . . . . . 143
4 Variable Speed Drive (VSD) load data for the LEOGO platform . . . 144
5 PQ load data for the LEOGO platform . . . . . . . . . . . . . . . . . . 145
6 Fixed speed motors data for the LEOGO platform . . . . . . . . . . . 146
xvii
Acronyms
M il.Sm3 Million Standard cubic meters. 139
Q in j Injection Rate. 140
Q oil Oil Production. 140
AVR Automatic Voltage Regulator. 39, 40, 54, 55
CSC Current Source Converter. 17
CSV Constant Stator Voltage. 70
DCF Distortion Component factor. 36
DFIG Doubly Fed Induction Generator. viii, 9–11, 15, 32, 51, 57, 60, 62, 63, 67–
69, 73, 74, 81, 83–86, 89, 92, 93, 100, 129
DG Distributed Generation. 47
DSP Digital Signal Processing. 36
ENTSOE European Network of Transmission System Operators for Electricity. 17
ESS Energy Storage System. 1, 14, 31, 44, 50–52, 54, 77, 78, 112, 115, 122, 127,
129–131
GOR Gas Oil Ratio. 140
GSC Grid Side Converter. 10, 17, 51, 57, 69, 75–77, 93, 100, 106, 111
GTO Gate Turn Off transistor. 17
IEPE Institute of Electrical Power Engineering. 35, 36, 49
IGBT Insulated Gate Bipolar Transistors. 14, 17
IGCT Integrated Gate Commutated Thyristor. 14
xix
xx M. Heidari: Control Models for Providing Virtual Inertia
KHI Kawasaki Heavy Industries. 39, 40, 49
LEOGO Low Emission Oil and Gas Open. 5, 6, 53, 139
MOSFET Metal Oxide Semiconductor Field Effect Transistor. 14, 17
MPPT Maximum Power Point Tracking. 51, 67, 71, 74, 75, 89, 92, 100, 101, 106,
107, 129
NCS Norwegian Continental Shelf. 5
NREL National Renewable Energy Lab (US). 1
NTNU Norges teknisk-naturvitenskaplige universitet. 2
PCC Point of Common Coupling. 37, 76, 86, 111, 112, 129
PFCC power Factor Correction Capacitor. 16
PLL Phase-Locked Loop. 19, 32, 34, 36, 40, 41, 44, 47, 49, 71
PMSG Permanent Magnet SG. 7, 9, 16
PSD Power Spectral Density. 59
PWM Pulse Width Modulation. 22–25, 28, 32, 41, 44, 70, 71, 100, 111
R&D Research and Development. 6
RES Renewable Energy Source. 1, 14, 18, 31, 129
ROCOF Rate of Change of Frequency. 1, 9, 13, 44, 45, 125, 130
RSC Rotor Side Converter. 51, 57, 69, 70, 76, 93, 100, 106, 129
SCIG Squirrel Cage Induction Generator. 15, 16
SG Synchronous Generator. 1–3, 5–8, 10, 13, 16, 18, 31, 32, 34, 35, 37, 39, 41,
43, 44, 47, 49, 54, 115, 129
SMES Superconducting Magnetic Energy Storage. 17
SOC State of Charge. 44
SPC Synchronous power Controller. 43, 49
SRM Spectral Representation Method. 59
SVPWM Space Vector Pulse Width Modulation. 25, 28, 71, 72, 129
Tables xxi
THD Total harmonic Distortion. 20
TSR Tip Speed Ratio. 65, 68
UPF Unity power Factor. 70
UPS Uninterrupted Power Supply. 139
VIC Virtual Inertia Converter. 77, 78
VISMA Virtual Synchronous Machine. 35
VOC Virtual Oscillator Controller. 19, 47
VSC Voltage Source Converters. 17
VSD Variable Speed Drive. xvii, 144
VSG Virtual Synchronous Generator. 39, 41, 43, 44
VSM Virtual Synchronous Motor. 32, 34
WC Water Cut. 140
WRIG Wound Rotor Induction Generator. 15
ZDC Zero D-axis Current. 70

Chapter 1
Introduction
1.1 Introduction
The continuous increase of demand for carbon emission reduction entails the
integration of the Renewable Energy Source (RES) into the electricity grids. One
of the major concerns of this change is the downgrading of the system inertia.
Power converters which are the main bridge between RES and the electricity grid
do not possess any inherited inertia [1, 2].
National Renewable Energy Lab (US) (NREL) refers to inertia in the power
systems as "the stored energy in large rotating masses which gives them the tend-
ency to remain rotating" [3]. The inertia in the system is mainly the factor that
stabilises Rate of Change of Frequency (ROCOF) after a contingency event. Con-
sequently, if the inertia constant of the system components is not adequate, the
system might experience a drastic ROCOF. On the other hand, high inertia, rein-
forces the grid and prevents cascading failures and undesirable load shedding. In
order to tackle this problem the concept of virtual inertia was introduced.
Virtual, synthetic, or hidden inertia is referred to any combination of power

electronic components, Energy Storage System (ESS), RES, and control algorithms
that virtually mimics the inertial response of the conventional power systems [4–
6]. This virtual inertial response is the result of modifying the classic control meth-
ods of the power electronic converters. This modification generally aims to extract
more energy from any stored form of energy behind the converters while reducing
the ROCOF in comparision with a scenario in which the converters are controlled
with the classic methods. From the grid viewpoint, this response is the same as
the one from a Synchronous Generator (SG) in the event of a contingency.
1.2 Motivation
The oil and gas sector plays a vital role in financing the Norwegian economy.
However, it is also one of the country’s key carbon emitters. Roughly 20% of Nor-
1
2 M. Heidari: Control Models for Providing Virtual Inertia
way’s total greenhouse gas emissions come from the offshore gas turbines oper-
ating on the Barents, North and Norwegian Seas. More environmentally friendly
solutions, like offshore wind, have the potential to cut drastically those carbon
emissions. However, as mentioned in section 1.1, wind generators and their power
electronic converters lack the inherent inertia of SG that stabilizes electrical sys-
tems. Therefore, a deep understanding of virtual inertia strategies and their sim-
ulation models is necessary in order to study the impact of the integration of
wind-farms into offshore platforms.
This thesis focuses on modeling control strategies for maintaining the stabil-
ity of an offshore platform electrical system with a large contribution of wind
energy. The project is a part of the Low emission Sub-Project on Energy Systems
and Digital Solutions of the Low emission Research Centre, a cooperation between
Norges teknisk-naturvitenskaplige universitet (NTNU) and SINTEF A. S. with the
goal of developing technologies and solutions for reducing the offshore green-
house gas emissions on the Norwegian Continental Shelf by 40% within 2030
and move towards zero emissions in 2050.
1.3 Objectives and Research Questions

The project is divided into smaller parts with separate goals for convenience
and in the end assembled back together. The following objectives are defined to
develop the model:
• Selection and modeling the wind farm and its control strategy
• Selection and modeling the energy source to provide virtual inertia
• Selection and modeling the control strategy to emulate the virtual inertia
• Modeling the grid of the oil and gas platform
• Assemble each part to build the complete model
• Successfully run the model and illustrate the operation of the virtual inertia
control strategy
During the modeling of the grid of the oil and gas platform, the following
research questions have to be answered:
• What are the important loads on the grid of oil and gas platform? How can
they be modeled? How much detail is required for the purpose of this study?
• What are the conventional energy sources on the platform? Should they be
altered in any way to integrate the wind energy? Is it possible/necessary to
completely remove these generators?
Chapter 1: Introduction 3
• How should the wind turbines be modeled? How are they connected to the
oil and gas grid? What should be the control strategy for power converters?
What are the limitations to be considered in the modeling?
• Which methods can be implemented for providing virtual inertia? What is

the source of energy behind this inertia?
• How is the power distributed between the sources and the loads? What is
the role of each component and how essential is their contribution?
1.4 Methodology
To achieve the set goals of this thesis, different control schemes are tested
and implemented in the Simulink graphical modeling tool, within the MATLAB
numerical integrated environment. The main contributions of the thesis can be
identified as:
• Provide an overview of the existing methods to provide virtual inertia;
• Develop a control strategy to emulate virtual inertia for the platform;
• Evaluate the effectiveness of the proposed method in the electric grid the
oil and gas platform.
1.5 Thesis Outline

Chapter 2 provides the background and concept of inertia in the power sys-
tems. In this chapter, the inertial response of the traditional power plants with
SGs is studied. Furthermore, the characteristics and loads of the understudy oil
and gas platform are introduced. Subsequently, the changes in the wind turbine
technology is discussed and the necessity for virtual inertia is investigated. Addi-
tionally, this chapters discusses power converters and their relevant background.
Chapter 3 thoroughly studies the methods to provide virtual inertia in the literat-
ure, while presenting a background to the changes in the proposals for tackling
the issue. Chapter 4 discusses the proposed method to emulate virtual inertia for
the offshore wind turbines that support the oil and gas platform and presents the
case study. Chapter 5 discusses steps taken for the modeling of the entire system.
Chapter 6 gathers the simulation results, pointing out the satisfactory emulation
of the virtual inertia in the system. Finally, conclusions and recommendations for
future work are offered in Chapter 7.
Chapter 2
Background
2.1 Introduction
The operation of the traditional power plants is entirely different from the op-
eration of the offshore wind parks. However, in order to understand the essence of
inertia in the power system, the traditional power plants must be part of the study.
In the traditional power plants, during the event of an imbalance in the available
power supply and load, the SGs speed up or slow down via the governor to com-
pensate for the imbalance. In this chapter, the background of the electric system of
the the oil and gas platforms is introduced. Afterward, the inertial response of the
SGs is provided and a connection is made to the offshore wind farms and changes
happening in the methods to provide inertia for wind farms are discussed. A thor-
ough background for power converters and their relevant technologies are also
presented.
2.2 Oil and Gas Platforms

The oil and gas offshore platforms consist of multiple energy-demanding units
such as drilling, processing, exporting, and accommodation. Most platforms on
the Norwegian Continental Shelf (NCS) generate their own electrical power by
gas turbines. The gas turbines are also used to directly drive compressors and
pumps. These gas turbines generate about 80% of the total CO2 and emissions
from offshore installations [7]. As mentioned in section 1.2, this thesis investig-
ates methods to provide virtual inertia for oil and gas platforms supplied by wind
turbines. Hence, an analysis and modeling of a typical oil and gas platform is es-
sential.
A consortium consisting of Equinor, Sintef A. S., and Siemens A. S. started a

project in 2016 to implement the integration of offshore wind with a stand-alone
electric grid for oil and gas offshore installations. The Low Emission Oil and Gas
Open (LEOGO) reference platform is selected as the main structure to study in
5
this thesis [8]. LEOGO is a hypothetical oil and gas developed by the Low Emis-
sion centre of the SINTEF A. S. for Research and Development (R&D) purposes.
Figure 2.1 illustrates all the components in the hypothetical platform. This figure
shows the connection of the energy storage, wind turbines, gas turbines and other
components of the grid of the oil and gas platform. Since the main goal of this
thesis is investigating the control methods to provide virtual inertia, less emphasis
is put on modeling the loads of the platform itself, meaning that except the gas and
wind turbines and energy storage unit, the rest of the components in figure 2.1
are not modeled in detail but a whole representation of the electric loads are con-
sidered. Nevertheless, the specific model of the system is thoroughly explained in
Annex 7.2 for future reference. In this Annex, the details of the power suppliers of
the platform and its main electrical loads are explained and the electrical grid of
the platform is illustrated in Simulink. The complete model of the built system of
this study is presented in chapter 4 and the modeling of each element is available
in chapter 5.
Figure 2.1: LEOGO platform main components and mass/energy flow [8]
2.3 Source of Inertia in the Power Systems

2.3.1 SG Model and Operation Principles
The SGs are the main source of energy supply to the electric grids. A SG con-
sists of a set of windings that interact inside a magnetic field to convert mechan-
ical energy into electricity: stator windings which are 3 AC winding distributed
at 120◦ spaced along the circumference of the stator magnetic circuit and rotor
winding which is a DC field winding. The shape of the rotor can be either roun-
ded rotor (in turbogenerators- high speed prime movers) or salient pole rotor (in
hydrogenerators- low speed prime movers). The field of the rotor can be created
Chapter 2: Background 7
either by a DC excitation (adding a DC field winding) or by using permanent mag-

nets in the rotor (Permanent Magnet SG (PMSG)).
When the rotor shaft is turned by an external force, the rotor poles also rotate
and create a rotating magnetic field BR . As the magnetic field rotates, it passes
from the stator windings and according to Faraday’s law, an electromotive force
(emf ) is induced in the stator windings as illustrated in figure 2.2. The frequency
of the output voltage depends on the number of pole pairs (P) and the rotational
speed of the rotor also known as angular velocity (ω) in r ad/s, Eq. 2.1 shows this
relation:
P ·ω
f = (2.1)
2π
Figure 2.2: SG components and stator voltages [9]
The internal voltage EA of the SG can be calculated from Eq. 2.2. In this equa-
tion, Nc is the number of conductors at angle zero. The induced voltage is pro-
portional to the rotor flux ϕ for a given angular frequency. When generator is not
loaded, the internal voltage is equal to the terminal voltage VΦ and when loaded,
current flows in stator windings which result in the rotating stator magnetic field.
The voltage induced in the armature is the summation of the voltages induced by
both rotor and stator fields. The output terminal voltage of the SG VΦ ,would be
achieved by deducting the voltage drops from self and leakage inductances and
resistance of the armature coils from the resultant voltage.
p
EA = 2πNc ϕ f (2.2)
VΦ = EA − IA(Rs + jX s ) (2.3)
If a load is connected to the output terminal of the generator, the current IA

starts to flow and the transmitted active and reactive power P, Q can be calculated:
Figure 2.3: Simplified SG model
S = 3VA · IA∗ , (2.4)

Zs = Rs + jX s (2.5)
2
|EA| · |VΦ | |V |
P =3 · cos (γ − θ ) − 3 Φ · cos γ (2.6)
|Zs | |Zs |
|EA| · |VΦ | |V |2
Q=3 · sin (γ − θ ) − 3 Φ · sin γ (2.7)
|Zs | |Zs |
in which, EA, VΦ are the internal and terminal voltages, γ, θ are the phase angle
between EA, VΦ and VΦ , IA vectors and Zs is the generator impedance. Considering
that the value of Rs is quite negligible in the SG, the active and reactive powers
can be simplified:
|EA| · |VΦ |
P =3 · sin θ (2.8)
|X s |
|EA| · |VΦ | · cos θ − |VΦ |2
Q=3 (2.9)
|X s |
2.3.2 Traditional Power Plants

The main idea behind providing inertia for the grid in the traditional power
plants was utilizing the capability of the SGs to slow down or speed up if necessary
to cope with the power imbalances and provide a reasonable inertial response. The
inertial response of a SGs follows the swing Eq. 2.10 [10]:
2H d 2 δ
= Pm − Pe = Pa (2.10)
ω d t2
in which H is the inertial constant, ω is the angular velocity (synchronous
speed), δ is the rotor angle and Pm and Pe are the mechanical and electrical power
respectively. The difference between Pm and Pe represents the accelerating/de-

celerating power (Pa ) for the generator. In the normal operation with no power
imbalance, the electrical and mechanical powers are equal and therefore, Pa = 0
and the generator runs at synchronous speed. The loads connected to the gen-
erator can vary (according to eq 2.8) much faster than the mechanical power of
the generator because the generator inherits inertia and this can lead to instability.
Figure 2.4: Low and high inertia concepts
The kinetic energy that is injected to the grid has to be replaced to provide
support again if needed. Figure 2.4 illustrates the concept of the power imbal-
ance in the system and the effect of inertia on the ROCOF. As the figure shows,
lower inertia translates to faster ROCOF, which could be disastrous. Moreover,
higher inertia constant allows larger imbalances to occur with slower ROCOF.
Additionally, the inertial constant affects the frequency nadir as depicted in fig-
ure 2.5. Frequency nadir is the point in which the frequency the highest drop in
value during a contingency event. Lower inertia leads to a lower frequency nadir,
resulting in a higher probability of system collapse and failure [11].
2.3.3 DFIG model and operation principles

DFIGs have been used in wind turbines as a generator for a long time and
although now PMSGs are preferred as the newer generators, a considerable por-
tion of the installation still use DFIG. PMSGs are usually low speed and have no
gearbox and their control strategy is straightforward but they are more expensive
(additional cost because of the permanent magnets) and require full-rated con-
verters due to the type 4 connection (explained in section 2.4.1) to the grid. On
the other hand, lower cost (no need for permanent magnets and smaller scaled
converters, usually one-third of the generator rated power) is an important ad-
vantage for DFIG, although considered as old-fashioned technology which needs
gearbox [13]. This section explains the operation principles of DFIG based on [14]
and chapter 10 of [15]. In this project, DFIG was selected instead of PMSG tech-
nology, hence the DFIG model is introduced more detailed.
A DFIG is supplied via a type three connection to the grid (explained in section
Figure 2.5: The inertial response of an SG with H = 1 (red) and H = 3 (green)

to a contingency event [12]
2.4.1). Figure 2.6 illustrates the connection of the DFIG to the grid. Rotor and grid
side filters are sometimes used for filtering the currents of the converters. Also,
a capacitor is used in the DC link. The selection and design of the capacitor and
filters are explained in chapter 5 under Grid Side Converter (GSC) control section.
The model of the converter controllers are also presented in chapter 6.
Figure 2.6: DFIG connection to the grid [15]
One of the main differences between DFIG and SG is that in DFIG the fre-
quency in different windings are not identical. Eq 2.11 denotes the relation between
the frequencies.
ωs = ω r + ω m (2.11)
ωm = P · Ωm (2.12)
ωs − ω m ω
s= = r (2.13)
ωs ωs
in which ωs , ω r and ωm are the frequency of the stator, rotor voltages/currents

and rotor electrical speed, respectively. Ωm is the mechanical speed of the shaft,
P is the number of pole pairs, and s is the slip of the machine. Note that, these
equation are applicable both on speed and frequency. According to typical design
of the DFIG, maximum slip can be considered 1/3 which means approximately
30 % of the stator power flows through the rotor. Since in most applications, the
stator is directly connected to the grid, ωs is equal to the grid frequency, known
as synchronous frequency. A DFIG can operate in three different modes according
to the previously mentioned speeds/frequencies:
• Sub-synchronous mode: ωm < ωs → ω r > 0 → s > 0
• Hyper-synchronous mode: ωm > ωs → ω r < 0 → s < 0
• Synchronous mode: ωm = ωs → ω r = 0 → s = 0
The relation between the stator and rotor winding turns are presented by para-
meter u, therefore the stator and rotor emf relations can be defined as:
0 Es
E rs = s (2.14)
u
In general to facilitate the analysis of the model, all the rotor parameters can
be moved to the stator side. Figure 2.7 illustrates the schematic of this model.
The rotor values when moved to the stator side (as shown in the figure) are as
followed:
0
0
2 0
2
Ir 0 0
R r = RR · u ; Lσr = Lσr · u ; I r = ; Vr = Vr · u; E rs = E rs · u (2.15)
u
in which all the superscripted values are the original rotor values and the val-
ues without superscripts are the reflected values to the stator.
The power flow in the machine according to figure 2.7 can be expressed as
followed:
Ps + Pr = Pcus + Pcur + Pm (2.16)

in which Ps , Pr , Pcus , Pcur and Pm are stator, rotor, loss in stator, loss in rotor and
shaft’s mechanical powers. If Pm is positive, the machine is operating as generator
Figure 2.7: DFIG circuit [15]
Figure 2.8: DFIG modes of operation [15]
and if Pm is negative, the machine runs as a motor. The sign of Ps is always oppos-
ite of Pm and the sign of Pr determines if the machine is in hyper-synchronous,
sub-synchronous or synchronous modes. Figure 2.8 illustrates this concept.
The power parameters of the machine can be calculated from the classic power
equations:
Ps = 3Re{Vs · Is∗ }, Pr = 3Re{Vr · I r∗ } (2.17)

Q s = 3I m{Vs · Is∗ }, Q r = 3I m{Vr · I r∗ } (2.18)
Substituting the values from KVL in circuit of figure 2.7:

Ps = 3Rs | Is |2 +3ωs L m I m{Is · I r∗ }, Pr = 3R r | I r |2 −3sωs L m I m{Is · I r∗ } (2.19)

