You are on page 1of 115

Faculty of Electrical Engineering, Mathematics &

Computer Science
Delft University of Technology

Transient Behaviour of Grounding Grids

Author: Michail Kouros

In partial fulfilment of the requirements for the degree of


MSc Electrical Power Engineering

Thesis committee:
Dr.ir. M. Popov
Prof.ir. M.A.M.M. van der Meijden
Dr. M. Ghaffarian Niasar
I. Tannemaat

Thesis Defense
July 2020
Transient Behaviour of Grounding Grids

TU Delft Supervisor:
Author: Dr. Marjan Popov
Michail Kouros
Student number: TenneT Supervisors:
4920260 Imre Tannemaat
Kostas Velitsikakis

Arnhem, The Netherlands


Acknowledgments
Initially, I would like to thank my university supervisor, Dr. Marjan Popov, for supervis-
ing this work and for all the significant feedback he provided me with.

Moreover, I would like to express my sincere gratitude to my supervisor, Imre Tannemaat.


I fully appreciate that Imre was always there to help and guide me during this project and
during my first project at TenneT. His help was extremely crucial for finalising this report.

In addition, I would also like, to thank Kostas Velitsikakis and Chris Engelbracht for their
remarks during this project as well as Bas Hoeijmakers for his guidance throughout my
first project at TenneT.

Finally, I am very thankful to my family for providing me with their encouragement and
love throughout my life.

i
Abstract
To ensure a safe power system which operates correctly, a proper grounding system is
necessary to provide connections to the ground at specific points. A grounding system
is generally designed: (1) to reduce voltages occurring in the grounding system for short
circuit situations (at power frequency) to an acceptable level in terms of step and touch
voltages and to prevent damage to systems, (2) to provide good connection and a con-
ducting path for the lightning protection system to function correctly and (3) for Electro-
Magnetic Compatibility (EMC) based on generic rules to reduce the risk of damage and
interference of systems.

More specifically, the grounding system is crucial for the good protection of the primary
components (mainly the power transformer) against overvoltages caused by lightning
strikes, usually with the aid of surge arresters. For the Dutch power system, the ground-
ing system has until now been represented as a simple resistance to ground for transient
studies. However, literature suggests that this fails to take into account the properties of a
grounding system and that their transient behaviour should be studied more extensively.

The lightning response of grounding grids can be assessed by means of simulations and
by experimentation results. To perform such studies, TenneT has different software tools
available. By using the XGSLab software tool, the behaviour of grounding systems was
studied. Thereafter, its transient behaviour was examined using the EMTP/ATP software
tool. Using this software, a range of different simulations were conducted, such as: sim-
ple grounding conductors, a grounding system of high voltage towers and, finally, an
existing substation model.

ii
Contents

Acknowledgements i

Abstract ii

List of Figures vi

List of Tables ix

1 Introduction 2
1.1 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Lightning Phenomena and their Effect on Grounding Systems 5


2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Lightning discharge effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Different lightning current waveshapes . . . . . . . . . . . . . . . . . . . . 7
2.3.1 The Heidler Waveshape . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Lightning protection scheme-earth termination system . . . . . . . . . . 10
2.4.1 Type A arrangements . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4.2 Type B arrangements . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4.3 Lightning surges on A type arrangements . . . . . . . . . . . . . . 11

3 Methods to Study the Transient Behaviour of Grounding Systems 13


3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Transmission Line Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Circuit Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.4 Electromagnetic Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.5 Hybrid Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.5.1 Applied software: XGSLab Ver 9.3.1.1 . . . . . . . . . . . . . . . . . 21

4 Broad Frequency Behaviour of Grounding Systems 24


4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 Frequency dependency of soil and soil ionization . . . . . . . . . . . . . . 24
4.3 Behaviour of grounding systems when conducting lightning currents . . 25
4.3.1 Grounding systems with constant soil parameters . . . . . . . . . 26
4.3.1.1 Vertical Conductor . . . . . . . . . . . . . . . . . . . . . . 26

iii
CONTENTS iv

4.3.1.2 Horizontal Conductor . . . . . . . . . . . . . . . . . . . . 30


4.3.1.3 Ring Conductor . . . . . . . . . . . . . . . . . . . . . . . . 32
4.3.1.4 Small grounding grid . . . . . . . . . . . . . . . . . . . . 35
4.3.2 Grounding systems with frequency dependent soil . . . . . . . . . 37
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5 Modelling Approach - Frequency Dependent Models 39