Q s = 3ωs Ls | Is |2 +3ωs L m Re{Is · I r∗ }, Q r = 3sωs Ls | I r |2 +3sωs L m Re{Is · I r∗ }
(2.20)
Additionally, the electromagnetic torque of the shaft can be written as:
ωm
Pmec = Tem · ωm = Tem (2.21)
P
Lm
Tem = 3P L m I m{I r∗ · Is } = 3P I m{φs · I r∗ } =
Ls
3P · I m{φs∗ · Is } = 3P · I m{φ r · I r∗ } =
Lm Lm
3 P · I m{φ r∗ · Is } = 3P I m{φs · φ r∗ } (2.22)
Lr σLs L r
where σ = 1− L m2
/Ls L r . The copper losses can be neglected in the calculations
to perform faster analysis and since the stator voltage is constant, the rotor voltage
will only depend on slip/speed. With these assumptions in mind, the operation of
the machine can be divided into four categories as demonstrated in figure 2.8.
2.3.4 Inertia for Wind Turbines

Unlike traditional power plants, the wind turbines have to compensate for the
inertia by other means, due to the decoupling between grid and generators via the
converters. One method to provide Inertia could be through the kinetic energy in
the blades [3]. This service requires the operation of the wind turbine to be in the
overloaded power zone for 100 ms and while slowing down the turbine and the
reduction of ROCOF, other generators in the grid will have the window to inject
more power to the grid and balance the system. Although, extra caution has to be
applied in extracting this energy, since the wind turbine requires certain kinetic
energy to operate continuously and if this energy is extracted, it must be replaced
as quickly as possible, same as in SGs. Moreover, this is not always the situation
in the industry and power deficit can be the main issue, therefore in this project
the scenarios are defined based on power deficit.
Wind turbine installations are mostly shifting towards operating at variable

speeds and connection via back-to-back converters, decoupling the grid and the
turbine entirely [16]. Figure 2.9 illustrates the inertial constant of wind turbine
technologies based on the size of the generation unit. The trend in the literature
of wind turbines implies that with the increasing demand for variable speed wind
turbines, the inertial response has to be much more scrutinised.
Expanding the control algorithms of the wind turbine is the key to introdu-
cing virtual inertia, and the desired response for the system can be implemented
by producing feedback signals from terminal current and voltage of the point of
common coupling [18].
Figure 2.9: Inertia constant for different wind turbine technologies [17]
2.4 Converter Technology
Power converters are used for converting AC or DC voltages to AC or DC

voltages with different amplitude and/or frequency. DC to AC converters, also
known as inverters, are commonly used to connect the DC sources such as RESs
and ESSs to the AC electrical grid. At the same time, the opposite applies to AC
to DC converters (rectifiers). Additionally, to synchronize two grids with differ-
ent frequencies converters can be used. A typical converter consists of a set of
static and/or mechanical parts. The most recent converters are utilizing the semi-
conductor technology of controllable or uncontrollable switches (such as diodes,
thyristors, Insulated Gate Bipolar Transistors (IGBT), Integrated Gate Commut-
ated Thyristor (IGCT), Metal Oxide Semiconductor Field Effect Transistor (MOS-
FET)) to perform switching in the circuits and obtaining the desirable waveforms.
Figure 2.10 illustrates a simple controllable 3 phase back to back converter

that connects a three phase generator to an AC grid via a DC link in between,
and IGBT switches. As connecting the generator side with a DC link to the grid
in long distance HV applications has proven to be less costly than traditional AC
lines, more and more DC links are emerging worldwide to connect the RES plants
to electrical grid [19]. Power converters can be categorized based on different
principles such as application, types of connection to the grid, voltage or current
source based, grid forming or grid following and generally depending on the con-
trol methods defined for the converter. As the links between power plants, grids
and customers, converters play a substantial role in controlling the grid properties
and ensuring power quality.
Figure 2.10: A 3 phase back to back converter
2.4.1 Types of grid connection for wind turbines

Wind turbines are usually classified by the control strategies implemented
for mechanical power and generator speed. A common classification method is
stall vs. pitch regulated turbines. In the stall regulated category, the wind turbine
blades are shaped in such way that for very high speed winds, the airfoils generate
less aerodynamic force, thus resulting in reduction of torque. While in the pitch
regulation, the blades are twisted around their axis and as the speed of the wind
increases, the blades are pitched into a new position to capture maximum energy
or protect the blade if necessary. Another categorization of the wind turbines are
called types. 4 main types are defined for wind turbines:
• Type 1:
limited variable speed (type 1), which consists of a Squirrel Cage Induction
Generator (SCIG) connected directly to a transformer and operates at grid
frequency. The operating speed of the turbine, under steady condition is
almost a linear function of the torque for a certain wind speed.
• Type 2:
Similar to type 1, type 2 wind turbine has limited variability in the speed,
while it consists of a Wound Rotor Induction Generator (WRIG). The con-
nection is also similar to type 1 and directly via a step up transformer, al-
though a variable resistor is added in the rotor circuit. The resistor is usually
a set of resistors and power electronics components external to the rotor, and
the current flows in the slip rings connected to the rotor. If the resistor is
mounted directly on the rotor, the need for slip rings is eliminated and the
model is called Weier design. The addition of resistors enables the control
for the current and provision of constant power, even during gust winds and
disturbances.
• Type 3:
Type 3 wind turbines, commonly known as DFIG, benefit from variable
speed control with partial power electronic conversion technology. In this
Figure 2.11: Different wind turbine types [20]
model in addition to advantages of type 2, a variable AC excitation system is

added to the rotor circuit, which is supplied through slip rings and a voltage
sourced converter. The converter is then connected back to back with a grid
side converter that exchanges power with the electric grid.
• Type 4:
Type 4 turbines have great flexibility as they are both variable speed and
incorporate full power electronic conversion technology. In this model, the
gearbox can be eliminated as the machine can spin at slow turbine speeds
and considerably less frequencies than the grid. The generator can be wound
rotor synchronous machines, PMSG, or SCIG.
Another type of wind turbine technology (type 5) is also available which is not
used quite often. In this type, a typical variable speed wind turbine is connected
to the grid by a speed/torque converter which is coupled with a SG. The converter
changes the variable speed of the rotor shaft to a constant output speed.The to-
pology of the four main types can be found in figure 2.11.
Grid codes of different countries have certain requirements for wind farms and
the properties such as voltage, active and reactive power must be kept in certain
ranges. Other requirements are also set by grid codes and generally all of them
can be categorized into 5 sections of frequency stability, robustness, system res-
toration, general system management and voltage stability. As mentioned before,
the control capabilities of a wind turbine generator depend on the generator type.
Types 1 and 2 can not control the voltage and they use power Factor Correction
Capacitor (PFCC)s to maintain active, reactive powers and voltage in a desirable

region. Types 3 and 4 are capable of controlling the voltage by varying reactive
power at specified active power and voltage set points. Types 3 and 4 are more
common as they offer not only this flexibility but also fulfill other requirements of
the grid codes.
2.4.2 Voltage source vs current source converters

European Network of Transmission System Operators for Electricity (ENTSOE)
defines Voltage Source Converters (VSC) as "self-commutated converters that con-
nect HVAC and HVDC systems using devices suitable for high power electronic
applications, such as IGBTs." The simplest difference between voltage source and
Current Source Converter (CSC) as evident by their name is the DC side para-
meter. If the voltage on the DC side is constant, it considered a voltage source
converter, while if the current of the DC link is constant, the converter is current
sourced.
The notable characteristics of the two technologies are as followed:
VSC-
• Input voltage is constant
• Output voltage is independent of the load
• Output current depends on the load
• Needs feedback diodes
• Complicated commutation circuit
• MOSFET, IGBT, Gate Turn Off transistor (GTO) can be used
CSC-
• Input current is constant
• The magnitude of the output current is independent of load
• Output voltage depends on the load
• Does not need feedback diodes
• Simple commutation circuit
• MOSFET, IGBT, GTO can not be used
VSCs are preffered to CSCs in most of the applications for their benefits. Ac
motor drive, Superconducting Magnetic Energy Storage (SMES) are application
that GSC is preferred to VSC [21].
(a)
(b)
Figure 2.12: (a) CSC, (b) VSC topologies [21]
2.4.3 Grid forming vs grid following converters

The RESs such as wind farms are connected to the grid thorough the power
converters. These converters are operated in either grid following or grid forming
modes. Currently, most of the converters operate in the grid following mode but in
islanded mode of microgrids they should be capable of operating in the grid form-
ing mode. If the converter regulates the power output by measuring the angle of
the grid voltage, it is operating as a grid following unit and follows the grid voltage
angle and frequency without actively controlling them. On the other hand, the grid
forming mode engages with controlling frequency and voltage angle, widely used
in microgrid applications [22].
In the grid forming units, the rotational speed of the SG is directly linked
with the frequency and by changing the speed, the frequency can be controlled.
However, for the wind turbines with less support from SG, other algorithms are
proposed to achieve the grid forming/following modes of operation for the con-
verters. These methods are discussed in the next chapter.
Figure 2.13 depicts a simple example of the algorithm for the grid following
and forming converter units in a P V plant. As illustrated, in the grid following
mode, voltage angle and frequency are inputs and power is the output of the
control algorithm. However, in the grid forming unit, the power is input and the
voltage angle and frequency are the outputs of the control algorithm.
2.4.4 PLL units

Grid connected converters commonly require a dedicated synchronization to
provide the phase and frequency of the grid voltage, in order to be connected to
Figure 2.13: A simple Grid forming (down) and grid following (up) algorithm
for a P V power plant [23]
the grid smoothly. A Phase-Locked Loop (PLL) is a control system that generates
an output signal whose phase is related to the phase of the input signal. This im-
plies that the frequency of both input and output signals are the same. Different
PLL models can also generate signals with frequencies of multiple times of the
input frequency.
The basic concept of operation of the PLL units are rather simple, however the
mathematical analogy and the elements involved can be complicated. The dia-
gram for a basic PLL is provided in figure 2.14. In this model, the signals from
Virtual Oscillator Controller (VOC) and reference signal are connected to the in-
put ports of the phase detector. Here the phase signals from the two inputs are
compared and an error signal is generated. Then a low pass loop filter, receives the
error signal from the phase detector and removes any high frequency harmonics
that the signal might have. Then the error signal is passed back to the VOC as its
tuning voltage. The operation of the loop is in such way that it reduces the dif-
ference in phase of the two main signals and synchronises their frequencies. The
looping operation continues until the phase detector can not generate the error
signal and the loop will be locked. In chapter 5, the proposed model for synchron-
ization is discussed.
The main advantage of the PLL units as mentioned, are the synchronisation
capability they offer to the system. However, PLLs are inherently nonlinear, noise
sensitive and difficult to tune. Extra caution is required when designing PLLs be-
cause multiple PLL units can compete with each other and cause complexity, re-
duced performance and instability [24, 25].
2.4.5 Harmonics
Harmonics are basically the integer multiples of the fundamental frequency
of a trigonometric sin or cos signal. Based on fourier analysis, any signal can be
Figure 2.14: PLL unit diagram
expanded to sinusoidal waveforms with different harmonics. In the power sys-

tems, harmonics are a result of non-linear loads that draw non-sinusoidal current
from a sinusoidal voltage source such as motor drives or converters (which com-
pose a large portion of the grid itself). Harmonics on nature can be extremely
dangerous specially in the power grids. They can cause voltage distortion, forced
increase in the nominal utility sizing, reduced power factor, equipment malfunc-
tion, component overheating, resonance phenomenon and so on. Different grid
codes set standards for acceptable ranges of harmonics which are defined as Total
harmonic Distortion (THD). According to IEEE standard of 519-1992 the total ac-
ceptable THD is from 3.0 % to 10.0 % according to application. THD is defined as
below:
q
V22 + V32 + V42 + ...
THD = (2.23)
V1
in which Vn is the RM S value of the nth harmonic.
In power converters, it is inevitable to have THD to some extent. Reason is that

the process of transforming voltage from ac to dc and dc to ac in nature is based
on switching components which operate with pulses, later on explained in detail.
In order to reduce the amount of harmonics injection many methods have been
developed thus far. Reactors or chokes, passive and active filters, harmonic cancel-
lation methods with high pulse count rectification and drive isolation transformer
are the general methods. In the design of the power converters in this study the
filters act an essential role. The main concerns about harmonics in this thesis are
about power factor (active power delivered to the loads) and voltage distortion.
The design procedure of the filters is explained in chapter 5.
Figure 2.15: harmonics definition and waveform
2.4.6 Vector transformation
The control and analysis of a three phase system can be difficult if one decides
to model every components in the abc reference frame. Therefore, vector trans-
formation has been used since long time ago to facilitate the analysis of system
components, control systems and so on. Two of such vector transforms are the
Park and Clarke transforms.
2.4.6.1 Clarke vector transformation
The Clarke transformation changes three rotating vectors in fixed abc axis into
stationary vectors on rotating axis reference frames (αβγ):
1 1
 
1 − −

Vα  2 2  V (t)
 p p  a
  2 3 3
Vb (t)
Vβ  =  (2.24)

0 −
3

2 2 

 Vc (t)

Vγ 
1 1 1
2 2 2
And the inverse of the Clarke transformation:

 
1 0 1  
Va (t)
 
p  Vα
1 3

  
Vb (t) =  − 1
Vβ  (2.25)
 
 2 2
p
Vc (t) 1 3
 
Vγ
− − 1
2 2
2.4.6.2 Park transformation

The Park transformation, also known as dq0 transformation, rotates the refer-
ence frame of the αβ to an arbitrary frequency.
cos (θ ) sin (θ )
0 V 
   
Vd α
  2  
 Vq  = − sin (θ ) cos (θ ) 0Vβ 
  (2.26)
3
V0 Vγ
0 0 1
The Clark and Park transformation combined together are the basis of the
vector transformation used in the three phase converter control system, called
the DQZ reference frame:
2π 2π
 
cos (θ ) cos (θ − ) cos (θ + )
 
VD  3 3  V (t)
v   a
  t2 2π 2π 
 VQ  = − sin (θ ) − sin (θ − ) − sin (θ + )
Vb (t) (2.27)
 
3
 p 3 3
VZ p p  Vc (t)

2 2 2

2 2 2
And its inverse:
p 
2

cos (θ ) − sin (θ )

Va (t)

 2  V 
v  p  D
 t2
cos (θ − 2π ) − sin (θ − 2π ) 2
Vb (t) =  VQ  (2.28)
  
3 3 3 2 
Vc (t) p  VZ


2π 2π 2
cos (θ + ) − sin (θ + )
3 3 2
2.4.7 Modulation techniques

The most extended modulation technique for inverters is Pulse Width Mod-
ulation (PWM). This technique is relatively easy to implement and can reduce
Figure 2.16: abc to αβ to dq transformation
the low frequency harmonics to some extent. The method operates on the basis
of switching at high frequencies and trying to maintain an average value equal
to the reference input signal (which is usually sinusoidal). The method works by
comparison of a reference signal Vmod (modulator) with a high frequency triangu-
lar wave Vcar (carrier) and giving the resultant signal to the gate of the switches.
Two of the most common PWM techniques are bipolar and unipolar modulation.
2.4.7.1 Bipolar PWM
Bipolar PWM is commonly used in half bridge and occasionally full bridge con-
verters. In this method, a DC voltage with alternating sign is applied at the output
such that an average value at each carrier signal period is achieved. Figure 2.17(a)
shows the bipolar technique, in which Vr e f is Vmod and the triangular waveform
VC is the carrier. The following function is applied to obtain the switching com-
mand:
¨
1 if Vmod ≥ Vcar
S= (2.29)
0 if Vmod < Vcar
resulting in the output voltage equal to:
¨
+0.5Vdc if Vmod ≥ Vcar
VO = (2.30)
−0.5Vdc if Vmod < Vcar
2.4.7.2 Unipolar PWM

Unipolar PWM is only used in full bridge converters because it requires applica-
tion of zero output voltage. In this method, the output voltage alternates between
+Vd c and 0 during positive cycle of the modulator, while alternating between −Vdc
and 0 during negative cycle of the modulator. Figure 2.17(b) shows the unipolar
technique, in which V ∗ is Vmod and the triangular waveform is the carrier. The
following equations demonstrate this method:
• Case one (Vmod > 0):
¨
1 if Vmod ≥ Vcar
S1 = (2.31)
0 if Vmod < Vcar
S2 = 0 (2.32)
Then the output voltage would be:

¨
+Vdc if Vmod ≥ Vcar
VO = (2.33)
0 if Vmod < Vcar
• Case two (Vmod < 0):
S1 = 0 (2.34)
¨
1 if Vmod ≥ Vcar
S2 = (2.35)
0 if Vmod < Vcar
Then the output voltage would be:

¨
0 if Vmod ≥ Vcar
VO = (2.36)
−Vdc if Vmod < Vcar
For a correct modulation the following conditions must be at place:
ma < 1 (2.37)
mf 1 (2.38)
in which ma and m f are amplitude and frequency modulation indexes and are
define as:
Vmodmax
ma = (2.39)
Vcarmax
f car
mf = (2.40)
f mod
(2.41)
(a) (b)
Figure 2.17: (a) Bipolar, (b) Unipolar PWM
Figure 2.18: Three phase inverter with switching states
In case of bipolar modulation if m f is an odd integer, then the output voltage

will have an even symmetry. On the other hand, in unipolar modulation m f is
an even integer so that most of the harmonics present in bipolar modulation are
canceled in unipolar. In this project, SVPWM is used and its basics are explained
in the next subsection. The detials of the modulation technique is crucial for un-
derstanding the operation of the switches developed later on in the project.
2.4.7.3 Space Vector Pulse Width Modulation (SVPWM)
In the three phase inverters the SVPWM method is widely used which is on
the basis of the explained PWM methods. This method, in high frequencies gives
out almost sinusoidal voltages at the point of connection. Figure 2.18 illustrates
a three phase inverter. Each switch pair has a switching function, resulting in 8
possible scenarios. These functions are presented in Equation 2.42 to 2.44 .
¨
1 if S1 closed, S2 open
SA = (2.42)
0 if S1 open, S2 closed
¨
SB = (2.43)
¨
SC = (2.44)
Considering the switching functions, the operation of this circuit generates

three phase to phase and phase to neutral output voltage magnitudes as in Equa-
tion 2.45 to 2.50.
VAB = (SA − SB ) · VDC (2.45)

VBC = (SB − SC ) · VDC (2.46)
VCA = (SC − SA) · VDC (2.47)
2 1 1
VAN = ( SA − SB − SC ) · VDC (2.48)
3 3 3
−1 2 1
VBN = ( SA − SB − SC ) · VDC (2.49)
3 3 3
−1 1 2
VC N = ( SA − SB + SC ) · VDC (2.50)
3 3 3
According to this terminology, the voltages are calculated from the dc input as
shown in table 2.1 and 2.2. VL L0 to VL L7 are called voltage vectors. These vectros
are shown on α − β plane as in figure 2.19.
From the 8 possible switching states, first and last one are evidently not prac-
tical to produce voltage and the rest are used (Those two can be used for adding
dead time to the switching pattern). The six sections between each VL L vector
(forming a circle) is a possible location for the output voltage vector of the in-
verter (Although due to the nature of averaging, the actual area is a hexagon).
The voltage synthesizing can be implemented by a triangular wave as trigger, as-
suming that the desired voltage reference is available (from the control system
and after vector transformation).
The strategy is as followed: First determine the desired voltage vector is in

which one of the six sectors, in order to understand which voltage vectors (mean-
ing switches) need to be utilized. Then the weight (amplitude) of each vector can
be found via the desired voltage magnitude and the prior knowledge of the DC
link voltage. These weights are exactly the time that the corresponding switch
Table 2.1: Switching states and line voltages
SA , S B , S C VAB VBC VCA VL L

−
→ ~
000 0 0 0 V L L0 = 0
π
−
→ p j
100 V DC 0 -V DC V L L1 = 2VDC e 6
3π
−
→ p j
110 0 V DC -V DC V L L2 = 2VDC e 6
5π
−
→ p j
010 -V DC V DC 0 V L L3 = 2VDC e 6
7π
−
→ p j
011 -V DC 0 V DC V L L4 = 2VDC e 6
9π
−
→ p j
001 0 -V DC V DC V L L5 = 2VDC e 6
11π
−
→ p j
101 V DC -V DC 0 V L L6 = 2VDC e 6
−→ ~
111 0 0 0 V L L7 = 0
Table 2.2: Switching states and phase voltages
SA , S B , S C VAN VBN VC N VL L
−
→ ~
000 0 0 0 V L L0 = 0
s
−
→ 2
100 2/3 V DC -1/3 V DC -1/3 V DC V L L1 = VDC e j0
3
s 1π
−
→ 2 j
110 1/3 V DC 1/3 V DC -2/3 V DC V L L2 = VDC e 3
3
s 2π
−
→ 2 j
010 -1/3 V DC 2/3 V DC -1/3 V DC V L L3 = VDC e 3
3
s 3π
−
→ 2 j
011 -2/3 V DC 1/3 V DC 1/3 V DC V L L4 = VDC e 3
3
s 4π
−
→ 2 j
001 -1/3 V DC -1/3 V DC 2/3 V DC V L L5 = VDC e 3
3
s 5π
−
→ 2 j
101 2/3 V DC -2/3 V DC 1/3 V DC V L L6 = VDC e 3
3
−
→ ~
111 0 0 0 V L L7 = 0
Figure 2.19: Graphical representation of the voltage vectors in each state
1
must be left turned on in each period (T = ). As an example, consider that the
fsw
control system generates a modulating signal of Vr e f ∠α and α drops in the first
section between VL L1 and VL L . Therefore, projecting Vr e f on these vectors:
Vr e f ∠α = (τ1 × fsw ) · VL L1 + (τ3 × fsw ) · VL L3 (2.51)
τ1 and τ3 can be found accordingly.
The maximum obtainable voltage on the output happens when the resultant
voltage vector (from the summation of the two vectors on boundaries of the selec-
ted section with the calculated weight amplitude) is placed exactly in the middle
of the section with an angle of 30 degrees. Accordingly the voltage magnitude will
be:
1
VL L,ma x RM S = VDC × cos 30◦ × p (2.52)
2
The value of the output voltage is lower than the desired voltage (that was
given as modulation signal). This happens because when adding the weighted
voltage vectors, signals are averaged and in the process some of the voltage is
diminished. Another problem with SVPWM is physical limitations, which limits
the voltage and current drawn from the inverter. This issue can be tackled by
injecting third harmonic before the PWM to the modulation signal. Although the
third harmonic has to be canceled before entering the grid with a transformer.
Figure 2.20: Third harmonic injection (dotted lines are the resultant signals)
2.4.7.4 Third harmonic injection

As mentioned, other than the lowered voltage due to the averaging, the limit-
ations of design of components sets an even lower margin for the voltage output.
To mediate this issue the third harmonic injection strategy is used. Third har-
monic injection happens right before injecting the voltage reference modulator
signal generated from the control system. Basically, the strategy involves adding
another sine wave with three time the desired frequency. This also does not in-
terfere with possible motors in the model, since the motors are grounded with
a neutral point which oscillates with three times the fundamental frequency and
therefore the motor will not be affected.
The third harmonic injection is used in this project and it is generated from
equation 2.53. Figure 2.20 shows the effect of third harmonic injection. The peak
of the signal will be reduced and limited.
M a x(Vf ) − M in(Vf )
V3 f = (2.53)
2
Vmod = Vf − V3 f (2.54)
Next chapter provides a thorough examination of developed methods to provide

virtual inertia and their control algorithms while highlighting their advantages
and disadvantages.
Chapter 3
Existing Literature for Methods

to Emulate Virtual Inertia
3.1 Introduction
One pf the most common control methods of wind turbines is maximising the
output power generated. These control methods need to be designed with caution
in order to prevent any instability issues in the system. With the aim of large in-
tegration of RES into the electricity grid, certain adaptations are essential in order
to avoid stability issues. One of these issues as introduced in the previous chapter,
arises from the concept of inertia of the system or more specifically, reduction of
the inertia.
Multiple methods are introduced in the literature to emulate virtual inertia for
generators. Before the advancements of the power electronic technology, reserved
SG units at partial loads injected the kinetic energy in their rotating mass to the
grid [10] (in case of power deficit in the system) or as condenser units drew energy
from the grid (in case of power excess in the system) [26]. However, this method
proved to be costly inefficient and impose high capital and operation costs on the
grid operator [3]. Soon after expansion of the ESS technologies it was proposed
that a solution to the power imbalances in the grid could be the spread of the ESSs
in the grid in forms of battery, hydrogen storages and flywheels [27–29]. Never-
theless, the idea was not commercialised yet and similar to the reserved SG units,
it was not the most convenient solution. Moreover, other disadvantages such as
limit life-cycle,safety and noise, limited space and low round trip-efficiency of this
technology was another downturn for industrial application at that time.
While the ESS cost has decreased in the recent years, subsequently with the
advancements in the power electronic technology, new solutions emerged and
currently, the state of the art methods to provide virtual inertia, incorporate a set
of these solutions together. Some methods present a sophisticated modeling of
the exact inertial response of the SGs while others provide less accurate models
31
with faster and simpler implementations. Figure 3.1 provides a set of different
methods to provide virtual inertia. These methods operate on different principles
such as modeling the exact behavior of SGs, simulating frequency-power response,
swing equation. While most of these methods circulate on the basis of modeling
SGs, some tend to model other rotating masses such as induction machines. This
chapter investigates the various methods to provide virtual inertia for the grid.
Another categorization for the methods to provide virtual inertia was provided
by Cheema [30] and based on the need for an external storage unit and details of
modeling. This categorization is presented in figure 3.2.
3.2 SG Model Based Methods
3.2.1 Synchronverters
Synchronverters model the exact behaviour of the SGs from the grid point of
view with heavy numerical calculations. Synchronverters are voltage sourced grid
following converters and benefit from the inherent synchronisation mechanism.
Although the initial synchronverter models required a dedicated synchronisation
unit, such as PLLs to provide the phase and frequency of the grid voltage, the im-
proved versions are completely self-synchronised [31].
A synchronverter consists of two parts, first part is the power part which in-
cludes all the components in the link between dc and ac side, and the second part is
the electric part which includes control, sensing and protection circuits. Figure 3.3
and 3.4 present the control models developed by Zhong [24]. In figure 3.3, Dp ,
1/Js, 1/s and the blue highlighted block represent the Virtual Synchronous Mo-
tor (VSM) and implements the torque, active and reactive power calculations of
a SG (similar but not identical to equations 2.19, 2.20 and 2.22 for a DFIG in sec-
tion 2.3.3) and additionally, generates the signal for converter modulation (this
signal would be the same as the voltage at the terminal of the virtual SG). J is the
generator moment of inertia, Te is electromagnetic torque (Tem ), m f magnitude of
the mutual inductance between stator and field coils, θ is the angle between rotor
axis and stator winding phase, e, P, Q are no-load generated voltage, reactive and
active output powers, respectively. The emf (e) calculated form the VSM equa-
tions enters a PWM block to produce the driving signals for the converter. The
output current of the converter that flows into the inductors are considered as
stator current (i) and fed back to the VSM model. Meanwhile, the frequency and
voltage must be regulated via controlling the active and reactive power, respect-
ively. The proposed topology satisfies these conditions as the mechanical friction
coefficient (Dp ) can play the role of frequency droop coefficient, eliminating the
need for an additional control loop by regulating the frequency/speed (θ̇ ) of the
synchronverter and generating the phase angle θ for the emf. The field excitation
Chapter 3: Existing Literature for Methods to Emulate Virtual Inertia 33
Figure 3.1: Methods to provide virtual inertia

Figure 3.2: Methods to provide virtual inertia [30]
current M f .i f is generated from the reactive power and a voltage droop control is
added that controls the voltage with the voltage droop coefficient Dq . Satisfying
all the conditions of the normal operation, the PLL is added to synchronise the
converters with the grid before connection.
The upgraded version as shown in figure 3.4 removes the PLL and virtual cur-
rent (is ) is generated from the voltage difference of e and Vg and the current
injected to the VSM can be either grid current (i g ) or is . Additionally, a PI control-
ler regulates the output ∆T to be zero and generates the reference frequency θ̇
for the synchronverter. Moreover, the addition of the switches Sc , S P and SQ en-
ables different modes of operation for the controller. SC in position 1 while SQ is
off and S P is on, allows the reference real and reactive power to be sent to the
grid. If active and reactive power references (Pset , Q set ) are zero simultaneously,
the synchronverter is self-synchronised. If is is zero, the voltage is synced. If SC
is set to position 2, other modes of operation are enabled. After the connection
has been made, if S P is in ON state, ∆T will be set zero by the PI controller. As a
result, electromagnetic and mechanical torques will be equal and P = Pset . This
mode is named the set mode. Active and reactive power have their own set modes
called the P and Q mode. If S P is OFF, the PI controller is bypassed and synchron-
verter is in frequency droop mode, called PD mode. Correspondingly, the voltage
droop mode is Q D mode. Table 3.1 provides all the operational modes of the self-
synchronised synchronverter.
The advantage of this model is the fact that the controller implements the
exact natural inertial response of a SG during a contingency event. Additionally,
the Voltage source implementation of this topology saves the topology from the
grid transient currents, not to mention that this can be used as a grid forming
unit to emulate virtual inertia. Moreover, the synchronverter technology is not
dependant on the parameters of the wind turbine generator itself. Nevertheless,
Table 3.1: The operation modes of the self-synchronised synchronverter
Switch SC Switch S P Switch SQ Mode

1 ON ON N/A
1 ON OFF Self-synced
1 OFF ON N/A
1 OFF OFF N/A
2 ON ON P mode, Q D mode
2 ON OFF P mode, Q mode
2 OFF ON PD mode, Q D mode
2 OFF OFF PD mode, Q mode
the complexity of the model and numerical calculations is a disadvantage for this
model.
Figure 3.3: Synchronverter control model
3.2.2 Institute of Electrical Power Engineering (IEPE) Lab Topology

The Institute of Electrical Power Engineering (IEPE) lab topology named Virtual
Synchronous Machine (VISMA), was introduced by Hesse [32] in 2009 and de-
veloped further on 2011 [33]. Figure 3.5 illustrates this model in general format.
Similar to synchronverters, the model represents a SG while a notable difference
between the two models is current source based topology of the VISMA, unlike
synchronverter. A current source based converter essentially has the disadvantage
of injecting current harmonics and unbalances of the grid into the control system.
In order to alleviate these disturbances, a wideband compensation method is used.
Figures 3.5 and 3.6 illustrate the VISMA model and its detailed subsystems in se-
quence. This topology, although more complex, has no evident advantages over
the synchronverter topology.
Figure 3.4: Self-synchronised synchronverter model
The compensation method is implemented by a Digital Signal Processing (DSP)

unit and a hysteresis controller. First the PLL unit generates a normalized 2π angle
ramp synced with the first phase of the grid. The integrated sum of the center fre-
quency and controller results in an angle ramp. A modulo limiter is considered to
avoid overflow. Due to the ramp shape of the signal, a sine function is required in
the feedback signal for comparison. The resulting synced ramp signal then passes
the phase generation subsystem to build the desired phase voltages of the grid. The
amplitude of the voltage is then formed in the amplitude generation subsystem.
Then the noise component is removed by subtracting the measured phase voltages
and the synthesized values. The noise component is inverted and weighted with
the Distortion Component factor (DCF) and in the end, fed to the hysteresis con-
troller. This method is quite known in the noise cancellation methods and the
imperative condition of correct operation of the method is a fast acting hysteresis
controller.
The updated control model (IEPE lab topology) that produces the reference
current signal for the hysteresis block is presented in figure 3.7. The control system
in this topology is quite similar to the synchronverter topology. In this model,
the output of the power output of the converter can be controlled by adjusting
the value of mechanical power and the electromotive force (emf ) sets the grid
voltage. The upgraded model is more flexible since the inertia constant and the
Figure 3.5: VISMA model
damping force can be adjusted easily in the SG model. In this model, the grid
current is measured and the voltage reference value is calculated from the stator
circuit analysis:
d i1
e1 − u1 = i1 · Rs + Ls · (3.1)
dt
d i2
e2 − u2 = i2 · Rs + Ls · (3.2)
dt
d i3
e3 − u3 = i3 · Rs + Ls · (3.3)
dt
d ~i r e f
~ g r id = ~i r e f · Rs + Ls ·
~e − u (3.4)
dt
T T
where ~e = e1 e2 e3 is the induced emf in the stator winding, u
~ g r id = u1 u2 u3
is the grid voltages at the Point of Common Coupling (PCC), Rs and Ls are the
stator resistance and inductance respectively. Then the reference current can be
calculated as:
~i r e f (s) = (~e(s) − u
~ g r id (s))/(Rs + Ls · s) (3.5)
Finally, the interaction between rotor and stator can be modeled by the elec-
tromechanical power balance equation:
(a) (b)
(c) (d)
(e) (f)
Figure 3.6: VISMA (a) hysteresis controller, (b) PLL, (c) phase generation,
(d) amplitude generation, (e) grid synthesizing, (f) Distortion compensation
subsystems
1 dω dω
Tm − Te = · + kd · f (s) · (3.6)
J dt dt
Pe
Te = (3.7)
Z ω
θ= ω · dt (3.8)
in which, J is the moment of inertia, kd is the mechanical damping factor, f (s)

is the phase compensation function, ω is the angular speed of rotation, θ is the
angular position of the rotor, Te and Tm are the electrical and mechanical torque.
The phase compensation term guarantees that damping force counteracts the pos-
sible oscillations of the rotor in the opposite phase. As mentioned previously, the
induced emf can be adjusted and set to grid voltage:
 
sin θ
 
e1
 
 2 
~e = e2  = E p · sin θ − · π (3.9)
 
3
e3
 
 2 
sin θ + · π
3
Figure 3.7: Updated IEPE Controller topology for SG
3.2.3 Kawasaki Heavy Industries (KHI) Lab’s Topology

The Kawasaki Heavy Industries (KHI) topology is a Virtual Synchronous Gen-
erator (VSG) model that works on the basis of phasor diagrams of the SGs [34]. In
this model the entire SG is not implemented and instead the equivalent Automatic
Voltage Regulator (AVR) and governor are modeling the dynamics of the SG [35].
Figure 3.8 depicts the KHI topology [30]. The model is a current controlled based
converter in d-q axis coordinates. The d axis is placed along the phase voltage of
the generator and the q axis leads by 90◦ . E F , Vg , Vd , Vq denote the internal emf
and terminal voltage of the generator and the d,q axis voltages , respectively. Ad-
ditionally, θ , x, r are the phase angle, reactance of the generator and armature
resistance. The armature current in the d-q axis can be then calculated from the
following equations:
∗
Id
= Y Ed − Eq − Vd Vq (3.10)
Iq∗

1 r x Ydd Ydq
Y= 2 = (3.11)
r + x 2 −x r −Ydq Ydd

Ed cos θ
=| E f | (3.12)
Eq sin θ
This model consists of both governer and AVR and utilises a PLL unit for syn-
chronisation, explained:
Figure 3.8: KHI SG topology
• Governer Model
The governer model shown in figure 3.10 determines the angular speed
of the rotor. The required angular speed ωR is found by finding difference
between the active power reference and its actual value. A first order delay
is used and the rated angular speed is compared with the output of the delay
subsystem.
• AVR Model
A quadrature-current compensation is used for determining the internal emf
as figure 3.10 illustrates. In real generators, the AVR and time constant of
the field winding determine the gain of the voltage feedback. The propor-
tional integral used does not have any phase delays or gain drops in the
high frequencies.
• PLL Model
The PLL used in this topology is presented in figure 3.10 and variables θ and
ω denote the angle and angular velocity of the rotor. The error between ω
and θ is find via the following equation:
2 X
sin (ϕ − nπ/3) cos (θ − nπ/3) =
3 n=0,1,2 (3.13)
sin (ϕ − θ ) ' (ϕ − θ ) if |ϕ − θ | 1
As the equation shows, the angular error is the inner product of the output
voltage of the inverter for each phase divided by peak values and a three
phase vector with a phase lead of π/2. The angular error only occurs on the
d-axis voltage component. Figure 3.9 presents this fact.
Figure 3.9: Angular deviation in the KHI PLL
The angular velocity ω can be find via the angular error:
ϕ − θ ' Vq /Vg (3.14)

K P I Vq
ω = (K P P + ) (3.15)
s Vg
3.3 Swing Equation Based Methods
3.3.1 Ise Lab’s Topology

Ise topology utilises the swing equation 2.10 as the heart of the VSG model and
was developed by the Ise Lab [18]. In each cycle, the swing equation is solved in
VSG control block and the frequency ω is calculated as shown in figure 3.11.
Then it is passed to an integrator to produce the phase angle, θ to feed the PWM
generator[36]. The complexity of this model is quite less than models operating
on the basis of modeling SG.
(a)
(b)
(c)
Figure 3.10: KHI topology (a) Governer, (b) AVR and (c) PLL circuits
Figure 3.11: Ise topology

3.3.2 Synchronous Power Controller (SPC) Topology
Another swing equation based model is the Synchronous power Controller

(SPC) model, developed by Rodriguez [37] for the control of PV generation sys-
tems. In this model, the converter is implemented as a cascaded control system.
The model has a second order transfer function as shown in figure 3.12 to improve
the damping and increase inertia of the system simultaneously. This method mod-
els the electrical admittance of the VSG enabling the control of the system without
the need for additional filter. On the other hand, this robustness of the controller
comes with the price of more difficulty in tuning the controller.
As mentioned, this model does not aim to mimic the behaviour of the SG
but it models a second order overdamped response that in case of perturbations
provides the satisfactory damping. The electromechanical transfer function of the
SPC is developed as below:
PM ax
Pel ec J · ωs
= (3.16)
Pinp k PM ax
s2 + s+
J · ωs J · ωs
where PM a x , Pel ec , Pinp , J, k and ωs represent the maximum value of the active
power delivered to the SG, the actual delivered power, the input power, the inertia,
the damping factor and the synchronous frequency of the generator respectively.
Figure 3.12: SPC topology

3.4 Frequency Power Based Methods

3.4.1 VSYNC Topology
The VSYNC research group introduced a topology that comprised a VSG model,
an energy storage unit as source of power and an LCL grid filter [38]. The Eigen-
frequency of the filter is positioned about halfway between the nominal power
frequency and converter switching frequency. This model is current based and
current reference is provided by the PLL circuit. Note that in this model, the PLL
is not used as a synchronisation unit but rather as a tool to provide the current
reference. The current reference is then used to drive the controller.
Figure 3.13 presents the VSYNC model and its current control block. The cur-
rent reference block produces the error current signal (idq ). KSOC is assigned in
such way that when the change in the State of Charge (SOC) of the ESS is at its
maximum. the active signal (P) would be equal to the nominal VSG output power.
Additionally, K v is set accordingly to achieve the maximum reactive power output
from the VSG for a specific voltage deviation. The frequency is estimated based
on the zero-crossing methodand the set point for current is calculated from the
following equation:
d∆ω
KI + K P ∆ω
dt
Isp = (3.17)
VDC
d∆ω
where is the ROCOF, K P is expressed in k g m2 /s2 and K I is a dimen-
dt
sionless factor.
3.4.2 Virtual SG
Virtual SG was developed on the basis of modeling the frequency and power
response [39] of a SG. The model is quite similar to the VSYNC topology. However,
a dynamic frequency regulation is provided, enabling dispatch of the SG. The
dynamics of the system can be presented as:
d∆ω
PSV G = K D ∆ω + Kl (3.18)
dt
where PSV G is the output power, K D and Kl are the damping and inertial con-
stants, respectively. The damping constant is similar to frequency droop, helping
to reduce the frequency nadir, while the inertial constant provides fast dynamic
frequency based on changes in the frequency. This characteristic is specially use-
ful in the islanded microgrids where ROCOF can be extremely high. As shown in
figure 3.14, the PLL measures the changes in the frequency and ROCOF. Then the
active power is calculated using Eq. 3.18 and the current reference of of the con-
troller is given by Eq. 3.19 and finally, the current controller produces the signals
for the PWM generator.
Figure 3.13: VSYNC topology and its reference current block
2 Vd PS V G − Vq Q
I d∗ = (3.19)
3 Vd2 + Vq2
Figure 3.14: VSG topology
Although this topology has some advantages for the islanded mode, it can not
be used as a grid forming for microgrids. Additionally, the system only provides
inertial response to the frequency variations and not input power variations. Fur-
thermore, the derivative term for ROCOF makes the system prone to noise and
instability.
3.5 Uncategorized Methods
3.5.1 Droop Based Approach

The droop based method is mainly applied to autonomous operation of isolated
microgrid systems [40, 41]. Assuming an inductive grid impedance, the frequency
droop of the grid can be calculated from Eq. 3.20 [4]:
ω g = ω∗ − m p (Pout − Pin ) (3.20)
where ω∗ is the frequency of the local grid and Pin , Pout are reference and
measured active power. The model developed from Eq. 3.20 is presented in fig-
ure 3.15. In this model, high-frequency components of the inverter are filtered out
by a low-pass filter. The filter also introduces a delay in the measured value which
can be interpreted as virtual inertia and although elementary, this model is quite
usfeful as mentioned before. According to the developed model in figure 3.15, the
frequency droop equation can be rewritten:
1 1
Pin − Pout = (ω∗ − ω g ) + T f sω∗ (3.21)
mp mp
The disadvantage of this model is extremely slow transient response and the
fact that the grid is not always inductive. However, adding a virtual impedance
can alleviate this issue [42].
Figure 3.15: Droop-based topology
3.5.2 Rate Limiter Approach