5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 Fitting Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.2.1 Vector Fitting Method . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.2.2 ARMAFIT Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3 Modelled Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.1 Vertical Conductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.1.1 Case A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.3.1.2 Case B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3.2 Ring Conductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.3.2.1 Case A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.3.2.2 Case B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.3.3 High Voltage Towers . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3.3.1 DTC-NDR380 Wintrack Tower 18 . . . . . . . . . . . . . 46
5.3.3.2 DTC-NDR380 Wintrack Tower 19 . . . . . . . . . . . . . 47
5.3.4 Boxtel 150kV Substation . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.3.4.1 Network Approach, N-ports system . . . . . . . . . . . . 51
5.3.4.2 Surge arrester at location A . . . . . . . . . . . . . . . . . 54
5.3.4.3 Surge arrester at location B . . . . . . . . . . . . . . . . . 54
5.3.4.4 Surge arrester at location C . . . . . . . . . . . . . . . . . 55
5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6 Transient Simulations - Results 58


6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.2 Vertical Conductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.2.1 Time domain comparison XGSLab - EMTP/ATP . . . . . . . . . . . 59
6.2.2 EMTP/ATP comparison of frequency dependent model and resistor 60
6.3 Ring Conductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.3.1 Time domain comparison: XGSLab and EMTP/ATP . . . . . . . . 61
6.3.2 EMTP/ATP comparison of frequency dependent model and resistor 62
6.4 High Voltage Towers - Doetinchem line . . . . . . . . . . . . . . . . . . . . 63
6.4.1 EMTP/ATP comparison of frequency dependent model and resistor 63
6.4.2 Effect of different tower footing impedance values . . . . . . . . . 65
6.5 Boxtel150kV Substation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.5.2 Model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.5.3 Calculation results direct strike . . . . . . . . . . . . . . . . . . . . 69
6.5.3.1 Different surge arrester locations . . . . . . . . . . . . . 69
6.5.3.2 Different grounding structure options . . . . . . . . . . 70
6.5.4 Calculation results backflashover . . . . . . . . . . . . . . . . . . . 71
CONTENTS v

6.5.4.1 Different surge arrester locations . . . . . . . . . . . . . 71


6.5.4.2 Different grounding structure options . . . . . . . . . . 71
6.6 Improve the topology of the grounding grid to reduce overvoltages . . . . 72
6.6.1 Addition of vertical electrode(s) at power transformer . . . . . . . 72
6.6.1.1 Two electrodes of 15m . . . . . . . . . . . . . . . . . . . . 72
6.6.1.2 Effective length of an electrode . . . . . . . . . . . . . . . 73

7 Conclusions 77

References 79

Appendix A - Models of Frequency Dependency of Soil 83

Appendix B - ARMAFIT Instructions 88

Appendix C - Vector Fitting Values 90

Appendix D - Kizilcay Models 91

Appendix E - Heidler Function as Defined in MODELS 93

Appendix F - Double Peaked Waveshape 93


List of Figures

1.1 Different types of grounding structures . . . . . . . . . . . . . . . . . . . . 3

2.1 Different types of discharge. Adapted from encyclopedia Britannica . . . 5


2.2 The four different types of cloud to ground discharge. Adapted from [2] . 6
2.3 Representations of first return stroke currents . . . . . . . . . . . . . . . . 8
2.4 Standard lightning waveshape defined with the Heidler function. Adapted
from [11] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Length of vertical/horizontal conductor as function of the soil resistivity,
for different classes of lightning. Adapted from [11] . . . . . . . . . . . . . 11
2.6 Equivalent high frequency lumped circuit for a vertical conductor. Adapted
from [12] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.1 Transmission line model. Adapted from [17] . . . . . . . . . . . . . . . . . 15