The rate limiter approach was introduced by Zhang [43] in 2016. Similar to
droop based approach, this model is not built on the idea of a complex modeling
for controller but instead designing a simple, yet efficient controller to provide
virtual inertia. This model utilizes a low-pass filter in its core. Figure 3.16 shows
this model. In the model, the signal resulted from comparison of the nominal
frequency and the power error is passed through a rate limiter function:
u(i) − y(i − 1)
r= (3.22)
t(i) − t(i − 1)
where u(i), t(i) are the current input and actual time step of the limiter block
while y(i − 1), t(i − 1) are the output of the block in the previous time step and
the previous time step. A non-linear function is then defined for the output of the
block which is the reference frequency f r :

∆t · R + y(i − 1)
 if r > R
y(t) = ∆t · F + y(i − 1) if r < R (3.23)

u(i) otherwise
in which R and F are rising slew rate and falling slew rate, respectively.
Figure 3.16: Rate limiter topology
3.5.3 Other methods

Methods to provide virtual inertia are abundant and a couple of other import-
ant topologies is introduced in this section. The VOC topology models a non-linear
oscillator and is capable of maintaining synchronisation without any communica-
tion [44]. This topology although quite complex, is highly practical for grids with
numerous Distributed Generation (DG)s.
Inducverters are one of the relatively new approaches and operate on the basis
of mimicking the behaviour of induction machines instead of SG. An advantage
of this model is the auto-synchronised PLL-less topology [45]. As figure 3.17 il-
lustrates the model consists of a complex and detailed modeling of the induction
machine, a block representing inertial response and an adaptive virtual imped-
ance. The detailed model of the induction machine is out of scope of this thesis
and is neglected.
Figure 3.17: Inducverter control topology

Chapter 4
Proposed Methodology and Case

Studies
4.1 Comparison Between the Existing Methods

The SG based models (synchronverter, IEPE, KHI and so on) are to some ex-
tent or entirely modeling the dynamics of the SG and most of these models re-
quire a PLL unit for synchronization with the grid. Although a derivative of the
frequency is not required for these models, they have no over-current protection
and this could cause numerical instability. On the other hand, the swing equation
based models (ISE, SPC and so on) are simpler than SG based models. Similar
to the previous category, these models do not require a frequency derivation, use
PLL and have no over current protection. Additionally, power and frequency os-
cillations could be a problem in these models. The frequency power based model
(VSYNC, virtual SG and so on) are mostly current source implementation and nat-
urally have over current protection while the previous models were mainly voltage
source based models. However, these models require frequency derivations, which
can inject noise and cause instability problems in the system regardless of the pos-
sible instability problems of the PLL. The droop-based and rate limiter models are
communication-less models that provide a slow transient response to contingency
events naturally. Some of the novel models tend to implement the dynamics of in-
duction machines or linear oscillators.
4.2 Selected Approach to Emulate Inertia

After analysing the literature of methods to provide virtual inertia, the aim was
set to define a strategy that provides a response similar to the existing methods,
while having the least possible complexity. Before selecting the method to provide
virtual inertia in case of power deficit, the source of energy required for injecting
additional energy to the grid of the platform must be determined. Since the case
49
scenarios defined later in this chapter are based on power deficit, this energy must
be extracted form either (1) the available elements in the grid or (2) an external
element. In the literature of case (1), as explained in the previous chapter, dif-
ferent methods suggested extracting energy from (i) the wind turbine or (ii) the
capacitor in the DC link. In this project, the wind turbine is considered to run at
nominal rotational speed and therefore, braking the turbine to extract the kinetic
energy in the rotating mass of blades and generator in case of power deficit is not
an option, not to mention the changes required in the converter controllers. Also,
the DC link capacitor between the converters is practically not sufficiently large
enough to provide the required energy for virtual inertia emulation and it has to
carefully integrated in the control algorithm which increases the complexity of
the system and can cause instability issues in voltage and power on its own. Fur-
thermore, the gas turbines in the grid of the platform are only present to provide
part of the loads that can not be supplied via the wind turbines and note that the
ultimate aim of the project is to reduce the utilization of carbon emitting tech-
nologies such as gas turbines, which alternatively translate to limited use of gas
turbines and investigating a solution in which the gas turbines do not participate
in virtual inertia emulation. The analysis provided due to the characteristics of the
understudy platform grid leads to one option, which is the addition of an external
unit. Following the explanations in section 3.1 regarding addition of an external
unit (case (2)) in a system, the installation and operation costs of the ESSs have
decreased in a sense that they proved to be a feasible solution for the power defi-
cit problem in near future. For all the reasons explained, a separate ESS unit was
considered to be the provider of the energy required for emulation of the virtual
inertia.
Figure 4.1: Project study case concept

Chapter 4: Proposed Methodology and Case Studies 51
The concept of the oil and gas platform is illustrated in figure 4.1. This
concept is built from scratch and to make the project simpler, each component is
designed separately, tested and then assembled to other components. In the end,
the entire system is evaluated. The ESS is connected via a converter to the grid of
the platform and the virtual inertia algorithm is embedded in the converter con-
trol. The complete model of the virtual inertia algorithm and the converter control
with ESS unit, detailed model of the wind turbine and its converters, and the gas
turbine are provided in the next chapter. Next section explains the cases defined
during the project to build the concept model. Cases a, b, c are simply the pro-
gressions of the project in each step to built the final model, while cases d, e, f and
g (actual case studies of the project) examine the effectiveness of virtual inertia
algorithm during a power deficit scenario. The progressions were an essential part
of the project analysis in order to better understand the dynamics of the platform
and choosing the most suitable methods to model the the system. The progres-
sions determined whether a choice was made correctly and in scope of the project
or not and how each component must be altered to best achieve the final goal of
the project which is emulating virtual inertia. After these progressions were made
and satisfying results from the of the project model until that point were reached,
the case studies were defined to test and analyse the effectiveness of the virtual
inertia provision method.
4.3 Case studies

4.3.1 Case a
This case models the wind turbine and its drive train with pitch controller
(wind turbine package). The aim is to build a wind turbine model that provides a
constant torque for the input of the wind turbine generator. This case is defined
for variable wind speeds and an optimum pitch angle for the wind turbine.
4.3.2 Case b
After modeling the wind turbine in the previous case, the DFIG is modeled in
this section considering two different models for wind turbine (first model from
case a, and second model from simplified wind turbine model of case a). Two
control strategies for the Rotor Side Converter (RSC) are considered and tested,
named direct speed control with PI controller and indirect control of speed for
Maximum Power Point Tracking (MPPT) system. The RSC on the DC side is con-
nected to a constant voltage source for testing the circuit.
4.3.3 Case c
The combination of the two control strategies and models in the previous case
are tested in this case with a connection to the grid with GSC, with presence of
DC link capacitor and grid side filters. The most stable design is chosen as the
representation for the wind farm in the concept model of the platform and it
delivers the base load of the platform.
4.3.4 Case d
The developed wind farm model of the case c is connected to a gas turbine. The
gas turbine is responsible for maintaining the grid and keeping the voltage and
frequency constant while the wind farm supplies the base load of the platform.
The gas turbine also supplies a small portion of the platform loads (certain loads
that can not be supplied from the wind farm and need gas turbine). A step change
in the load is considered to analyse the effect on frequency when ESS is not con-
nected and virtual inertia algorithm is not present. At his point the grid platform is
completely built and the frequency deviations should be evident in the simulation
results. The next cases are expected to alleviate the frequency deviations of this
case.
4.3.5 Case e
The virtual inertia is implemented from this case forward. In addition to the
model in previous case, the ESS unit is modeled and the virtual inertia emulation
algorithm is implemented on it in the form of a frequency droop control with a P
controller. The effectiveness of the virtual inertia emulation method is analysed.
4.3.6 Case f
The same scenario as in case e but with a PI controller is designed and tested.
4.3.7 Case g
The same scenario as in case f but with an added dead band to the frequency
in the ESS unit controller.
Chapter 5
Modeling
5.1 Introduction
The modeling of complex and detailed projects could be extremely confusing
if not approached properly. In this study, in order to have a better analysing cap-
ability, the model is dismantled into different sections and then put back together
in the end. In this chapter, the detail modeling and methodologies are explained.
Different sections that are modeled separately are as followed:
• platform grid
• Wind turbines and their converters
• Virtual inertia provision unit
5.2 Platform grid

5.2.1 Loads
As mentioned in chapter 2, the detailed modeling of the platform grid is not
essential for designing a virtual inertia algorithm. Therefore, the platform model
was implemented with less detail and as a constant load. Nevertheless, some vari-
ations are considered in the load to analyse the efficiency of the virtual inertia
algorithm. Based on the model presented in Annex 7.2, the total value of the
loads in the platform are estimated around 45 MW. The parameters of the sim-
ulation of the electrical grid of the LEOGO platform are presented in Annex 7.2.
The value of the load itself does not necessarily impact the modeling of the virtual
inertia algorithm and therefore, the total loads of the platform are scaled down
from 45 MW to 2 MW. The reason for scaling down is considering less capacity for
the wind farm and less number of turbines. An additional 5 kW is considered for
the loads that can not be supplied through the wind energy but supplied through
the gas turbine. The model of the loads are considered as constant power loads.
53
5.2.2 Gas turbine

The platform also relies on three gas turbines and back up diesel generator that
initially supplied the 45 MW loads of the platform. In the modeling of the system
only one gas turbine is considered and the diesel generator is replaced by the ESS
that provides the virtual inertia. To model the exact models, the wind turbines
and gas turbines can be replicated (check appendix).
One gas turbine is modeled to support a portion of the loads that can not be
supplied through wind. In addition, it helps to keep the voltage of the grid con-
stant and contributes to maintaining the frequency at the nominal value. Table 5.1
provides the values for the salient pole gas turbine and the platform grid. The gas
turbine is modeled as a SG with an AVR and governor.
Table 5.1: Gas turbine generator data
Parameter Value Description

fn 50 [Hz] Nominal grid frequency
Vn 690 [V] Nominal grid voltage (LLRM S )
Pl oad 2.005 [MW] Total loads on the platform
Pn gas 10 [kW] Nominal power of the gas turbine
Xd 2.24 [pu] Steady state generator reactance (d-axis)
Xd 0 0.17 [pu] Transient generator reactance (d-axis)
Xd ” 0.12 [pu] Sub-transient generator reactance (d-axis)
Xq 1.02 [pu] Steady state generator reactance (q-axis)
Xq 0 0.13 [pu] Transient generator reactance (q-axis)
Xq” 0.08 [pu] Sub-transient generator reactance (q-axis)
Td 0 0.012 [s] Transient time (d-axis)
Td ” 0.003 [s] Sub-transient time (d-axis)
Tq ” 0.003 [s] Sub-transient time (q-axis)
Rs 0.037875 [pu] Stator resistance
H 1.028 [s] Inertia coefficient
F 0.02056 [pu] Friction factor
P 2 [-] Pole pairs
dw0 0 [%] Initial speed deviation (% of nominal speed)
th0 0 [deg] Initial angle of the rotor
ia0 1 [pu] phase a initial current magnitude
ib0 1 [pu] phase b initial current magnitude
ic0 1 [pu] phase c initial current magnitude
pha0 0 [deg] phase a initial phase
phb0 -120 [deg] phase b initial phase
phc0 120 [deg] phase c initial phase
Vf0 1 [pu] Initial field voltage
Chapter 5: Modeling 55
5.2.2.1 AVR
AVR plays an important role in generation and in this project, it is used for
generating the field voltage and maintaining the output voltage of the gas turbine.
A typical AVR consists of five components [46]:
• Amplifier: amplifies the input reference voltage
• Exciter: generates the field voltage from the reference signal
• Generator: relation between the field voltage and the gas turbine voltage
terminal
• Sensor: feedback for measuring the output voltage and comparing to the
reference signal
• Additional controllers: for controlling the plant model of the first three com-
ponents
Figure 5.1 illustrates the block diagram of an general AVR and its components.
The first four components are usually modeled with a first order transfer function
and the controller is typically a PID controller, while other control strategies can
also be implemented.
Figure 5.1: A typcial AVR model
AVR system can be extremely complicated but in this project to avoid com-
plications, the IEEE type one topology for the AVR is used. The reference signal
Vr e f is given to the AVR along the stator voltage in d-q axis and zero input for sta-
bilizer port (stabilizer port provides additional stabilization of the power system
oscillations, which is neglected in the study). Table 5.2 presents the values used
in the modeling of the AVR for the gas turbine.
5.2.2.2 Governor
Governors are used for controlling the rotational speed of generators and de-
pending on the application and type of the generator, they will have different con-
trol systems. The generator speed is coupled with grid frequency and therefore,
Table 5.2: Gas turbine AVR data

Tr 0.02 [s] Low pass time constant
Ka 300 [-] Regulator gain
Ta 0.001 [s] Regulator time constant
Ke 1 [-] Exciter gain
Te 0 [s] Exciter time constant
Tb 0 [s] Transient gain reduction
Tc 0 [s] Transient gain reduction
Kf 0.001 [-] Damping filter gain
Tf 0.1 [s] Damping filter time constant
Efmin -11.5 [pu] Regulator output limit (min voltage)
Efma x 11.5 [pu] Regulator output limit (max voltage)
Kp 0 [-] Regulator output gain
Et0 1 [pu] Initial value of terminal voltage
Vf0 1.28 [pu] initial value of field voltage
Vst a b 0 [pu] Stabilization reference signal
Vr e f 1 [pu] The desired stator terminal voltage
the governor is actually controlling the frequency. Some of the old and obsolete
governor models are GAST, GAST2, GASTWD, GFT8WD, WESGOV and they are
replaced with newer version such as GGOV1. For this project, a separate governor
was designed with less complexity than the standard GGOV1 model.
A separate simple governor is designed modeling turbine delay and a droop

control for turbine speed. Table 5.3 presents the values used in the modeling of
the governor for the gas turbine. Figure 5.2 illustrates the schematic of the gov-
ernor built for the gas turbine. The
Table 5.3: Gas turbine governor data

K g ov 10 [-] Governor delay gain
D g ov 0.05 [-] Governor droop gain
Pml im 1 [pu] Maximum output of the governor
Pm r e f 0.5 [pu] Reference gas power
ωr e f 1 [pu] Reference generator speed
Figure 5.2: Governor model
5.3 Wind turbine and related converters

The wind turbine and its related converters consist of the following components
ad they are tested separately first and assembled together later:
• Wind model
• Wind turbine, drive train and pitch controller
• Simplified wind turbine model
• DFIG
• RSC
• GSC
Table 5.4 contains parameters for modeling the wind profile, wind turbine,
drive train and pitch controller and it has been continuously referenced during
this chapter. Note that wind turbine is also scaled down according to the load and
the nominal power of the wind turbine is 2.4 MW, while in practice, the offshore
wind turbines have much higher power ratings. Since the idea for the wind farm
is to only participate in supplying the base load and not getting involved in virtual
inertia emulation, the size of 2.4 MW was selected.
5.3.1 Wind model

The aim of this section is to explain the wind model used in this project. The
data for wind (average value, based value, cut-in, cut-out and turbulence of wind)
are presented in table 5.4. The model that has been utilized in this study is the
model developed by RISO National Laboratory of DTU based on Kaimal spec-
trum [47]. This model was accessed from an ongoing research, thanks to Daniel
Mota- Ph. D. candidate at NTNU. The wind model considers both rotational turbu-
lence and tower shadowing effect to on the wind speed. The model uses a built-in
band-limited white noise generator that is explained in this section. As figure 5.3
illustrates, the inputs to the wind model are the rotor diameter (in this case radius
of the rotor, extracted from equation 5.11), the length scale factor (L considered
Table 5.4: Wind turbine data

Vwind 12 [m/s] Average Speed of the wind and
Base wind speed of the wind turbine
Vcut in 3 [m/s] Cut in wind speed
Vcut out 20 [m/s] Cut out wind speed
Tblwind 6 [%] Turbulence intensity of wind
ρai r 1.225 [kg/m3 ] Air density
βma x 45 [deg] Maximum pitch angle
βopt 0 [deg] Optimal pitch angle
kβ 500 [-] Proportional gain of the pitch controller
rβ ±2[−] Rate of change of pitch angle
Hwt 4 [s] Wind turbine inertia constant
SS 0.3 [-] Shaft stiffness of drive train
DC 1 [-] Mutual damping coefficient of drive train
N 1 [-] Gear ratio
λopt 8.1 [-] optimal tip speed ratio of the wind turbine
C popt 0.48 [-] optimal performance coefficient of the wind turbine
C1 0.5176 [-] Performance function coefficient
C2 116 [-] Performance function coefficient
Pnwt 2 [MW] Nominal power of the wind turbine
P bwt ≈ Pnwt [kW ] Base power of the electrical generator (PnDF I G )
Rwt 42 [m] Wind turbine radius
W base 155 [rad/s] Base rotational speed
W bpu 1 [pu] Base rotational speed of the wind turbine
(pu of base generator speed)
Pma x Vwind 1 [-] Maximum power at base wind speed (pu of Pnwt )
equal to 0.2R wt ) which is used in calculating shadowing effect and finally, turbu-
lence intensity. The model generates a random time series of wind speed with the
average value of Vwind .
Figure 5.3: Wind speed model
Generally, wind speed can be defined with four components, named the aver-
age, gust, ramp and turbulence wind speeds [48]. Out of these four components
only the turbulence is not given analytically and is yet the most challenging part to
model. Researchers attempt to model wind turbulence with the assist of stochastic
processes. One of these methods is the Spectral Representation Method (SRM),
developed by Shinozuka and Jan [49]. SRM produces quite realistic samples ac-
cording to Power Spectral Density (PSD). The presented wind model in this pro-
ject, applies SRM to Kaimal PSD function [50] as shown in figure 5.4. The original
Kaimal PSD is defined as:
[ln (h/z0 )2 ]−1 · l · .Vwind

s( f ) = (5.1)
(1 + 1.5( f · l/Vwind ))5/3
in which f , h, Vwind , l, z0 are frequency [Hz], height of the wind hub [m], the
average wind speed [m/s], turbulence length scale [m] and roughness length [m].
Equation 5.1 can be simplified as followed:
c1
s( f ) = (5.2)
(1 + c2 | ω |)5/3
The model presented in this project, inspired from RISO model and based on
Kaimal PSD function, consists of certain constants and function obtained from
Equation 5.1. The function in figure 5.4 is:
p
T bl wind · L · Vwind
f (u) = p (5.3)
100 2
and the coefficients k, BD, C D, E D and F D in figures 5.5, 5.6 can be calculated
as:
k = 0.0182/1.3463 (5.4)
1.3653
BD = (5.5)
L
1.3463( )
2πVwind
0.9846
CD = (5.6)
L
1.3463( )2
2πVwind
3.7593
ED = (5.7)
L
1.3463( )
2πVwind
1
FD = (5.8)
L
1.3463( )2
2πVwind
Figure 5.4: kaimal model for wind
The sample generated wind profile of the site was achieved with an average
value of 12 m/s, presented in figure 5.7.
5.3.2 Wind turbine, drive train and pitch controller (Wind turbine
package)
5.3.2.1 Wind turbine
The wind turbine chosen for study is a three bladed variable speed horizontal
axis wind turbine with a two mass drive train. Figure 5.8 illustrates the model of
the wind turbine, drive train and pitch controller and their connections referred
to from now on as wind turbine package. In this section, the components of fig-
ure 5.8 are explained. The results of the simulation of this model are presented
in chapter 6. The gains k1 and k2, are the nominal rotational speed and torque
of the DFIG. The wind turbine package is implemented in pu but the inputs and
Figure 5.5: kaimal filter
Figure 5.6: Kaimal coefficients