3.2 Circuit representation of a short length conductor. Adapted from [21] . . 16
3.3 Circuit representation for each segment. Adapted from [22] . . . . . . . . 17
3.4 Equivalent circuit of each segment. Adapted from[29] . . . . . . . . . . . . 20
3.5 Impedance of the grounding grid as function of the frequency. a) For a
lightning discharge at the corner of the grid, b) For a lightning discharge
at the middle of the grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.6 GPR of the grounding grid. a) For a lightning discharge at the corner of
the grid, b) For a lightning discharge at the middle of the grid . . . . . . . 23
3.7 Influence of the earth termination connection to the structure impedance
to ground: a) Horizontal conductor, b) Grounding grid-energization at the
center. Adapted from [3] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4.1 Grounding structures modelled: a) Vertical Conductor, b) Horizontal Con-


ductor c) Ring Conductor, d) Grounding Grid . . . . . . . . . . . . . . . . . 26
4.2 The effect of soil resistivity on the impedance of the vertical conductor . 27
4.3 The effect of permittivity on the impedance of the vertical conductor as
function of the frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4 Impedance and phase angle as function of frequency for different soil re-
sistivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.5 Impact of the permittivity on the vertical conductor impedance at high
frequency (10 MHz) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.6 The effect of soil resistivity on the impedance of the horizontal conductor 31

vi
LIST OF FIGURES vii

4.7 The effect of permittivity on the impedance of the horizontal conductor


as function of the frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.8 Impact of the permittivity on the impedance of the horizontal conductor
at high frequency (10 MHz) . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.9 The effect of soil resistivity on the impedance of the ring conductor . . . 33
4.10 The effect of permittivity on the impedance of the ring conductor as func-
tion of the frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.11 Impact of the permittivity on the impedance of a ring conductor at high
frequency (10 MHz) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.12 The effect of soil resistivity on the impedance of the grounding grid . . . 35
4.13 The effect of permittivity on the impedance of the grounding grid as func-
tion of the frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.14 Impact of the permittivity on the impedance of a grounding grid at high
frequency (10 MHz) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.15 Comparison of frequency dependent soil with a soil with constant soil pa-
rameters of ρ = ρ 0 and ²r = 10 . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5.1 Flowchart of the steps taken . . . . . . . . . . . . . . . . . . . . . . . . . . . 39


5.2 Perform frequency scan using ATPdraw . . . . . . . . . . . . . . . . . . . . 41
5.3 Equivalent frequency dependent model for a vertical conductor, case A . 42
5.4 Frequency response of a vertical conductor, comparison of XGSLab and
ATP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.5 Equivalent frequency dependent model for a vertical conductor, case B . 43
5.6 Frequency response of a vertical conductor, comparison of XGSLab and
ATP, Case B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.7 Ring conductor, XGSLab and ATP comparison, Case B . . . . . . . . . . . 45
5.8 Ring conductor, XGSLab and ATP comparison, Case B . . . . . . . . . . . 45
5.9 Ring conductor impedance - Case B . . . . . . . . . . . . . . . . . . . . . . 46
5.10 Grounding system of Wintract tower 18 . . . . . . . . . . . . . . . . . . . . 47
5.11 High voltage tower 18, XGSLab and ATP comparison . . . . . . . . . . . . 47
5.12 Grounding system of Wintract tower 19 . . . . . . . . . . . . . . . . . . . . 48
5.13 Grounding grid of Boxtel 150kV substation along with the locations of cur-
rent injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.14 Drawing of Boxtel substation . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.15 Example of an 1-port system . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.16 Example of a 2-port system . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.17 π circuit approach, 2-port system . . . . . . . . . . . . . . . . . . . . . . . . 52
5.18 Transformer point impedance (Z22) . . . . . . . . . . . . . . . . . . . . . . 53
5.19 Substation - Case A, XGSLab and ATP comparison . . . . . . . . . . . . . . 54
5.20 Substation - Case B, XGSLab and ATP comparison . . . . . . . . . . . . . . 55
5.21 Substation - Case C, XGSLab and ATP comparison . . . . . . . . . . . . . . 56
5.22 Grounding grid as defined in the EMTP/ATP software . . . . . . . . . . . . 57

6.1 Time domain comparison between XGSLab and ATP . . . . . . . . . . . . 59


6.2 Ground potential rise for the vertical conductor for different lightning wave-
shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
LIST OF FIGURES viii

6.3 Heidler function as defined in EMTP/ATP . . . . . . . . . . . . . . . . . . . 61


6.4 Time domain comparison between XGSLab and ATP . . . . . . . . . . . . 61
6.5 Ground potential rise for the ring conductor for different lightning wave-
shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.6 Transmission line system of Doetinchem - 380kV, 7 Wintrack towers con-
nected . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.7 Overvoltages in a 380kV/150kV transmission line system, for an lightning
waveshape of 31kA (4.2/70 µs) . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.8 Overvoltages in a 380kV/150kV transmission line system,for an lightning
waveshape of 31kA (4.2/70 µs) . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.9 Overvoltages in a 380kV/150kV transmission line system, for different soil
resistivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.10 Tower overvoltages, worst case scenario . . . . . . . . . . . . . . . . . . . . 67
6.11 Substation network as defined in the EMTP software tool . . . . . . . . . . 68
6.12 Impedance value for different placement of surge arrester . . . . . . . . . 69
6.13 Overvoltages at transformer and surge arrester for a discharge current of
10 kA and 1.2/50 µs directly at phase A - for surge arrester at different
positions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.14 Overvoltages at transformer and surge arrester for a discharge current of
20 kA and 1.2/50 µs- comparison between different soils . . . . . . . . . . 70
6.15 Overvoltages at transformer and surge arrester for a discharge current of
100 kA and 1.2/50 µs at shielding wires - for surge arrester at different
positions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.16 Overvoltages at transformer and surge arrester for a discharge current of
100 kA and 1.2/50 µs, at the shielding wires . . . . . . . . . . . . . . . . . . 72
6.17 Transformer point impedance - for additional conductors in the ground-
ing grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.18 Effect of different lengths of vertical conductors - for different soil resis-
tivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.19 Transformer point impedance - for additional vertical conductors with ef-
fective length in the grounding grid . . . . . . . . . . . . . . . . . . . . . . . 75
6.20 Voltage decrease in the transformer terminal, by optimizing the ground-
ing grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
List of Tables

2.1 Standard lightning parameters . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.1 Aspects taken into account in the used software . . . . . . . . . . . . . . . 22

4.1 Typical values of soil permittivity . . . . . . . . . . . . . . . . . . . . . . . . 25

5.1 Characteristics of the Vertical Conductor for two cases . . . . . . . . . . . 42


5.2 Characteristics of the Ring Conductor for two cases . . . . . . . . . . . . . 44
5.3 Characteristics of high voltage towers . . . . . . . . . . . . . . . . . . . . . 46

6.1 Characteristics of the Heidler function for XGSLab and EMTP . . . . . . . 59


6.2 Distance of surger arrester from transformer . . . . . . . . . . . . . . . . . 68

ix
LIST OF TABLES 1

List of Abbreviations
EMTP: ElectroMagnetic Transient Program
TD: Time Domain
FD: Frequency Domain
LPL: Lightning Protection Levels
LPS: Lightning Protection Scheme
GPR: Ground Potential Rise
TLM: Transmission Line Method
FEM: Finite Element Method
MoM: Method of Moments
VF: Vector Fitting
1
Introduction
The Dutch Transmission System Operator, TenneT TSO B.V and consulting companies,
conduct electromagnetic transient studies, for the Dutch high voltage grid in terms of
insulation coordination. In these studies, simulations are performed to determine if
the limits of the components are exceeded and to find ways to mitigate the overvolt-
ages at acceptable costs. Currently, for transient studies, the grounding system is usually
modelled as a simple resistor or as an ideal ground. However, literature has recently
shown that this can greatly underestimate the genenerated overvoltages in the system
[1]. Therefore, for this thesis the broad frequency behaviour of complex grounding sys-
tems is studied using the software XGSLab. Thereafter, by using a proven methodology,
they are modelled on the EMTP/ATP software as frequency dependent impedances and
highly accurate transient simulations are carried out.

Grounding systems have always existed in conjunction with power systems to provide
an electrical connection to earth so that currents are dissipated into the ground without
building unacceptable voltages or current concentrations. This ensures a safe environ-
ment and a lower risk of damage to the equipment. An illustration of some common
grounding structures is shown in Figure 1.1.