Figure 5.7: Wind speed profile of the site
outputs of the DFIG must be in physical units according to the Matlab model.
DFIG model has been explained later on in this chapter and the data of the DFIG
necessary to calculate k1 and k2 can be found in table 5.6. Note that since DFIG
is operating in generator mode, the input torque (Tmin ) has to be negative, while
the wind turbine package, produces a positive torque reference (Tm the output of
the wind turbine is also negative, but the drive train model operates with positive
values of input and output torque).
Figure 5.8: Wind turbine, drive train and pitch controller (wind turbine pack-
age)
k1 = ω r base (5.9)
PnDF I G
k2 = − (5.10)
ω r base
The same wind turbine block of the Matlab was rebuilt for this project in order
to have access to the parameters in the model such as C P and λ that are not access-
ible in the wind turbine block of Matlab. Matlab description of the wind turbine
model: "The block implements a variable pitch wind turbine and the performance
coefficient (C P ) is implemented internally as the mechanical output power of the
turbine over wind power. C P is a function of wind speed, rotational speed and pitch
angle (β) with maximum value at β = 0". The model parameters are depicted in
figure 5.9 and the values assigned to them are provided in table 5.4. However,
as mentioned in the description by Matlab, this model does not include the drive
train. Hence, a separate drive train was developed according to the model in [51].
The pitch control is also a simple proportional gain based on the same reference.
According to the values assigned, the turbine characteristic curve can be plot-
ted for different wind speeds. Figure 5.10 is sketched from the default values of
the Matlab model as in table 5.5, while figure 5.11 is sketched from the data of
the rebuilt model provided in table 5.4. According to the set values for the turbine
parameters, both wind turbine and DFIG have the same nominal speed and out-
put mechanical power. Although practically, the values are not the same and the
default values set by Matlab are preferred, for simplicity and convenience these
values have been set the same. However, this assumption does not affect the per-
formance of the developed model and it could be operated with unequal value for
the mentioned parameters.
Figure 5.9: Wind turbine block parameters
Table 5.5: Default wind turbine data of Matlab
Parameter Value
Pnwt 1.5 [MW]
P b wt 1.5/0.9 [MW]
Vwind 12 [m/s]
W b pu 1.2 [-]
Pmax Vwind 0.73
β 0 [deg]
The wind turbine model is presented in figure 5.12. The model inputs are wind
speed profile (coming from the kaimal model [m/s]), Generator speed (coming
from drive train [pu]) and pitch angle ( coming from pitch controller [deg]) and
Figure 5.10: Wind turbine power characteristics for default MATLAB values
Figure 5.11: Wind turbine power characteristics for the modeled wind turbine
the output of the wind turbine is mechanical torque load on the shaft (which is
the input to the drive train [pu]). This model develops the known wind turbine
power and torque equations:
Pwind = 0.5air · (π · R2wt ) · Vwind

3
(5.11)
Pm
C P (λ, β) = ; CP < 1 (5.12)
Pwind
Pm
Tm = (5.13)
Wwt
and the gains highlighted in red are calculated as followed:
1 1
k1 = = (5.14)
VWbase Vwind
1
k2 = (5.15)
Wb pu
k3 = λn (5.16)
1
k4 = (5.17)
C Pn
Pmax Vwind · Pnwt
k5 = (5.18)
Pbwt
These gain values are directly affected by the parameters defined in figure 5.9
and presented in table 5.4. Note that k1 is one over the base wind speed consider-
ing the same values for both average wind speed and base wind speed, the output
of the gain would be in per unit of the average wind speed. The performance coef-
ficient function C P depends on both Tip Speed Ratio (TSR) (λ) and pitch angle
(β). C P is calculated in the subsystem highlighted in blue in figure 5.12 from the
following equations:
C P (λ, β) = C1 (c2 /λi − C3 β − C4 ) · e(−C5 /λi ) + C6 λ (5.19)

1 1 0.035
= − (5.20)
λi λ + 0.08β 1 + β 3
in which C1 to C6 are the performance coefficients parameters defined in

table 5.4. Accordingly, the nominal value of C P and λ are set to the optimum val-
ues derived from this function. The maximum C P is calculated at β = 0 & λ = 8.1
and equal to 0.48. Furthermore, since the power needed from the wind farm is
known previously, the blade radius can be designed for the optimum values and
the desired power. Note that R wt is essential as an input parameter to the wind
profile modeling with Kaimal spectrum as it is related to the length scale.
Figure 5.12: Wind turbine model
5.3.2.2 Drive train

Drive trains in wind turbines are responsible to transfer the energy from the
wind turbine blades to the generator and they consist of gearboxes. Two types of
common gearboxes for wind turbines are planetary and parallel type. The num-
ber of the gearboxes rotating together in the drive train depends on the speed of
the shaft. Higher speeds require more and bigger gear and usually planetary or a
combination of planetary and parallel are used. The term of mass is also used to
indicate the complication of the drive train [52].
As mentioned, the output torque calculated from the wind turbine goes through
the drive train (Figure 5.13). A two mass drive train [51], models the shaft stiff-
ness, mutual damping (second part) and the inertia (first part) of the wind tur-
bine (Note that the wind turbine model in figure 5.12 does not incorporate the
turbine inertia). The initial values of the discrete integrators are defined such that
the steady state value is calculated faster. The gains highlighted in red are gains
k1 = 0.5/H wt and k2 = ω r base and gear ratio is considered as one for convenience
(N = 1). The governing equations behind the drive train model are as followed:
dωwt
2H wt = Tm − Ts (5.21)
dt
θst a
= ωwt − ω r (5.22)
dt
θst a
Ts = SS · θst a + DC · (5.23)
dt
Equations 5.21, 5.22, 5.23 model the inertial response of the wind turbine,
shaft twist angle and the shaft torque , respectively, where,
H wt = Inertia constant of the wind turbine [s]

θst a = Shaft twist angle [d e g]
ωwt = Wind turbine speed [pu]
ω r = DFIG rotor speed [pu]

Ts = Shaft torque [pu]
SS = Shaft stiffness [-]
DC = Damping coefficient [-]
and the values for SS, DC and H wt can be found in table 5.4.
Figure 5.13: Two mass drive train model
5.3.2.3 Pitch control

Multiple pitch control strategies have been introduced in the literature. Pitch
control is used to reduce/increase the angle of the wind turbine blades with the
incoming wind such that the power output of the turbine is controlled. The pitch
controller is commonly used for MPPT purposes with a wide variety of simple to
complex algorithms. The idea of using pitch control instead of stall control is cru-
cial specially in high wind speeds and offshore wind farms. Pitch controls fall into
categories of conventional, collective, individual, electric, robust and soft control-
lers [53].
In this project a conventional pitch control system is used. The pitch control-
ler [51] presented in figure 5.14 compares the rotor speed of the connected gen-
erator to the shaft of wind turbine (DFIG) with the reference speed of 1 pu and
the error is multiplied by a gain (kβ = 500) and given as the pitch signal to the
wind turbine in figure 5.12. The pitch is limited to the values is table 5.4 with
βopt , βma x , and rβ .
5.3.3 Simplified wind turbine model

The wind turbine model explained in section 5.3.2 can be simplified without
loosing valuable information for the purpose of the project. The simplified version
is analysed against the detailed model. The reason for developing an additional
Figure 5.14: Pitch controller
model for the wind turbine is that details of the model of section 5.3.2 do not
directly affect the virtual inertia algorithm or system response to a contingency
event. Therefore, an additional model for the wind turbine has been developed
with less detail and presented in this section. Figure 5.15 demonstrates the model.
In this model, first TSR is calculated and the value is used in a look up table to find
the corresponding C t . The value of C t is then used to calculate the input torque
signal for the DFIG. C t is the torque coefficient and f 1 and f 2 are the TSR and
torque equations (in figure 5.15).
CP
Ct = (5.24)
λ
f1=λ (5.25)
2
f 2 = 0.5ρair · R3wt · Vwind · Ct (5.26)
Figure 5.15: Simplified wind turbine model
In the simplified model the pitch angle is considered to be constant (β = 0)

and a gear ratio of 1:100 is selected. The look up table reads the data from the
figure 5.16.
5.3.4 DFIG
DFIGs are the conventional design for the wind turbines and in this project,
the wind turbine is coupled with a DFIG. The DFIG has a wound type rotor, three
phases and rotor is set as reference frame. No saturation is considered for the
model and the initial conditions are all set to zero. This generator is mainly re-
sponsible for supporting the main load of the platform with a connection through
Figure 5.16: (left) Torque coefficient-λ, (right) Power-Vwind
back-to-back converters (type 3). This enables the control of DC link voltage, gen-
erator speed, active and reactive power delivered to the platform. The converters
used in the model are ideal with switching frequency of 16kHz. The value for
switching frequency although not in accordance with the size of the converters,
does not affect the general performance of the system. The generator data are
presented in table 5.6.
Table 5.6: DFIG data

Pn 2 [MW] Nominal power of the DFIG
ω r base 155 [rad/s] Base rotational speed
fn 50 [Hz] Nominal frequency
Vn 690 [V] Nominal three phase-to-phase voltage
u 1/3 Stator/rotor turn ratio
sma x 1/3 Maximum slip
Rs 0.0026 [Ω] Stator resistance
Rr 0.0029 [Ω] Rotor resistance
Lls 0.087 [mH] Leakage inductance
Lm 2.5 [mH] Linkage inductance
J 127 [kg.m2 ] Moment of inertia
P 2 [-] Pole pairs
The model of the DFIG used is the original block of the Matlab and in order
to implement the control strategies for the RSC and GSC, the model explained
in [14] and chapter 10 of [15] are implemented.
5.3.5 Rotor Side converter (RSC)

The control of RSC differs according to the goal of the converter (Zero D-axis
Current (ZDC), Unity power Factor (UPF), Constant Stator Voltage (CSV)) but the
majority of the control methods implement the following procedure:
• Define parameters of interest that are desired to be controlled such as gen-

erator speed, active or reactive power and so on
• implement (an outer layer) controller for the parameters of interest and
generate reference signals for currents or voltages depending on the control
strategy
• implement (an inner layer) controller for currents/voltages
• implement any additional calculations to generate the desired PWM signals

for the converter.
In this project, the ZDC method is selected. For convenience of the design and
control, the converter controllers are not implemented in the abc reference frame
but in dq0. Therefore, while modeling all the vectors on d and q reference frames,
one can separate the controllers in each reference frame with decoupling terms.
In the ZDC control strategy, the d-axis reference for current is set to zero, resulting
in all the power being controlled by the q-axis vectors. Figure 5.17 illustrates the
RSC control strategy to generate the PWM signals.
Figure 5.17: Vector control of the RSC
The d-axis current reference is set to zero as mentioned and the q-axis current
is generated from the outer layer controller for the rotor speed. Two controller are
designed for the outer controller and they are evaluated against each other. First
controller follows the speed reference and the other one, implements an indirect
speed control with a MPPT scheme. The output of the outer layer controller is the
torque reference, which then is multiplied by a gain to give the current reference
for the q-axis. The current references are then regulated with the inner layer con-
trollers and after adding the decoupling terms, the d and q-axis voltage references
are generated and sent to the PWM generator. In this method, the three phase ro-
tor currents are sampled and sent as a feedback to the inner layer controllers of
the current. A 3rd harmonic injection algorithm (section 2.4.7.4) is also imple-
mented before all the PWM inputs on the reference voltages. The PWM model
implements the SVPWM technique explained in section 2.4.7.3 for all the con-
verters in this project. PLL is developed for extracting the frequency and angle of
the line voltages.
5.3.5.1 PLL circuit

The PLL circuit used in this model was inspired by [54]. The operation of the
model is quite similar to the explanations given in section 2.4.4. Sampled voltage
is transformed to d-q reference frame and the angles are calculated. The rest of the
procedure is illustrated in figure 5.18. The data used for the PLL are also available
in figure 5.19. This design outputs the frequency, the angle and the sin and cos of
the angle. This PLL is used in all the cases where a PLL is needed.
Figure 5.18: PLL circuit
5.3.5.2 Angle calculation for RSC

The angle used in calculating the park transformations can be calculated from
equation 5.27 and 5.28.
π
θs = θ P L L − (5.27)
2
θ r = θ P L L − P · θm (5.28)
in which θm , θ r , θs , θ P L L and P are the mechanical angle of the shaft, electrical

angle of the rotor, stator and PLL.
Figure 5.19: PLL data
5.3.5.3 Decoupling terms in RSC
To consider the effect of compensation for coupling between d and q axis the
decoupling terms are used after the current controllers from Eq 5.29 and 5.47
while neglecting the effect of the converter and possible delays in the system:
Vd r = ed r − ω r σL r iqr (5.29)
Lm ~
Vqr = eqr + ω r σL r id r + ω r | ψs | (5.30)
Ls
in which ed r and eqr are the outputs of the current regulators and Vd r and
Vqr are the reference voltage signals for the SVPWM and ψs is the approximated
stator flux [15] calculated as followed:
s
2
Vn
~s |= 3
|ψ (5.31)
2π f n
|ψ~s | is also used to generate the q axis current reference from the torque
(figure 5.17).Vd r and Vqr are then transformed to abc reference frame with inverse
park transformation and fed to the SVPWM block. The decoupling of the d and q
axis enables the separate control for the d and q axis currents to generate ed r and
eqr separately with equal controllers for both d and q axis.
5.3.5.4 Current regulators
The conventional method for controlling dq currents is using a PI controller

tuned by modulus optimum criterion [55]. The modulus optimum is selected for
cases in which the system has one dominant time constant along several minor
one. The idea of this method is to cancel the largest time constant of the system
with the controller. Assuming the controller has no time delay from switching of
the converter, sampling/measurements and other calculations in the controller,
the plant model of the DFIG on the rotor side can be written as:
1 1
G plant = · (5.32)
R r 1 + τi · s
σ · Lr
where τi = and σ = 1 − L m 2
/Ls L r as defined in section 2.3.3. Con-
Rr
sidering that the inner loop controllers (current controllers) must be faster then
the outer loop controllers (speed controller), while slower than the switching fre-
quency of the converter, the natural frequency of the current controller is set to
one decade lower than the switching frequency:
ωni = 0.1 fsw = 1.6kHz (5.33)
Therefore, the proportioanl and integral gains of the current controller are:
k pi = (2ωni · σ · L r ) − R r = 0.5771 (5.34)

kii = ω2ni · L r · σ = 491.5995 (5.35)
Note that the exact same controllers are used for both d and q axis currents.
5.3.5.5 Speed regulator
The conventional method for controlling outer loop parameter (speed/ voltage
and so on) is using a PI controller tuned by symmetrical optimum criterion [55].
The reason is that adding another pole to the system transfer function, builds
a third order transfer function that can not be tuned with modulus optimum. Al-
though assuming that the inner layer controller is much faster than the outer layer
controller, the transfer function of the inner layer controller can be simplified to a
constant gain, resulting in a reduce order of the third order transfer function [56].
Hence modulus optimum can be applied again. the plant model of the DFIG speed
can be found from the inertial response of the DFIG:
PDF I G
G plant = (5.36)
J DF I G · s
Considering that the inner loop controllers (current controllers) are faster then
the outer loop controllers (speed controller), the natural frequency of the speed
controller is set approximately three decade lower than the switching frequency
and two decade lower than the current controller natural frequency:
ωnω ' 0.01ωni = 20Hz (5.37)

k pω = (2ωnω · J DF I G )/PDF I G = 2540 (5.38)

kiω = ω2nω · J DF I G /PDF I G = 25400 (5.39)
5.3.5.6 MPPT
For designing the MPPT algorithm an indirect speed control [14] is considered.
While the wind speed changes, the algorithm adapts the reference torque signal
of the DFIG in order to extract the maximum power from the wind turbine. Fig-
ure 5.20 illustrate MPPT on general power and torque charts. While working in
maximum power point, the turbine characteristics are as followed:
R wt · Wwt
λopt = ; C P = C P max ; C t = C t max (5.40)
Vwind
According to the optimal values torque is calculated as:
Topt = Kopt · ω2wt (5.41)

R5wt
Kopt = 0.5ρair π 3 C P max (5.42)
λopt
(a) (b)
Figure 5.20: MPPT graph on (a) P − ω, (b) T − ω charts for different wind
speeds
The speed of the generator is sent as feedback to the MPPT algorithm and
after applying Eq 5.41, the reference torque signal is generated. The value for Kopt
according to wind turbine parameters in 5.4 is calculated 2.96 × 105 . Figure 5.21
demonstrates the implemented algorithm.
Figure 5.21: MPPT algorithm
5.3.6 Grid Side Converter (GSC)

A similar approach is taken for designing the GSC. Figure 5.22 illustrates the
GSC controller. Instead of d axis, the q axis current reference is set to zero and
the d axis reference is generated by the DC link voltage controller. The rest of the
strategy is quite similar.
Figure 5.22: Vector control of the GSC
5.3.6.1 DC link capacitor
The Dc link capacitor [57–59] is sized in the industry according to three cri-
teria named, ripple voltage/current rating, DC voltage rating, resonant frequency
rating. The KCL for the DC link:
d VDC
idc = C DC · + iL (5.43)
dt
where i DC and i L are the DC current (connected to the RSC) and the load cur-
rents (connected to the GSC). In the back to back converters, the most important
criteria for the design of the DC link capacitor is the voltage ripple. A typical value
for the voltage ripple is 0.01 to 0.02 pu. Accordingly, C DC = 80mF is selected to
satisfy this condition with parameters of table 5.6 and a DC voltage level of 1150
V.
5.3.6.2 GSC filters

The addition of power converters in power systems causes injection of har-
monics. These harmonics have to be limited in order for the grid components to
function appropriately. In the back to back converter strategy, usually a filter is de-
signed to filter the harmonics and output a smoother signal with low ripple. In this
project a simple RL filter is designed to rectify this issue and placed between the
GSC and PCC as shown in figure 5.22. The parasitic resistor is usually in the range
of 0.001 to 0.01 pu and as a general criteria, the inductance of the filter considered
must limit the current limit to maximum 10% of the rated current amplitude [60].
The value of the inductance typically can be calculated from Eq 5.44.
(VDC − D · VDC ) · D
Lf l = (5.44)
2∆imax,p−p · fsw
p
∆imax,p−p = 0.1 2I base (5.45)
The maximum current ripple occurs for duty cycle of D = 0.5. The values of
the filter selected for the system are R f l = 0.02mΩ and L f l = 0.4mH.
5.3.6.3 Decoupling terms in GSC

The decoupling terms for compensation in the GSC that generate the dq refer-
ence voltage signals(Vd g , Vq g ) for the inverter are as followed:
Vd g = ed g − (2π f n )L f l · iqg + ud g (5.46)

Vq g = eqg + (2π f n )L f l · id g + uqg (5.47)
where ed g and ed g are the output signals of the current regulators, ud g and
uq g are the feed-forward of the sampled grid voltage.
5.3.6.4 Current regulators

The design of the current regulators of the GSC are similar to the design of
the RSC current regulators explained in section 5.3.5.4. The difference is that the
plant transfer function is calculated from Eq 5.48:
1 1
G plant = · (5.48)
R f l 1 + τi · s
Lf l
where τi = . The natural frequency of the current controller:
Rf l
ωni = 400Hz (5.49)
k pi = (2ωni · L f l ) − R f l = 0.3016 (5.50)

kii = ω2ni · L f l = 56.8489 (5.51)
5.3.6.5 DC voltage regulator

The voltage regulator for the GSC is tuned according to the same method in
section 5.3.5.5.
k pv = 1.2 (5.52)
kiv = 360 (5.53)
5.4 Virtual inertia provision unit

The Virtual Inertia Converter (VIC) control strategy is quite similar to the GSC
control strategy with adaptations for implementing the inertial response. Instead
of extracting energy form the DC link (as in back to back converter), the energy
is supplied via an ESS. The ESS used in this project is a controlled current source.
The voltage of the ESS is kept constant at VESS with a very fast PI controller and the
output of the controller is the current reference for the controlled current source.
The current is passed through a switching converter which is controlled with the
VI algorithm. Figure 5.23 depicts the VIC controller. The q axis current reference
is set to zero, while the d axis current reference is generated from the frequency
controller, which is responsible for the inertial response for the frequency. The
ac side filters are designed with the same method as in section 5.3.6.2. The de-
coupling terms are also the same as section 5.3.6.3. The current references of the
current regulators are calculated from the reference active and reactive powers of
the inverter:
P = Vd · id + Vq · iq (5.54)
Q = Vq · id − Vd · iq (5.55)
Considering zero q axis voltage and zero reactive power reference:

Figure 5.23: Vector control of the VIC
P
idd,r e f = (5.56)
Vd
iq,r e f = 0 (5.57)
where P is the active power reference (output of the frequency controller, VI

algorithm). Table 5.7 demonstrates the VIC data for simulation.
Table 5.7: VIC data

fn 50 [Hz] Nominal frequency of the grid
Vn 690 [V] Nominal line to line RMS voltage
V ESS 1150 [V] ESS DC voltage
fsw 15 [kHz] Switching frequency
Rf l 0.5290 [Ω] Filter resistance
Lf l 33.37 [mH] Filter inductance
The strategy for VI in this project is on the same base as frequency power
based methods explained in section 3.4. The frequency variation is controlled
and dq reference currents are generated. This method although not having the
derivative term, still has the issues incorporated with methods of section 3.4. The
VI algorithm is developed in three cases, named case e, f, and g:
5.4.1 Case e
In this case, the VI is implemented with a P controller. Figure 5.24 demonstrates
the model:
Figure 5.24: Case e- Virtual inertia provision with P controller
where, k P is the proportional gain equal to 1000.
5.4.2 Case f
In this case, the VI is implemented with a PI controller. Figure 5.25 demon-
strates the model:
Figure 5.25: Case f- Virtual inertia provision with PI controller
where, k P and ki are the proportional and integral gain equal to 1000.
5.4.3 Case g
In this case, the VI is implemented with a PI controller alongside a dead zone
for reacting to frequency. Figure 5.26 demonstrates the model:
Figure 5.26: Case g- Virtual inertia provision with PI controller and dead zone
where, k P and k are same as previous case and dead zone is 0.005 pu.
Chapter 6
Simulation Results
In this chapter, the progressions of the model and the case studies (as explained
in section 4.3 cases a to c are the progressions and cases d to g are the case studies)
developed for the project are presented.
6.1 Case a: Wind turbine drive train and pitch controller

(Wind turbine package)
In this section, the simulation results for the wind turbine package are presen-
ted separately to validate its operation. All the values are calculated in pu and the
operation is tested via changing the feedback rotor speed (Wr ) of DFIG as step
inputs of 1, 0.9 and 1.1 pu at t = 0, 20, 55 seconds. The same model as figure 5.8
is presented below:
Figure 6.1: Wind turbine package simulation model
The results of the simulation indicates that the wind profile oscillates randomly
with an average value of 12 m/s (figure 6.3). The turbine successfully tracks the
DFIG rotor speed (figure 6.5) and its oscillations are damped according to damp-
ing designed in the drive train. The pitch angle (figure 6.4) changes and at first
accelerates the turbine (as expected and seen in figure 6.5) by increasing the pitch
angle and reducing the output torque/power of the turbine. If the speed reference
would have remained at 1.1 pu for a longer period, the pitch angle would continue
to increase until maximum pitch angle of 45 degrees and after that the turbine
81
Figure 6.2: DFIG rotor speed reference (Wind turbine package simulation
model)
Figure 6.3: Wind profile, (Wind turbine package simulation)
Figure 6.4: Pitch angle, (Wind turbine package simulation)
Figure 6.5: Wind turbine speed, (Wind turbine package simulation)

Chapter 6: Simulation Results 83
Figure 6.6: Shaft input and output torque, (Wind turbine package simulation)
would be unable to track the reference signal. After lowering the rotor reference
speed to 0.9 pu, the pitch controller starts braking the turbine (figure 6.4) and
increasing the toque/power. Note that in figure 6.6 the input shaft torque and
wind turbine torque are the same signals (wind turbine torque is sketched from
internal signal in wind turbine package). The output shaft torque has some oscil-
lation which are damped in the drive train.
Note that in reality, according to the pitch control, nominal value of the wind
turbine and restrictions of safety and manufacturing (considering both wind tur-
bine and DFIG have the same nominal power), the wind turbine is unable to pro-
duce power which exceeds its nominal value for a long time but it is capable of
braking to reduce its output power. Additionally, if the DFIG is exceeding its nom-
inal rotor speed, the wind turbine coupled with the generator should not inject
more power, since it will speed up the generator even more and makes it unstable.
The results of the simulation shows that the wind turbine follows the DFIG
not the other way around while at the same time, the DFIG rotor speed is con-
trolled with the power converters. Additionally, figure 6.6 shows that with this
designed turbine the input mechanical torque signal of the DFIG system will be
oscillatory even if the reference speed is constant. This might cause oscillatory
output power for the DFIG and depending on the accepted ripple for the output
(if accepted at all), the design of the wind turbine should be changed. This issue is
investigated further after simulation of the DFIG and examining the output power.
6.2 Case b: DFIG only with RSC

6.2.1 Case b1: PI controller and wind turbine package
The DFIG is loaded with the shaft torque (output of case a) coupled with the
wind turbine and the electrical load of the platform. To simulate the model, a three
phase voltage is connected directly to the output terminals of the DFIG. The PI
controller is responsible for controlling the speed of the DFIG and generating the
current reference as explained in subsection 5.3.5.5. Figure 6.7 illustrates model
of case b1. The simulation aims to test the capability of the controllers to control
the dq currents and speed of the DFIG for the nominal values of the DFIG.
Figure 6.7: Case b1 simulation model
Figure 6.8 shows that the wind speed is successfully generated with random
distribution and an average value of 12 m/s. Figure 6.9 illustrates the speed of the
wind turbine which is approximately 1 pu with minor oscillations and figure 6.10
shows how the pitch controller is functioning in respond to changes from the DFIG
rotor speed. The changes in the pitch angle are not substantial and is in accord-
ance with the wind turbine speed that does not change much. Figure 6.11 shows
that the output torque of the shaft which is the input signal of the DFIG is oscil-
lating around 1 pu. Note the wind turbine and input shaft torques are the same
(it is trivial because the same wind turbine package of case a is used) but they os-
cillate considerably due to the nature of the variable wind speed. The drive train
is successfully damping these oscillations and giving a much smoother signal as
the output. Fig 6.12 is the same signal of the output shaft torque but in physical
units (N.m). This almost constant torque enables the DFIG to provide an output
power with presumably low oscillations. It has to be checked that if the power
oscillations are acceptable with this design or not.
Figure 6.8: Case b1- Wind speed (m/s)
Figure 6.9: Case b1- Wind turbine speed (pu)
Figure 6.10: Case b1- Pitch angle (deg)
Figure 6.13 shows the rotor speed of the DFIG with a desired nominal value
Figure 6.11: Case b1- shaft torque (pu)
Figure 6.12: Case b1- Mechanical input torque of generator (N.m)
of 155 rad/s. The PI speed regulator tracks the 155 reference signal quite accept-
able with minor deviations that do not exceed 5 rad/s. These deviations occur as
the input mechanical torque signal of the DFIG is not constant and has small os-
cillations. Since the PI speed regulator functions correctly, the current regulators
have the correct reference signals as depicted in figure 6.14 and 6.15 in red col-
ors (note that d axis reference current was set to zero). These figure also indicate
that the designed PI controllers for currents are also preforming well. Although
the design could have been changed to have lower ripples which is a trade off in
design between the costs and the performance. The stator and rotor current are
presented in figures 6.16 and 6.17 which indicate that with the given sinusoidal
voltage at the PCC, the DFIG produces almost sinusoidal currents, which again
correspond to the current regulators and their performance. A filter can also be
added to the system which is considered from case c forward.
With these performances, the DFIG output the electromagnetic torque of fig-
ure 6.18 which is following its reference. This fluctuating torque was expected
as the input torque of the DFIG was also oscillatory and the reason for that was
mainly the variable wind speed. In practice, the grid codes force the power quality
attribute to the wind farm operators and the fluctuations are compensated by the
grid because the grid is considerably stronger. With the recent advancements in
the offshore wind turbine technology and increased size of the turbines the shaft
Figure 6.13: Case b1- Rotor speed (rad/s)
Figure 6.14: Case b1- q axis RSC current (A)
Figure 6.15: Case b1- d axis RSC current (A)

Figure 6.16: Case b1- stator current (A)
Figure 6.17: Case b1- rotor current (A)

toque is even less oscillatory because of the high mass and inertia of the turbine
which enables the turbine to operate at the nominal values and nullifying the
effect of the variable wind speeds.
Figure 6.18: Case b1- Electromagentic torque (N.m)
6.2.2 Case b2: MPPT and Simplified wind turbine model

The same scenario as in case b1 with some changes is tested. Instead of the PI
speed regulator the MPPT algorithm of subsection 5.3.5.6 is used on the simpli-
fied wind turbine model of subsection 5.3.3. The controller focuses on extracting
maximum power form the generator while the wind speed is constantly changing
and the rotor speed is controlled indirectly. Additionally, the nominal power of
the wind turbine and DFIG are considered to be 2.4 MW instead of 2MW and the
nominal rotor speed is 200 m/s instead of 155 rad/s (these changes are explained
in the next paragraph). The simulation aims to test the effectiveness of the MPPT
algorithm in comparison with the PI speed regulator. The model developed in this
case is presented in figure 6.19.
Figure 6.19: Case b2 simulation model

Figure 6.20: Case b2- Rotor speed (rad/s)
Figure 6.21: Case b2- q axis RSC current (A)
Figure 6.22: Case b2- d axis RSC current (A)

Figure 6.23: Case b2- stator current (A)
Figure 6.24: Case b2- rotor current (A)
Figure 6.25: Case b2- Electromagentic torque (N.m)

Figure 6.20 shows the rotor speed of the DFIG and it takes about 4 seconds for
the simulation transients to finish and the system operate around its nominal val-
ues after 4 seconds. The reason for these transients is that the initial rotor speed
of the DFIG was set to 155 rad/s from previous simulation and it was not changed
in this case, while the nominal speed in this case is 200 m/s. This is because the
simplified wind turbine model of case b2 has a 2.4 MW nominal power at an av-
erage wind speed of 12 m/s (because the power curve data for a 2 MW was not
found while the data of a 2.4 MW was found as shown in figure 5.16), however
the wind turbine package in case b1 has the nominal power of 2 MW at the same
wind speed. Therefore, the DFIG rotor speed reaches a higher value in case b2
compared to case b1 (These transients and higher nominal rotor speed would not
have happened if the initial condition of the DFIG and the rated power of the wind
turbine were kept the same but they were not kept the same because the MPPT
was aimed to be tested even if the initial rotor speed was not the nominal value. It
is evident that the DFIG was not started from zero speed because with the design
values it would have taken a considerable amount of time for it to reach nominal
values and the computational restrictions of the PC in which the model was built,
would not allowed it). The simplified wind turbine model of case b2 with MPPT
which has a nominal power output of 2.4 MW at an average wind speed of 12 m/s,
gives 2 MW power at average wind speed of 11 m/s. Hence, if the wind profile is
generated with average value of 11 m/s and given as input to the wind turbine
model instead of 12 m/s, the problem is solved. Other then the the explained
issue, after 4 seconds of transients in the beginning of the simulation, the rotor
speed oscillates around its nominal value of 200 rad/s with minor changes of up
to 20 rad/s and the reason for these oscillations is the variable wind speed profile.
The PI regulators for dq currents were the same as previous case and they
have the correct reference signals as depicted in figure 6.21 and 6.22 in red colors
(note that d axis reference current was set to zero). These figure also indicate that
the designed PI controllers for currents are also preforming well and in fact much
better than the case b1. The stator and rotor current are presented in figures 6.23
and 6.24. The MPPT directly changes the input torque of the generator to extract
maximum power at each wind speed. Figure 6.25 shows that the reference for
electromagnetic torque is followed with less deviations compared to the previous
case. In a nutshell, cases b1 and b2, in case b1 a wind turbine of 2 MW was used,
while in case b2, wind turbine has 2.4 MW rated power. The difference is also
reflected in rotor speed. in case b2 the MPPT track the maximum power at each
wind speed and controls the torque directly and rotor speed indirectly while case
b1 is the opposite. Case b2 shows a more promising result than case b1 and has a
more stable electromagnetic torque.
6.3 Case c: DFIG with RSC and GSC

The previous cases b1 and b2 analysed the difference of each control strategy
for DFIG RSC. In cases c1, c2 and c3 the effect of these control methods are invest-
igated while interacting with GSC, its controller and the other added elements in
the system and after the simulation, the final model for wind turbine is selected.
Note the replacement of the DC voltage source with the DC link capacitor.
6.3.1 Case c1: PI controller with variable wind speed and wind tur-
bine package
For this analysis, the same structure as in case b1 (average wind speed of 12
m/s, wind turbine size of 2 MW) is used while adding the GSC and its controller,
the DC link capacitor and the GSC filters as shown in figure 6.26.
Figure 6.26: Case c1 simulation model

Figure 6.27: Case c1- Wind speed (m/s)
Figure 6.28: Case c1- Wind turbine speed (pu)
Figure 6.29: Case c1- Pitch angle (deg)
Figure 6.30: Case c1- shaft torque (pu)

Figure 6.31: Case c1- Mechanical input torque of generator (N.m)
Figure 6.32: Case c1- Rotor speed (rad/s)
Figure 6.33: Case c1- q axis RSC current (A)

Figure 6.34: Case c1- d axis RSC current (A)
Figure 6.35: Case c1- Electromagentic torque (N.m)
Figure 6.36: Case c1- stator current (A)

Figure 6.37: Case c1- rotor current (A)
Figure 6.38: Case c1- DC link voltage (V)
Figure 6.39: Case c1- q axis GSC current (A)

Figure 6.40: Case c1- d axis GSC current (A)
Figure 6.41: Case c1- q axis GSC voltage (V)
Figure 6.42: Case c1- d axis GSC voltage (V)

Figure 6.43: Case c1- rotor (purple), stator (red), and grid (yellow) voltage
frequency
Figure 6.44: Case c1: power delivered to the load (W)

Figures 6.27 to 6.37 indicate the exact same results as of in case b1. These
figures are sketched for the RSC controller parameters and only are presented to
show the correct interaction of the RSC while connected to GSC and therefore,
require no further explanation. The RSC controller is able to perform correctly
only when the GSC controller is operating well and provides it with the suitable
DC link voltage. This DC link voltage must be almost constant around the oper-
ating designed value. Hence, the RSC controller sees this DC link capacitor as a
constant voltage source. Figure 6.38 illustrates the DC link voltage. After the ini-
tial simulation transients of around 0.75 seconds the DC voltage regulator, which
is a PI controller, tracks the reference value of 1150 V with ripples no more than
100 V. Figures 6.39 and 6.40 illustrate the d and q axis currents of the GSC. The
currents have the correct references because the outer layer controller (Dc voltage
controller) operates correctly. Figures 6.41 and 6.42 show the voltages generated
after the decoupling terms which will be given to the PWM generator after con-
ditioning. Figure 6.43 depicts the frequency of the rotor, stator, grid and platform
frequencies. The stator and platform frequencies are measured form the same sig-
nal and they are identical. The grid frequency is the frequency of the GSC voltage
whcih is equal to the stator and platform frequencies. The rotor frequency is the
RSC voltage frequency which is lower then the stator frequency and in accordance
with the DFIG slip.
This control model as speculated, can not provide constant output power (fig-
ure 6.44) and more complications will have to be added to the project to com-
pensate the power and fix it for the 2 MW needed for the load. The reason for
irregular power output is the input torque of the generator, as the wind turbine is
not operating at its nominal wind turbine speed of 1 pu, the output torque of the
shaft which will be the generator input torque, is not constant and therefore, the
generator speed can not settle at its nominal value. The results proves that MPPT
algorithm might be a better option. Next case examines if the MPPT algorithm
gives results in accordance with the aim of the project.
6.3.2 Case c2: MPPT in variable wind speeds and simplified wind
turbine
For this analysis, MPPT with simplified wind turbine algorithm is used (average
wind speed of 11 m/s, wind turbine size of 2.4 MW) in order for the generator to
output 2 MW (according to figure 5.16) and the rest of the elements are similar
to case c1 as shown in figure 6.45.
Figure 6.46: Case c2- Rotor speed (rad/s)

Figure 6.49: Case c2- q axis RSC voltage (V)

Figure 6.50: Case c2- d axis RSC voltage (V)


frequency

Figures 6.46 to 6.53 indicate achieving similar results as in case b2. These
figures are sketched for the RSC controller parameters and only are presented to
show the correct interaction of the RSC while connected to GSC and therefore,
require no further explanation. Similarly, figures 6.54 to 6.58 show results for the
GSC controller which is the same controller in case c1. Hence, the results of the
simulation are similar. Although note that the DC link voltage fluctuates quite less
than the previous case. For this reason the reference signal for the d axis controller
is also less oscillatory and therefore, the current has better quality. Figure 6.59 de-
picts the frequencies in the system. Similar terminology is used as previous case.
The much notable difference is that the fluctuations are less and the rotor speed is
decreasing continuously which should settle eventually but due to computational
power problems was was not pursued.
The power delivered to the platform fluctuates around the 2 MW desired value
(figure 6.60). This was expected as the input wind speed is changing around 11
m/s. Therefore, the MPPT algorithm is continuously changing the input torque
signal of the generator leading to generator speed always fluctuating around its
steady state value. Therefore, the RSC current references will also fluctuate. It
is speculated that if the wind speed is considered constant, the fluctuation are
substantially lower and the power output would be much smoother. Next case,
examines the effect of constant wind speed.
6.3.3 Case c3: MPPT in constant wind speed and simplified wind tur-
bine
For this analysis MPPT wind simplified wind turbine algorithm is used along
constant wind speed input of 11 m/s in order for the generator to output 2 MW
(according to figure 5.16). The model is presented in figure 6.61.
Figure 6.62: Case c3- Rotor speed (m/s)
Figures 6.62 shows that the rotor speed reaches the steady state value after
approximately 6 seconds. Note that the initial speed was set to 155 rad/s which
in order to check the capability of the MPPT to track the speed indirectly and
Figure 6.65: Case c3- q axis RSC voltage (V)
Figure 6.66: Case c3- d axis RSC voltage (V)



frequency
as explained before in case b2. Figures 6.63 and 6.64 illustrate the correct refer-
ence tracking of the PI controllers and in figure 6.65 and 6.66 the voltage signals
for the PWM algorithm are presented. The electromagnetic torque settles quite
nicely after the initial 6 seconds transients (figure 6.67). The stator and rotor
current are much better than the previous cases because of less fluctuations in
the currents of the PI controllers (figure 6.68 and 6.69). The GSC controller is
unchanged and therefore, the results of figures 6.70 to 6.74 are quite similar to
previous case (although with lower transients an a much better steady state res-
ult). Figure 6.75 shows the frequency of different voltage signals in the system.
As the figure depicts, the frequency of the rotor settles after 10 seconds because
the current references, the DC link voltage and rotor speed almost settle. Most
importantly, figure 6.76 shows that the power measured at the PCC is settling at
2 MW, which was the aim of this whole progression so far. Case c3 is selected for
further analysis and simulation of other components of the platform.
6.4 Case d: Gas turbine and wind turbine

In this scenario, first case c3 is connected to the gas turbine and loads and
then the additional load of 5 kW is switched in and out at t = 13, 14 seconds as
shown in figure 6.77. With this setup the gas turbine is initially supplying 5 kW
and during the contingency has to provide 10 kW alongside the 2 MW coming
from the wind turbine. Since the gas and wind turbines are the only sources of
energy, having a constant output from wind turbine and not having the virtual in-
ertia in place, the frequency drops down to 46.8 Hz. This is because all the stress
is placed on the gas turbine to provide the load power.
Figure 6.78, 6.79 and 6.80 show the response of the system in this case. In
figure 6.78 the yellow waveform shows the change in the active power demand
of the load, the red waveform is the output power of the gas turbine and the green
waveform is the output power of the ESS which is disconnected with a large im-
pedance (it is still drawing some current, I forgot to solidly disconnect it but it
does not affect the results). Note that from this scenario forward, load power is
shown with a DC gain of -2MW in order to illustrate the share of gas turbine and
ESS.
Figure 6.79 shows the frequency response of the system. As it can be seen,
the frequency drops substantially to almost 46.8 Hz due to the added load but it
will recover after the load is set back to its initial value. Figure 6.80 shows that
the magnitude of the peak ac voltage at PCC remains constant and it will keep
its three phase sinusoidal waveform and the current drawn from the gas turbine
increases in the period of excess power. Notice the saturation of the input mech-
anical power of the gas turbine (due to the governor saturation) after 0.6 seconds
and decrease of the gas turbine rotor speed at the same time. If this contingency
were to continue more than one seconds, the speed would have reached even
lower values. Similar scenario would have happened to the frequency. After the
load is set back to initial value, the gas turbine speeds up again and the frequency
gets back to nominal value. Next case studies the effect of adding the ESS and
emulating virtual inertia.
Figure 6.77: Case d simulation model
Figure 6.78: Case d- active power of load [W] (yellow), output power of the
gas turbine (red) and output power of the ESS [W] (green)
Figure 6.79: Case d- grid frequency [Hz]