2
1. Introduction 3

Figure 1.1: Different types of grounding structures

A grounding system is considered to be effective when its impedance to ground is small


or ideally zero. There are many different ways to measure the impedance of a ground-
ing electrode (or grounding grid) and study its performance under different frequency
conditions. In this thesis, it will be shown that the impedance variation at different fre-
quencies is caused by multiple effects, such as:

• Soil parameters

• The frequency dependency of soil

• Soil Ionization

• The grounding structure and its size

• The amplitude and shape of the impulse current (frequency) and its feeding point

1.1. Research objectives


Based on the research goals and the findings from the literature review, the following
research questions were addressed and answered.

• What is an appropriate and detailed way of modelling grounding systems, such as:
tower footing and grounding grids with the available software?

– In what way can the frequency dependent model be defined in ATP/EMTP?


– What is an appropriate method to model the grounding grid (circuit, trans-
mission line, electromagnetic, hybrid)?

• What is the impact of the transient behaviour of grounding grids on the protection
of power transformers against lightning overvoltages?
1. Introduction 4

– What is the impact of the size of the grounding grid?


– What is the impact of the frequency dependency of soil and soil ionization
on the grounding grid?
– How can all relevant points of a grounding grid be modelled on the transient
software EMTP/ATP

• Which practical methods can be used to optimize the system configuration for
optimal protection of the power transformer against lightning overvoltages?

1.2. Thesis outline


A brief description of the content of each chapter is presented below:
• Chapter 2: In this chapter, the basics of lightning are explained; the different cur-
rent waveshapes that it can be represented by; a basic description of the Lightning
Protection Scheme (LPS).

• Chapter 3: In this chapter, the most well-known approaches for modelling ground-
ing structures for transient studies are described. Currently, there are four popular
ways that are in use. The software tool used for this thesis is based on the Hybrid
method (this method is a combination of the other three methods). Finally, a de-
scription of the XGSLab software tool is given.

• Chapter 4: In this chapter, the behaviour of the different effects that influence the
grounding systems when conducting currents into the earth are explained; Fur-
thermore, the behaviour of grounding systems is explained for: a) soils with con-
stant electrical parameters, b) soils that depend on frequency. Finally, by consid-
ering the different effects and the behaviour of the grounding systems, a decision
on what modelling assumptions will be used for the later chapters.

• Chapter 5: In this chapter, a description of the modelling assumptions is given; a


short description of the methods (Vector Fitting (VF) and ARMAFIT), which are
used to define equivalent frequency dependent models for transient studies in
EMTP/ATP; Finally, a comparison between the results from XGSLab and EMT-
P/ATP to validate the developed models.

• Chapter 6: In this chapter, the most significant results are shown by making use
of the frequency dependent (FD) models developed in chapter 5. By using these
models, the results obtained in XGSLab in time domain is compared to the ones
obtained in EMTP/ATP. Furthermore, the importance of defining FD models is
shown by comparing the models developed in chapter 5 with an equivalent resis-
tance. Finally, for the considered substation the following things are investigated:
1) What is the optimal place to install a surge arrester to protect the power trans-
former. 2) What is an effective way to reduce the generated overvoltages at the
transformer terminal.

• Conclusions: In this chapter, the most significant results during this study are sum-
marized along with future possible/recommended work.
2
Lightning Phenomena and their
Effect on Grounding Systems

2.1. Introduction
In this chapter, some general information about lightning is provided, such as: the dif-
ferent types of lightning; how often they occur; their direction-their polarity; the differ-
ent standard waveshapes; the remain sections concern the Lightning Protection Scheme
(LPS) as explained in the IEC and lightning in grounding conductors.

2.2. Lightning discharge effect


The prime source of a lightning discharge is a thundercloud where the electrification of
numerous cloud particles takes place. Lightning discharges that neutralize the charge
inside a thundercloud can be categorized into four different categories. As seen from
Figure 2.1, these are: intracloud, cloud to ground, intercloud and cloud to air.