M. Heidari: Control Models for Providing Virtual Inertia
Figure 6.80: Case d- gas turbine simulation results. output power [W] (top-left), input mechanical power [pu] (mid-left), generator
speed [pu] (bottom-left), three phase ac output voltage [V] (top-right), and three phase ac current output current [A] (mid-right)
114
6.5 Case e: Gas turbine, ESS with P controller for Virtual

inertia and wind turbine
In this scenario, similar to previous scenario, the additional load of 5 kW is
switched in and out at t = 13, 14 seconds. The P controller successfully imple-
ments the inertia emulation by simulating the droop control of a governor in SG.
The stress on gas turbine is reduced considerably, because the ESS starts injecting
power as soon the load is switched in and frequency deviation is observed. Fig-
ure 6.82 shows that the gas turbine power jumps to 10 KW at t = 13s to supply
the necessary load, while the ESS slowly injects power to the grid. After about
0.5 seconds that the ESS is contributing approximately 1.8 kW, the gas turbine
experience less output power then 10 kW. At t = 14s that the load is set back to
the initial value the ESS stops injecting power to the grid after a small transient
and the gas turbine also supplies the 5 kW initial load.
Figure 6.83 shows the operation of the ESS. At t = 13s when the additional
load is connected, the frequency start to drop. The sensors pick up this change and
feedback the frequency to the P controller. The controllers generates the reference
current signal for the d axis. The current controllers are operating smoothly with
very low oscillations and control the currents. According to the Park transforma-
tion, the q axis voltage is set to zero and d axis follows the RMS value of the output
voltage. The q axis voltage only deviates from its zero reference value during the
transients. The waveform of the current drawn from the ESS is directly linked to
the power output of the ESS. The power output (shown as Po_VSG) in the figure
justifies the operation of the ESS according to the goal that was defined (which
is injection of additional energy to the grid to relieve the stress on gas turbine
and emulate virtual inertia). The dP_in waveform shows the output signal of the
P controller which divided by the d axis voltage is the current reference. The con-
trollers can be controlled better to have less error and oscillations. The positive
contribution of the ESS is directly reflected in the frequency of the grid and the
frequency response of the system has been improved substantially and only drops
to 48.5 Hz in its nadir point.
Figure 6.84 illustrates the response of the gas turbine to load addition while
the ESS is connected. In this scenario, the mechanical input of the gas turbine was
not saturated and the speed drop was significantly less (0.974 pu compared to
0.938 pu in previous case), while the magnitude of the peak ac voltage remained
constant. Additionally, even if the load change would continue after t = 14s the
gas turbine would not loose its speed because the ESS designed is capable of
injecting infinite energy and at around t = 13.8, the grid reached a new steady
state condition. If more restrictions were added to the ESS model, the response
would have been different. After t = 14s the system goes back to the initial state
after a small transient.
Figure 6.81: Case e simulation model
Figure 6.82: Case e- active load power [W] (yellow), gas turbine output power
[W] (red), and ESS output power [W] (green)
Chapter 6: Simulation Results
Figure 6.83: Case e- ESS converter simulation results. d axis current and its reference [A] (top-left), q axis current and its reference
[A] (mid-left), Output active power and its reference [W] (bottom-left), Output reactive power [Var] (top-mid), PLL angle [deg] (mid-
mid), output voltage in d and q reference [V] (bottom-mid), ac output voltage of phase a [V] (top-right), ac output current of phase a
[A] (mid-right), and frequency [Hz] (bottom-right)
117
Figure 6.84: Case e- gas turbine simulation results. output power [W] (top-left), input mechanical power [pu] (mid-left), generator
118
6.6 Case f: Gas turbine, ESS with PI controller for virtual

inertia and wind turbine
In this scenario, the load changes similar to previous case. All the explanations
of the previous case also applies to this one. The PI controller successfully im-
plements the inertia emulation. The frequency response of the system has been
improved substantially and even better than the previous case (frequency nadir is
48.7 Hz instead of 48.5 Hz), although after disconnecting the additional load, the
steady state result is achieved later than the previous case. A derivative term could
also be added to the controller, but it was avoided in this method due to possible
stability issues. Figure 6.85, 6.86 and 6.87 show the response of the system in
this case. The system will reach steady state eventually at the initial conditions
but due to computational limits the system was only simulated until t = 17s.
Figure 6.85: Case f- active load power [W] (yellow), gas turbine output power
Figure 6.86: Case f- ESS converter simulation results. d axis current and its reference [A] (top-left), q axis current and its reference [A]
(mid-left), Output active power and its reference [W] (bottom-left), Output reactive power [Var] (top-mid), PLL angle [deg] (mid-mid),
output voltage in d and q reference [V] (bottom-mid), ac output voltage of phase a [V] (top-right), ac output current of phase a [A]
(mid-right), and frequency [Hz] (bottom-right)
120
Figure 6.87: Case f- gas turbine simulation results. output power [W] (top-left), input mechanical power [pu] (mid-left), generator
121
6.7 Case g: Gas turbine, ESS with PI controller and dead

zone for virtual inertia and wind turbine
In this scenario, the load changes are identical to last case. The deadzone limits
the input of the PI controller and as a result the frequency drops more than in case
f (drops to 48.45 Hz). The idea behind adding the deadzone was to reduce the
sensitivity of the ESS to the frequency changes in the system, so that the ESS
only would be activated if the frequency deviations violates a certain range. A
disadvantage of this method is the slower steady state result which is not desirable.
Therefore, the addition of the dead zone appears to be unnecessary when the PI
controller is implemented. The same explanations of the two previous cases are
also valid for this scenario. Figure 6.89, 6.90, and 6.88 show the response of the
system in this case.
Figure 6.88: Case g- active load power [W] (yellow), gas turbine output power
Figure 6.89: Case g- ESS converter simulation results. d axis current and its reference [A] (top-left), q axis current and its reference
[A] (mid-left), Output active power and its reference [W] (bottom-left), Output reactive power [Var] (top-mid), PLL angle [deg] (mid-
mid), output voltage in d and q reference [V] (bottom-mid), ac output voltage of phase a [V] (top-right), ac output current of phase a
[A] (mid-right), and frequency [Hz] (bottom-right)
123
Figure 6.90: Case g- gas turbine simulation results. output power [W] (top-left), input mechanical power [pu] (mid-left), generator
124
6.8 Comparison between cases with virtual inertia emu-

lation
The results of the previous cases (d to g) are compared in this section.
Figure 6.91: Frequency response of the platform grid in case d (blue), e (red), f
(yellow), and g (purple) to the load change
Figure 6.92: Virtual inertia block response in case d (blue), e (red), f (yellow),
and g (purple) to the load change
As figure 6.91 illustrates, case f offers the highest inertial support, while case
d offers the lowest. In case f, the frequency nadir is 48.7 Hz, while in case d the
frequency drops to 46.8 Hz. Also the ROCOF is considerably lower in cases e,f and
Figure 6.93: Gas turbine speed response in case d (blue), e (red), f (yellow),
and g (purple) to the load change
Figure 6.94: Gas turbine output power in case d (blue), e (red), f (yellow), and
g (purple) to the load change
g where the ESS contributes to the energy balance of the system and the system
restores faster. The contribution of the ESS is shown by the output power of the
virtual inertia block, which will be the reference power signal of the ESS converter.
As figure 6.92 illustrates, case f draws the highest power form the ESS as expected.
Figure 6.94 shows the stress on the gas turbine for the defined scenarios, where
case d has the highest impact on the gas turbine, as ESS is not injecting any power
to assist the gas turbine. The higher power extraction from ESS in case f is also
reflected in gas turbine power output, as in this case the lowest amount of power is
extracted from the gas turbine, meaning less stress during the contingency event.
As gas turbine speed and output power are linked, similar scenario happens for
gas turbine speed as depicted in figure 6.93.
Chapter 7
Conclusions and Future Work
7.1 Conclusions
This thesis provided a literature review on methods to provide virtual iner-
tia for systems with high integration of RES by a thorough presentation on the
background involved. Multiple methods were discussed and objectively all the
methods strive to provide a dynamic frequency response via the converters. The
methodology can be selected based on system characteristics, limitations, level of
complexity and other goals. Some of the methods provide more detailed frequency
response to the converter (much similar to SG) while others can be as simple as
a delay.
The complications of the project involved the operation of the grid as a whole,
while considering the interaction of the grid components, tuning of the controllers
and priorities for control. Gathering data from the LEOGE platform, it was noted
that several loads are dependant on the presence of the gas turbine and therefore,
the gas turbine was essential while it had other responsibilities in the grid such as
fixing the voltage of the PCC. It was concluded that the loads of the platform can
be modeled as a whole without affecting the operation of the system, the inertia
emulation algorithm or the interactions with the wind farm.
The wind turbines were scaled down same as loads of the platform for redu-
cing the computation load on the PC used for simulation, without affecting the
approach, methodology and the results of the project. The type three connection
was used for wind turbines with DFIG. The switching converters were controlled
by SVPWM method with signals generated form PI controllers. An MPPT algorithm
was shown to be the most suitable option for controlling the RSC. The main goal
for the controllers of the DFIG converters were to be able to control the DFIG so
that it delivers almost constant power to the grid of the platform and supply the
base loads, while eliminating the need for larger gas turbines or diesel generators.
An ESS was proved to be the best option for proving the energy support and
129
emulating virtual inertia during a contingency event. The model consists of a hy-
pothetically infinite ESS with a switching converter, that is controlled according to
the changes in the grid frequency. The developed virtual inertia algorithm for the
system was selected such that, the complexity of the project and interactions of the
grid components would be straightforward. Each component was validated sep-
arately to better analyse its dynamics and effects on the system as a whole. Then
the components were assembled together and the final simulations considered the
interaction of all the grid elements.
The developed model successfully emulates virtual inertia in the concept model
of the oil and gas platform by extracting the necessary energy from the ESS to slow
down the ROCOF, while the wind turbine supplies the base loads of the platform.
Between the proposed methods for emulating inertia, the PI controlled appeared
to provide the best frequency response compared to cases with PI with dead-band,
and P controller. A derivative controller was not used to prevent the possible sta-
bility issues. For the PI controller to provide a better inertial response, more en-
ergy was extracted from the ESS. On the other hand, the gas turbine experienced
less stress and was able to maintain its rotor speed at a higher value while the
input mechanical power of the generator was also not saturated. Preventing the
saturation of the gas turbine input translates to less consumption of fuel for the
generator which is aligned with the goal of reducing the carbon emission of the
oil and gas platform (Although most of the contribution is because of introducing
the offshore wind turbines).
7.2 Future Work

Other than the methods explained in chapter 3, newer methods to provide vir-
tual inertia are being developed. A considerable number of these methods rely on
artificial intelligence and optimization algorithms. Some of which, aim to minim-
ize the frequency nadirs while injecting more energy to the grid. Some other strive
for reducing the stress on the ESS, maximising its lifetime or optimise the sizing
of the ESS.
The other methods to provide virtual inertia can be implemented on this model
and compared with the results of this thesis to evaluate the advantages and dis-
advantages of each one in this case. The addition of neural network based al-
gorithms, which are widely used in the literature, can outperform the conven-
tional PI controllers in speed and accuracy for controlling different parameters of
the system. Furthermore, the modeling of the platform itself, can be more detailed
and specific limitations and conditions applied due to the nature of different loads
on the structure (for example considering unbalanced load and its dynamics). The
variable wind speed can be easily incorporated to the final model. Although the
power output of the wind farm will be fluctuated around the nominal output, the
Chapter 7: Conclusions and Future Work 131
imbalance can be compensated with an external energy source. Therefore, with

modification of the control model for the ESS, this goal can be achieved. Also,
the scenarios can be defined more elaborately and discuss different contingency
events for the offshore platform and the virtual inertia strategy can then be ana-
lysed more scrutinisingly. Additional ancillary services are defined in grid codes
and they can be implemented in the modeling of the system, making the study
even more generalised. The study can be expanded to the type of ESS technology
used and additional goals for the ESS such as lifetime, sizing and so on can be
part of the study.
Bibliography
[1] Y. Zheng, Virtual inertia emulation in islanded microgrids with energy stor-
age system, master of science thesis, delft university, 2016. [Online]. Avail-
able: http://resolver.tudelft.nl/uuid:51849a15- 2f89- 447b- 8661-
ac5d12b44edb.
[2] F. Blaabjerg, R. Teodorescu, M. Liserre and A. Timbus, ‘Overview of control

and grid synchronization for distributed power generation systems,’ Indus-
trial Electronics, IEEE Transactions on, vol. 53, pp. 1398–1409, Nov. 2006.
DOI : 10.1109/TIE.2006.881997.
[3] P. Denholm, T. Mai, R. W. Kenyon, M. O’Malley and B. Kroposki. (2020).

‘Inertia and the power grid: A guide without the spin, nrel,’ [Online]. Avail-
able: https://www.nrel.gov/docs/fy20osti/73856.pdf..
[4] U. Tamrakar, D. Shrestha, M. Maharjan, B. P. Bhattarai, T. M. Hansen and
R. Tonkoski, ‘Virtual inertia: Current trends and future directions,’ Applied
Sciences, vol. 7, no. 7, 2017, ISSN: 2076-3417. DOI: 10.3390/app7070654.
[Online]. Available: https://www.mdpi.com/2076-3417/7/7/654.
[5] R. Eriksson, N. Modig and K. Elkington, ‘Synthetic inertia versus fast fre-
quency response: A definition,’ IET Renewable Power Generation, vol. 12,
Sep. 2017. DOI: 10.1049/iet-rpg.2017.0370.
[6] A. Fernández-Guillamón, E. Gómez-Lázaro, E. Muljadi and Á. Molina-García,
‘Power systems with high renewable energy sources: A review of inertia and
frequency control strategies over time,’ Renewable and Sustainable Energy
Reviews, vol. 115, p. 109 369, 2019. DOI: 10.1016/j.rser.2019.109369.
[7] W. He, K. Uhlen, M. Hadiya, Z. Chen, G. Shi and E. del Rio, ‘Case study of
integrating an offshore wind farm with offshore oil and gas platforms and
with an onshore electrical grid,’ Journal of Renewable Energy, vol. 2013,
pp. 1–10, 2013. DOI: 10.1155/2013/607165.
[8] H. G. Svendsen, A. Holdyk, I. R. Mosgren and J. Wiik, Deliverable report:
Dsp5_2020_05, test platform specification, modelling and operational optim-
ization results, 2020.
[9] A. E. Tutorials, Synchronous generator as a wind power generator, Dec. 2021.
[Online]. Available: https://www.alternative-energy-tutorials.com/
wind-energy/synchronous-generator.html.
133
[10] P. Kundur, Power system stability and control. McGraw-Hill.

[11] M. Khatibi, F. Rahmani and T. Agarwal, ‘Comparative Analysis of Power Sys-
tem Model Reduction,’ arXiv e-prints, arXiv:2004.04112, arXiv:2004.04112,
Apr. 2020. arXiv: 2004.04112 [eess.SY].
[12] Frequency control in a power system - technical articles. [Online]. Available:
https://eepower.com/technical-articles/frequency-control-in-a-
power-system/#.
[13] P. Tavner, What are the disadvantages of a pmsg compared to a dfig? Feb.
2019. [Online]. Available: https://www.researchgate.net/post/What-
are-the-disadvantages-of-a-PMSG-compared-to-a-DFIG.
[14] G. Abad, Doubly fed induction machine: modeling and control for wind energy
generation applications. IEEE Press, 2011.
[15] H. Abu-Rub, M. Malinowski and K. Al-Haddad, Power electronics for renew-
able energy systems, transportation, and industrial applications. Wiley/IEEE,
2014.
[16] B. Kroposki, B. Johnson, Y. Zhang, V. Gevorgian, P. Denholm, B.-M. Hodge
and B. Hannegan, ‘Achieving a 100% renewable grid: Operating electric
power systems with extremely high levels of variable renewable energy,’
IEEE Power and Energy Magazine, vol. 15, no. 2, pp. 61–73, 2017. DOI:
10.1109/mpe.2016.2637122.
[17] A. Fernández-Guillamón, E. Gómez-Lázaro, E. Muljadi and Á. Molina-García,

‘Power systems with high renewable energy sources: A review of inertia and
frequency control strategies over time,’ Renewable and Sustainable Energy
Reviews, vol. 115, p. 109 369, 2019. DOI: 10.1016/j.rser.2019.109369.
[18] H. Bevrani, T. Ise and Y. Miura, ‘Virtual synchronous generators: A survey
and new perspectives,’ International Journal of Electrical Power Energy Sys-
tems, vol. 54, pp. 244–254, 2014. DOI: 10.1016/j.ijepes.2013.07.009.
[19] N. B. Negra, J. Todorovic and T. Ackermann, ‘Loss evaluation of hvac and
hvdc transmission solutions for large offshore wind farms,’ Electric Power
Systems Research, vol. 76, no. 11, pp. 916–927, 2006, ISSN: 0378-7796.
DOI : https://doi.org/10.1016/j.epsr.2005.11.004. [Online]. Avail-
able: https://www.sciencedirect.com/science/article/pii/S0378779605002609.
[20] Wind turbine generator model validation (wtgmv). [Online]. Available: https:
//www.esig.energy/wiki-main-page/wind-turbine-generator-model-
validation-wtgmv/.
[21] S. Azmi, K. Ahmed, S. Finney and B. Williams, ‘Comparative analysis between

voltage and current source inverters in grid-connected application,’ IET
Conference on Renewable Power Generation (RPG 2011), 2011. DOI: 10 .
1049/cp.2011.0138.
Bibliography 135
[22] D. Pattabiraman, R. H. Lasseter. and T. M. Jahns, ‘Comparison of grid fol-

lowing and grid forming control for a high inverter penetration power sys-
tem,’ in 2018 IEEE Power Energy Society General Meeting (PESGM), 2018,
pp. 1–5. DOI: 10.1109/PESGM.2018.8586162.
[23] D. Pattabiraman, R. H. Lasseter. and T. M. Jahns, ‘Comparison of grid fol-
lowing and grid forming control for a high inverter penetration power sys-
tem,’ 2018 IEEE Power Energy Society General Meeting (PESGM), 2018. DOI:
10.1109/pesgm.2018.8586162.
[24] Q. Zhong, ‘Virtual synchronous machines: A unified interface for grid in-
tegration,’ IEEE Power Electronics Magazine, vol. 3, pp. 18–27, 2016.
[25] L. wang Shang, J. Hu, X. YUAN and Y. HUANG, ‘Improved virtual synchron-
ous control for grid-connected vscs under grid voltage unbalanced condi-
tions,’ Journal of Modern Power Systems and Clean Energy, vol. 7, pp. 1–12,
Apr. 2018. DOI: 10.1007/s40565-018-0388-2.
[26] W. Kenyon, A. Hoke, J. Tan, B. Kroposki and B.-M. Hodge, ‘Grid-following
inverters and synchronous condensers: A grid- forming pair?,’ Mar. 2020.
DOI : 10.1109/PSC50246.2020.9131310.
[27] G. Delille, B. Francois and G. Malarange, ‘Dynamic frequency control sup-

port by energy storage to reduce the impact of wind and solar generation
on isolated power system’s inertia,’ Sustainable Energy, IEEE Transactions
on, vol. 3, pp. 931–939, Oct. 2012. DOI: 10.1109/TSTE.2012.2205025.
[28] M. F. M. Arani and E. F. El-Saadany, ‘Implementing virtual inertia in dfig-
based wind power generation,’ IEEE Transactions on Power Systems, vol. 28,
no. 2, pp. 1373–1384, 2013. DOI: 10.1109/TPWRS.2012.2207972.
[29] N. Kakimoto, S. Takayama, H. Satoh and K. Nakamura, ‘Power modula-
tion of photovoltaic generator for frequency control of power system,’ IEEE
Transactions on Energy Conversion, vol. 24, no. 4, pp. 943–949, 2009. DOI:
10.1109/TEC.2009.2026616.
[30] K. M. Cheema, ‘A comprehensive review of virtual synchronous generator,’

International Journal of Electrical Power Energy Systems, vol. 120, p. 106 006,
2020. DOI: 10.1016/j.ijepes.2020.106006.
[31] Q. Zhong, P.-L. Nguyen, Z. Ma and W. Sheng, ‘Self-synchronized synchron-
verters: Inverters without a dedicated synchronization unit,’ IEEE Transac-
tions on Power Electronics, vol. 29, pp. 617–630, 2014.
[32] R. Hesse, D. Turschner and H. Beck, ‘Micro grid stabilization using the vir-
tual synchronous machine (visma),’ Renewable energy power quality journal,
vol. 1, pp. 676–681, 2009.
[33] Y. Chen, R. Hesse, D. Turschner and H. Beck, ‘Improving the grid power
quality using virtual synchronous machines,’ in 2011 International Confer-
ence on Power Engineering, Energy and Electrical Drives, 2011, pp. 1–6. DOI:
10.1109/PowerEng.2011.6036498.
[34] Y. Hirase, K. Abe, K. Sugimoto and Y. Shindo, ‘A grid connected inverter

with virtual synchronous generator model of algebraic type,’ IEEJ Trans-
actions on Power and Energy, vol. 132, pp. 371–380, Jan. 2012. DOI: 10.
1541/ieejpes.132.371.
[35] D. Singh and K. Seethalekshmi, ‘A review on various virtual inertia tech-

niques for distributed generation,’ in 2020 International Conference on Elec-
trical and Electronics Engineering (ICE3), 2020, pp. 631–638. DOI: 10.1109/
ICE348803.2020.9122959.
[36] K. Sakimoto, Y. Miura and T. Ise, ‘Stabilization of a power system with a

distributed generator by a virtual synchronous generator function,’ in 8th
International Conference on Power Electronics - ECCE Asia, 2011, pp. 1498–
1505. DOI: 10.1109/ICPE.2011.5944492.
[37] P. Rodriguez, I. Candela and A. Luna, ‘Control of pv generation systems us-
ing the synchronous power controller,’ 2013 IEEE Energy Conversion Con-
gress and Exposition, 2013. DOI: 10.1109/ecce.2013.6646811.
[38] M. van Wesenbeeck, S. de Haan, P. Varela and K. Visscher, ‘Grid tied con-
verter with virtual kinetic storage,’ Undefined/Unknown, in Proceedings
2009 IEEE Bucharest Power Tech Conference, s.n., Ed., null ; Conference
date: 28-06-2009 Through 02-07-2009, IEEE Society, 2009, pp. 1–7, ISBN:
978-1-4244-2235-7. DOI: doi:10.1109/PTC.2009.5282048.
[39] D. Shrestha, U. Tamrakar, Z. Ni and R. Tonkoski, ‘Experimental verification
of virtual inertia in diesel generator based microgrids,’ in 2017 IEEE Inter-
national Conference on Industrial Technology (ICIT), 2017, pp. 95–100. DOI:
10.1109/ICIT.2017.7913065.
[40] F. Katiraei and M. Iravani, ‘Power management strategies for a microgrid

with multiple distributed generation units,’ IEEE Transactions on Power Sys-
tems, vol. 21, pp. 1821–1831, 2006.
[41] N. Pogaku, M. Prodanovic and T. C. Green, ‘Modeling, analysis and testing
of autonomous operation of an inverter-based microgrid,’ IEEE Transactions
on Power Electronics, vol. 22, no. 2, pp. 613–625, 2007. DOI: 10.1109/TPEL.
2006.890003.
[42] J. M. Guerrero, L. G. de Vicuna, J. Matas, M. Castilla and J. Miret, ‘A

wireless controller to enhance dynamic performance of parallel inverters
in distributed generation systems,’ IEEE Transactions on Power Electronics,
vol. 19, no. 5, pp. 1205–1213, 2004. DOI: 10.1109/TPEL.2004.833451.
[43] W. Zhang, A. Cantarellas, J. Rocabert, A. Luna and P. Rodríguez, ‘Synchron-
ous power controller with flexible droop characteristics for renewable power
generation systems,’ IEEE Transactions on Sustainable Energy, vol. 7, pp. 1572–
1582, 2016.
Bibliography 137
[44] B. Johnson, S. Dhople, A. Hamadeh and P. Krein, ‘Synchronization of par-

allel single-phase inverters with virtual oscillator control,’ English (US),
IEEE Transactions on Power Electronics, vol. 29, no. 11, pp. 6124–6138, Nov.
2014, ISSN: 0885-8993. DOI: 10.1109/TPEL.2013.2296292.
[45] M. Ashabani, F. D. Freijedo, S. Golestan and J. M. Guerrero, ‘Inducverters:
Pll-less converters with auto-synchronization and emulated inertia capab-
ility,’ IEEE Transactions on Smart Grid, vol. 7, no. 3, pp. 1660–1674, 2016.
DOI : 10.1109/tsg.2015.2468600.
[46] P. Mangera, F. H. Sumbung and D. Parenden, ‘Automatic voltage regulator

(avr) controller design based on routh’s crution stability analysis in diesel-
based power plants,’ Proceedings of the International Conference on Science
and Technology (ICST 2018), 2018. DOI: 10.2991/icst-18.2018.113.
[47] P. V. D. Male and E. Lourens, ‘Operational vibration-based response estima-
tion for offshore wind lattice structures,’ Structural Health Monitoring and
Damage Detection, Volume 7 Conference Proceedings of the Society for Ex-
perimental Mechanics Series, pp. 83–96, 2015. DOI: 10.1007/978-3-319-
15230-1_9.
[48] A. Abrous, R. Wamkeue and E. M. Berkouk, ‘Modeling and simulation of

a wind model using a spectral representation method,’ 2015 3rd Interna-
tional Renewable and Sustainable Energy Conference (IRSEC), 2015. DOI:
10.1109/irsec.2015.7455141.
[49] M. Shinozuka and C.-M. Jan, ‘Digital simulation of random processes and
its applications,’ Journal of Sound and Vibration, vol. 25, no. 1, pp. 111–
128, 1972. DOI: 10.1016/0022-460x(72)90600-1.
[50] J. C. Kaimal, ‘Turbulenece spectra, length scales and structure parameters
in the stable surface layer,’ Boundary-Layer Meteorology, vol. 4, no. 1-4,
pp. 289–309, 1973. DOI: 10.1007/bf02265239.
[51] T. H. M. S. Rashid, A. Routh, M. R. Rana, A. I. Ferdous and R. Sayed, ‘A
novel approach to maximize performance and reliability of pmsg based
wind turbine: Bangladesh perspective,’ 2018.
[52] H. Al-Hamadani, T. An, M. King and H. Long, ‘System dynamic modelling of
three different wind turbine gearbox designs under transient loading con-
ditions,’ International Journal of Precision Engineering and Manufacturing,
vol. 18, no. 11, pp. 1659–1668, 2017. DOI: 10.1007/s12541-017-0194-1.
[53] O. Apata and D. Oyedokun, ‘An overview of control techniques for wind
turbine systems,’ Scientific African, vol. 10, 2020. DOI: 10.1016/j.sciaf.
2020.e00566.
[54] G. Dudgeon, Wind farm model in simscape: 140 wind turbines. [Online].
Available: https://www.mathworks.com/matlabcentral/fileexchange/
68597-wind-farm-model-in-simscape-140-wind-turbines.
[55] C. Bajracharya, M. Molinas, J. Suul and T. Undeland, ‘Understanding of

tuning techniques of converter controllers for vsc-hvdc,’ Jun. 2008.
[56] L. Ping, X. Jian, K. Yong, Z. Hui and C. Jian, ‘For prototype super-conducting
magnetic energy storage (smes) [voltage source convertor control scheme],’
Proceedings IPEMC 2000. Third International Power Electronics and Motion
Control Conference (IEEE Cat. No.00EX435), DOI: 10.1109/ipemc.2000.
884543.
[57] M. Vujacic, O. Dordevic and G. Grandi, ‘Evaluation of dc-link voltage ripple

in seven-phase pwm voltage source inverters,’ 2018 IEEE International Tele-
communications Energy Conference (INTELEC), 2018. DOI: 10.1109/intlec.
2018.8612327.
[58] M. Salcone and J. Bond, ‘Selecting film bus link capacitors for high per-
formance inverter applications,’ 2009 IEEE International Electric Machines
and Drives Conference, 2009. DOI: 10.1109/iemdc.2009.5075431.
[59] M. Anwar and M. Teimor, ‘An analytical method for selecting dc-link-capacitor
of a voltage stiff inverter,’ Conference Record of the 2002 IEEE Industry Ap-
plications Conference. 37th IAS Annual Meeting (Cat. No.02CH37344), DOI:
10.1109/ias.2002.1042651.
[60] H. Brantsæter, Ł. Kocewiak, A. R. Årdal and E. Tedeschi, ‘Passive filter design

and offshore wind turbine modelling for system level harmonic studies,’ En-
ergy Procedia, vol. 80, pp. 401–410, 2015. DOI: 10.1016/j.egypro.2015.
11.444.
Appendix
Low Emission Oil and Gas Open (LEOGO) Platform

The platform has reservoir water pressure support by injection of production water
and water from subsea reservoir. A gas lift is used to reduce the weight of fluid
column in borehole and reduce the bottomhole backpressure. The annual oil and
gas production of the platform is considered to be 3.14 and 1.57 Million Standard
cubic meters (M il.Sm3 ), respectively. The platform components are as followed
[8]:
• Wells:
– 15 production wells
– 15 water injection wells (used to increase reservoir pressure)
• Energy supply:
– 3 gas turbine generators

– waste heat covers heat demand
– Diesel generator for emergency backup power
• Main electric load:
– separator trains (compressors)

– gas lift compressor
– gas re-compressor (increasing gas pressure after separator train)
– oil export pumps (for transport to neighbour platform)
– water injection pumps (produced water and seawater)
– seawater lift pumps for injection
– living quarters and auxiliary loads
– drilling (rarely in operation on a production platform)
The platform has three voltage levels of 11kV, 690V and 400V supplied through
various energy sources. The power sources of the LEOGO platform consist of main
gas turbines, an Uninterrupted Power Supply (UPS) system and an emergency
diesel generator.
139
Table 1: Field characteristics parameter values
Parameter Value Comment

Height from 100 m Typical NCS value. Sea depth +
seabed to platform height
platform
Wellhead 6 MPa (60 bar) Typical NCS pressure: 25 MPa
pressure reservoir / 10 MPa well head (at
seabed)
Gas injec- 20 MPa injection Q in j is assumed proportional to
tion rate for pressure Q oil
gas lift Injection Rate
(Q in j )/Oil Pro-
duction (Q oil ) =
220
Separator 4 MPa Typical values are 3-5 MPa(inlet
train inlet pressure is controlled by choke
pressure valve)
Separator 2 MPa (gas)
train outlet 0.3 MPa (oil)
pressure 0.7 MPa (water)
Gas export 6 MPa Normally 3-5 MPa.
pressure Ca 10 MPa out of compressor
(and drop to 6 MPa at export
point)
Oil export 3 MPa Ca 5 MPa out of oil pump (and 2
pressure MPa pressure drop in pipe)
Water in- 6 MPa Somewhat higher than wellhead
jection pressure
pressure
Well stream 50 Sm3 /s sum from all wells into mani-
flow fold/separator
Well stream Gas Oil Ratio These numbers have been adjus-
composition (GOR) = 500 ted to match the gas/oil/water
Water Cut (WC) production:
= 0.6 gas: 49.75 Sm3 /s = 4.3 mill
Sm3 /d a y
oil: 0.1 Sm3 /s = 8640 Sm3 /day
water: 0.15 Sm3 /s = 12960
kSm3 /d a y
(1 Sm3 = 6.2898 barrels)
Pipe diamet- 200 mm Used to compute pipe friction
ers (and hence pressure drop and
pump/compressor power de-
mand)
Appendix 141
Figure 1: LEOGO platform electrical grid
Figure 2: LEOGO platform-name of each component

142
Table 2: Generator data for the LEOGO platform
Generators
Basic AVR Governer and Engine
Tag Name Description Nominal frequency Nominal voltage Nominal generator power Nominal engine power Template Template
Gen1 Gas turbine generator 50 11kV 28 MW 21.8 MW IEEE_ AC8B IEEE_GGOV1
gen2 Gas turbine generator 50 11kV 28 MW 21.8 MW IEEE_ AC8B IEEE_GGOV1
Gen3 Gas turbine generator 50 11kV 28 MW 21.8 MW IEEE_ AC8B IEEE_GGOV1
Gen_EME_1 Diesel generator 50 690V 2.8 MW 21.8 MW IEEE_ AC8B IEEE_GGOV1

142
Appendix
Table 3: Transformer data for the LEOGO platform
Transformers
Tag Name Description Nominal power Primary nom. voltage Secondary nom. voltage Nominal Primary Primary Secondary Secondary
frequency reactance resistance reactance resistance
TRA_EME_01 1.6 MVA 11/0.69 kV Dyn11 ASEA-mod.TypTr2 1.6 MW 11 kV 690 V 50 0.00396875 0.00396875 0.029736325 0.00396875
TRA_EME_02 1.6 MVA 11/0.69 kV Dyn11 ASEA-mod.TypTr2 1.6 MW 11 kV 690 V 50 0.00396875 0.00396875 0.029736325 0.00396875
TRA_Dril_01 3.3 MVA 11/0.69 kV 1.6 GEAFOL_modified.TypTr2 3.3 MW 11 kV 690 V 50 0.0549806 0.001460315 0.0549806 0.001460315
TRA_Dril_02 3.3 MVA 11/0.69 kV 1.6 GEAFOL_modified.TypTr2 3.3 MW 11 kV 690 V 50 0.0549806 0.001460315 0.0549806 0.001460315
TRA_UTL_01 0.8 MVA 11/0.69 kV Dyn11 ASEA_mod.TypTr2 0.8 MW 11 kV 690 V 50 0.02961148 0.0048125 0.02961148 0.0048125
TRA_UTL_02 0.8 MVA 11/0.69 kV Dyn11 ASEA_mod.TypTr2 0.8 MW 11 kV 690 V 50 0.02961148 0.0048125 0.02961148 0.0048125
TRA_UTL_03 0.6 MVA 0.69/0.4 kV Transformer-mod.TypTr2 0.6 MW 690 V 400 V 50 0.02961148 0.0048125 0.02961148 0.0048125
TRA_UTL_04 0.6 MVA 0.69/0.4 kV Transformer-mod.TypTr2 0.6 MW 690 V 400 V 50 0.02961148 0.0048125 0.02961148 0.0048125
143
143
144
Table 4: VSD load data for the LEOGO platform
VSD load
Tag Name Description Nom- VSD VSD Nom- Nom- Nom- Nom- Nom- Pole In- Start- Synch. Use Load
inal power effi- inal inal inal inal inal pairs er- ing speed Load Power
voltage factor ciency act- voltage fre- power effi- tia torque load Power
ive quency factor ciency con- torq
power stant
VSD_OEX_01 50Hz\11 kV\1500 kW/11 kV/1491.TypAsmo 11 0.9 95.8 1.5 11 50 0.89 95 1 0.86 0.66 0.88 Yes 0.38
kV MW kV MW
VSD_OEX_02 50Hz\11 kV\1500 kW/11 kV/1491.TypAsmo 11 0.9 95.8 1.5 11 50 0.89 95 1 0.86 0.66 0.88 Yes 0.38
kV MW kV MW
VSD_REC_01 50Hz\11 kV\7860 kW/11 kV/1482.TypAsmo 11 0.91 96.1 7.86 11 50 0.8 95 1 0.8 0.55 0.91 Yes 5.8
kV MW kV MW
VSD_REC_02 50Hz\11 kV\7860 kW/11 kV/1482.TypAsmo 11 0.91 96.1 7.86 11 50 0.8 95 1 0.8 0.55 0.91 Yes 5.8

kV MW kV MW
VSD_SEP_01 50Hz\11 kV\7860 kW/11 kV/1482.TypAsmo 11 0.91 96.1 7.86 11 50 0.8 95 1 0.8 0.55 0.91 Yes 3.53
kV MW kV MW
VSD_WIN_01(1) 50Hz\11 kV\4800 kW/11 kV.TypAsmo 11 0.88 96.3 4 11 50 0.8 95 1 0.8 0.7 0.88 Yes 3.4
kV MW kV MW
kV MW kV MW
kV MW kV MW
VSD_WST_01 50Hz\11 kV\4800 kW/11 kV.TypAsmo 11 0.88 96.3 4 11 50 0.8 95 1 0.8 0.7 0.86 Yes 4.4
kV MW kV MW
VSD_WST_02 50Hz\11 kV\4800 kW/11 kV.TypAsmo 11 0.88 96.3 4 11 50 0.8 95 1 0.8 0.7 0.86 Yes 4.4
kV MW kV MW
VSD_DRL_01 50Hz\1000 kW/0.69 kV/1.TypAsmo 690 0.92 96.8 1 690 50 0.8 95 1 0.8 0.7 0.92 No 0
V MW V
V MW V
V MW V
V MW V
V MW V
V MW V
144
Appendix
Table 5: PQ load data for the LEOGO platform
PQ load
Tag Name Description Nominal Nominal Nominal Nominal Load Voltage de- Minimum
voltage active power reactive power factor mode pendency voltage
power
LQ1 General 400 0.5 MW 0 0.99 1 0 0.1

Load
LQ2 LV load 400 0.5 MW 0 0.99 1 0 0.1
Load Utility MV Load 690 3 MW 0 0.9 1 0 0.1

A
Load Utility MV Load 690 4 MW 0 0.9 1 0 0.1

B
Load EME_1 General 690 0.5 MW 0.1 MVAR 0.98 1 0 0.1

load
Load EME_2 General 690 0.5 MW 0.1 MVAR 0.98 1 0 0.1

load
145
145
146
Table 6: Fixed speed motors data for the LEOGO platform
Motors- fixed speed
Basic Motor Control Load torque
Tag Name Description Nominal ap- Nominal frequency Nominal voltage Template Starting torque Synch. Speed Use Load Load Power Torque Torque
parent power load torq. Power speed dep., speed dep.,
xF xL
ASM_EME_01 50Hz\500 563190 W 50 690 V Default 0.3 0.88 No 0 2 5

kW/0.69
kV/1.TypAsmo

kW/0.69
kV/1.TypAsmo

kW/0.69
kV/1.TypAsmo
ASM_SWL_01 50Hz\11 889648 W 50 11 kV Default 0.3 0.87 Yes 0.75 MW 2 5

kV\750 kW/11
kV.TypAsmo

kV\750 kW/11
kV.TypAsmo

kV\750 kW/11
kV.TypAsmo
ASM_ACO_01 \50Hz\11 3485819 W 50 14 kV Default 0.3 0.89 Yes 1.75 MW 2 5

kV\3000 kW/11
kV.TypAsmo
ASM_ACO_02 \50Hz\11 3485819 W 50 15 kV Default 0.3 0.89 Yes 1.75 MW 2 5

kV\3000 kW/11
kV.TypAsmo
146

No - Ntnu Inspera 79786156 64576396

Uploaded by

Copyright:

Available Formats

No - Ntnu Inspera 79786156 64576396

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

No - Ntnu Inspera 79786156 64576396

Uploaded by

Copyright:

Available Formats

Mohammad Heidari

Control Models for Providing Virtual

Electrical Grid of Offshore Platforms

Master’s thesis in Renewable Energies in the Marine Environment

Co-supervisor: Salvatore D'Arco

Control Models for Providing Virtual

Master’s thesis in Renewable Energies in the Marine Environment

Norwegian University of Science and Technology

A well-built grid can withstand a contingency event and have an acceptable

The proposed algorithm is implemented on a voltage source inverter that ex-

3.3 Swing Equation Based Methods . . . . . . . . . . . . . . . . . . . . . . 41

6.3 Case c: DFIG with RSC and GSC . . . . . . . . . . . . . . . . . . . . . . 93

2.1 LEOGO platform main components and mass/energy flow [8] . . . 6

3.1 Methods to provide virtual inertia . . . . . . . . . . . . . . . . . . . . . 33

4.1 Project study case concept . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5.1 A typcial AVR model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6.1 Wind turbine package simulation model . . . . . . . . . . . . . . . . . 81

6.8 Case b1- Wind speed (m/s) . . . . . . . . . . . . . . . . . . . . . . . . . 85

6.51 Case c2- Electromagentic torque (N.m) . . . . . . . . . . . . . . . . . . 103

1 LEOGO platform electrical grid . . . . . . . . . . . . . . . . . . . . . . . 141

2.1 Switching states and line voltages . . . . . . . . . . . . . . . . . . . . . 27

3.1 The operation modes of the self-synchronised synchronverter . . . . 35

5.1 Gas turbine generator data . . . . . . . . . . . . . . . . . . . . . . . . . 54

1 Field characteristics parameter values . . . . . . . . . . . . . . . . . . 140

M il.Sm3 Million Standard cubic meters. 139

Q in j Injection Rate. 140

Q oil Oil Production. 140

AVR Automatic Voltage Regulator. 39, 40, 54, 55

CSC Current Source Converter. 17

CSV Constant Stator Voltage. 70

DCF Distortion Component factor. 36

DSP Digital Signal Processing. 36

ENTSOE European Network of Transmission System Operators for Electricity. 17

GOR Gas Oil Ratio. 140

GTO Gate Turn Off transistor. 17

IEPE Institute of Electrical Power Engineering. 35, 36, 49

IGBT Insulated Gate Bipolar Transistors. 14, 17

IGCT Integrated Gate Commutated Thyristor. 14

KHI Kawasaki Heavy Industries. 39, 40, 49

LEOGO Low Emission Oil and Gas Open. 5, 6, 53, 139

MOSFET Metal Oxide Semiconductor Field Effect Transistor. 14, 17

NCS Norwegian Continental Shelf. 5

NREL National Renewable Energy Lab (US). 1

NTNU Norges teknisk-naturvitenskaplige universitet. 2

PFCC power Factor Correction Capacitor. 16

PMSG Permanent Magnet SG. 7, 9, 16

PSD Power Spectral Density. 59

R&D Research and Development. 6

RES Renewable Energy Source. 1, 14, 18, 31, 129

ROCOF Rate of Change of Frequency. 1, 9, 13, 44, 45, 125, 130

SCIG Squirrel Cage Induction Generator. 15, 16

SMES Superconducting Magnetic Energy Storage. 17

SOC State of Charge. 44

SPC Synchronous power Controller. 43, 49

SRM Spectral Representation Method. 59

THD Total harmonic Distortion. 20

TSR Tip Speed Ratio. 65, 68

UPF Unity power Factor. 70

UPS Uninterrupted Power Supply. 139