Figure 2.1: Different types of discharge. Adapted from encyclopedia Britannica

5
2. Lightning Phenomena and their Effect on Grounding Systems 6

From the four discharge methods, intracloud and cloud to ground discharge occur more
often compared to the others. However, cloud to ground discharge is the one responsible
for most lightning damage, injury and death [2]. For this reason, scientists have been
more concerned with this type of lightning. Cloud to ground discharge is divided into
four different types of stroke based on the direction of the initial breakdown and by the
polarity of the charge transfer, as seen from the clouds to ground.
1. Downward Negative
2. Upward Negative
3. Downward Positive
4. Upward Positive

Figure 2.2: The four different types of cloud to ground discharge. Adapted from [2]

The most common cloud to ground flash is the negative downward lightning, shown in
category 1 of Figure 2.2. This type of lightning occurs 90% of the time whereas the re-
maining 10% consists of the other three categories seen in the same figure. Electromag-
netic measurements in the field have proved that a downward negative flash is launched
2. Lightning Phenomena and their Effect on Grounding Systems 7

by an electrical breakdown process. This leads to a column of charge, called ’stepped


leader’. While the stepped leader is moving towards the earth, it might result in several
branches during its step and when it reaches a height ranging from tens to hundred me-
tres from the ground the electric field at the tip of the grounded structures may increase
to an extent that electrical discharges begin to initiate. These discharges will grow in
the opposite direction, and once a connection with the leader is successfully initiated,
the charge will result in a flow of a lightning current. When the connection between the
ground and the stepped leader is made, a wave of near zero ground potential will travel
along the channel at a speed close to the speed of light [3].

Finally, flashes which are directed upwards (category 2, 4) are usually initiated from tall
structures, making them less frequent compared to the downward positive seen in cat-
egory 3. Negative lightning flashes usually consist of more than one stroke (subsequent
strokes) as compared to positive flashes which usually consist of only one. The first
stroke of negative lightning flashes has usually an amplitude that is two times higher
than that of the subsequent strokes. For the protection of power systems from lightning
surges, usually only the downward first negative lightning return-stroke current is con-
sidered [4].

2.3. Different lightning current waveshapes


A great number of measurements to find the most accurate lightning characteristics have
been performed during the last years, resulting in different proposals for waveshapes.
The most frequently used waveshapes are described in this section.

The first lightning waveshape was represented as a current source, based on a double
exponential function [5]. However, later it was found that this function could not accu-
rately simulate the maximum steepness close to the current peak. Also, it was suffering
from a discontinuity at the onset time (t=0) on its first derivative [6]. Subsequently, a dif-
ferent representation was proposed by Heidler [7] which overcame the limitations of the
double exponential function. Furthermore, the Heidler function can also represent sub-
sequent lightning strokes by summing up two or more Heidler functions [8], but by doing
so the synthesized wave loses its existing concavity at the first front stroke [9]. Later, the
International Council on Large Electric Systems (CIGRE), suggested the synthesizing of
two independent expressions: one with an increasing steepness up to 90% of the peak
and the other one being used to represent the tail of the waveshape [10]. Even though
this current waveshape overcame all above mentioned limitations, Visacro [9] proposed
a different representation called ’Double Peaked’, which is produced by summing up 7
Heidler-functions with different characteristics. The reasoning behind this representa-
tion is that, according to Visacro, multiple peaks offer a better representation for the first
stroke currents. Additionally, if the front and tail time parameters of the CIGRE wave-
shape are not carefully selected, a discontinuity between its two different waveshapes
occurs.

A representation of all the different lightning current waveshapes, including the double
peaked is shown in Figure 2.3. The double peaked function was modelled using MODELS
2. Lightning Phenomena and their Effect on Grounding Systems 8

of EMTP/ATP and can be found in the Appendix F. In the figure, the initiation time of
the Triangular and the Heidler function was delayed, so that the current peak could be
reached almost at the same time for a better illustration.

35
Cigre
Double Peak
30 Heidler
Triangular

25
Current [kA]

20

15

10

0
10 20 30 40 50
Time[μs]

Figure 2.3: Representations of first return stroke currents

In the course of this thesis, the Heidler function will represent the lightning stroke as
it does not suffer from any discontinuities in its derivatives. It is considered accurate
and recommended in the IEC standards. The selection of its rise and decay parame-
ters is simpler compared to CIGRE and Double Peaked waveshapes. Finally, the Heidler
function is included on the XGSLab software tool, unlike the CIGRE and Double peaked
functions and an essential part of this thesis is a voltage response comparison between
XGSLab and EMTP/ATP. This is a very important step, as TenneT is interested to know
the various capabilities of the software tool and it also helps in validating the performed
work.

2.3.1. The Heidler Waveshape


As mentioned earlier, the standard lightning waveshape recommended from IEC for
lightning return-stroke is defined with the following Heidler function which is shown
in equation (2.1) and Figure 2.4.

n −t
Ip (t /t 1 ) t2
i= k · 1+(t /t )n
·e (2.1)
1

Where,
1
t n·t
−( t 1 )( t 2 ) n
• k =e 2 1 , correction factor

• Ip is the peak current

• n is the steepness factor of the peak current


2. Lightning Phenomena and their Effect on Grounding Systems 9

• t is the time

• t 1 front time constant

• t 2 decay time constant

Figure 2.4: Standard lightning waveshape defined with the Heidler function. Adapted from [11]

The Heidler function is based on years of statistical analysis of lightning strokes. The cur-
rent rise may be adjusted by the coefficients of the steepness factor for the peak current
(k) and the front time constant (t 1 ), while the decay can be adjusted by the decay time
constant (t 2 ). The parameters for the standard lightning for the current shapes of the
first positive impulse, the first negative impulse and the subsequent negative impulses
for different Lightning Protection Levels (LPL) from IEC 62305-1 are listed in Table 2.1
[11]. The maximum current values in Table 2.1, represent the threat of the lightning
stroke for the components and equipment that need to be protected. The equation in
this case, do not match the figure as there are two different ways to represent the Heidler
waveshape. In Figure 2.4, time T1 is the front time to rise up to 90%, while T2 is the time
to half value of the peak. Finally, current I as seen in the figure does not necessarily mean
it is the peak current.

Table 2.1: Standard lightning parameters

First positive impulse First negative impulse Subsequent negative impulse


Parameters LPL LPL LPL
I II III-IV I II III-IV I II III-IV
I(kA) 200 150 100 100 75 50 50 37.5 25
k 0.93 0.93 0.93 0.986 0.986 0.986 0.993 0.993 0.993
T1 [µs] 19 19 19 1.82 1.82 1.82 0.454 0.454 0.454
T2 [µs] 485 485 485 285 285 285 143 143 143
2. Lightning Phenomena and their Effect on Grounding Systems 10

The transient voltages or currents that are generated by lightning on power systems or
different human made systems are called lightning surges. Lightning surges on power
systems can be the result of either a direct lightning strike to the system or a lightning
strike in the vicinity of a system, resulting in electromagnetic and/or resistive coupling.
Lightning arresters and surge arresters are frequently used to protect the system from
lightning strikes and lightning surges. The effectiveness of both measures depends on
a well-designed grounding system, but the Dutch grounding system is primarily based
on the behaviour related to power frequency currents and voltages [4]. Therefore, this
study provides useful information on how grounding systems can be optimized to pro-
tect equipment and provide safe environment for currents and voltages of multiple dif-
ferent frequencies occurring from lightning surges.

2.4. Lightning protection scheme-earth termination system


Generally, an LPS according to IEC 62305-1 [11], has the following three listed functions:

• An air termination system, which is used to intercept the lightning strike to the
structure

• A down conductor system, which is used to conduct lightning currents to earth

• An earth termination system, which is responsible for dissipating the lightning


current into the earth

This thesis is concerned about the earth termination and ways to improve it, so when
a lightning current is dissipated into the earth, the ground potential rise is minimized.
Usually, a low value of 10 Ω is recommended for a resistance to ground calculated at low
frequency for global applications. In the Netherlands a typical value of <1 Ω is very often
applicable.
According to IEC 62305-1 [11], there are two basic types of grounding arrangements
which are analysed in the following subsections.

2.4.1. Type A arrangements


Type A arrangements include the horizontal and vertical ground electrodes that are con-
nected to the down conductor system and they should not form closed loops. In Figure
2.5, it can be observed that, the minimum length of the grounding electrode depends on
the classes of the LPS shown in Table 2.1. The minimum length for a horizontal electrode
should be equal to l 1 and for a vertical equal to 0.5l 1
2. Lightning Phenomena and their Effect on Grounding Systems 11

Figure 2.5: Length of vertical/horizontal conductor as function of the soil resistivity, for different classes of
lightning. Adapted from [11]

Finally, type A electrodes should be installed at a depth of at least z=0.5 metres and they
should be allocated as uniformly as possible, so that electrical coupling effects in earth
can be curtailed [11].

2.4.2. Type B arrangements


Type B arrangements include a ring conductor and foundation ground electrodes that
form a closed loop. Those electrodes can also be meshed. The mean radius r e of these
electrodes should not be less than the l1 value given in Figure 2.5. If the value is found
to be less than the l 1 value, then supplementary vertical or/and horizontal electrodes
should be added. Finally, a ring electrode is preferably buried at least at a depth of z=0.5
metres and that it is distanced at least 1 metre away from external walls [11].

2.4.3. Lightning surges on A type arrangements


For phenomena such as lightning, the effective impedance of a grounding conductor
(case A arrangements) is significantly different. To illustrate this, very briefly a circuit ap-
proach to calculate the impedance of a vertical conductor and a counterpoise is shown
in Figure 2.6. As seen in the figure, the impedance of a vertical conductor is not only
resistive, but it has a capacitive and an inductive part which play a critical as well as a
different role for the different frequencies of lightning. The inductive part will increase
the impedance of the grounding system, while the capacitive part will decrease it. To
understand this different behaviour, an extensive study of the behaviour of grounding
systems in broad frequency range was performed and is explained in chapter 4.
2. Lightning Phenomena and their Effect on Grounding Systems 12

Figure 2.6: Equivalent high frequency lumped circuit for a vertical conductor. Adapted from [12]

The equations shown in Figure (2.6), according to [12], can only be used for a preliminary
analysis, as their applicability is restricted to situations where the length of the ground-
ing conductor is less than 1/10 of the earth’s wavelength. This limits the frequency range
of these equations providing accurate results to lower frequencies.
3
Methods to Study the Transient
Behaviour of Grounding Systems

3.1. Introduction
The analysis of the grounding system at steady-state (power frequency) is not sufficient
to simulate the phenomena that occur during transients, such as lightning. In this chap-
ter, a review of the most common techniques for modelling grounding structures during
transient conditions is presented.

When modelling a grounding conductor at power frequency, the most important aspect
is the conductors resistance to ground, while capacitive and inductive phenomena are
usually neglected. Ideally, the value of the grounding resistance needs to be as small as
possible, so that any current flowing through the conductor or a large mesh of conduc-
tors can be easily dispersed into the ground. When a current is led into the soil through
a grounding system, this results in a voltage value called Ground Potential Rise (GPR).
This value is proportional to the grounding resistance, so a lower grounding resistance,
means a lower GPR. A very high GPR can be hazardous for people and animals in the
vicinity in terms of touch and step voltage. The resistance of any grounding structure,
such as a horizontal conductor, during steady-state conditions (power frequency), can
be easily determined either by an analytical formula that exist in standards or by mea-
suring it.

However, during high frequency transients the behaviour of the grounding system changes
significantly. The transients inserted into a grounding grid consist of different frequency
components, all experiencing a different conductor impedance. Also, capacitive and in-
ductive characteristics play a larger role. Hence, different conductors of the grounding
grid will result in different potentials and all conductors will become frequency depen-
dent. This means that every connection point of the grounding grid will be coupled by:
resistive, inductive and capacitive phenomena with each connection point and will be

13
3. Methods to Study the Transient Behaviour of Grounding Systems 14

also influenced by the soil they reside [13].

The behaviour of the grounding grid is considerably influenced from factors such as:

• The soil’s electrical properties

• The geometry of the grounding system

• The injection point and wave-shape of the current

The impedance of a grounding system should be lower than the equivalent impedance of
the remaining system during a lightning surge phenomenon, else, an electric discharge
might flow through the system and cause severe stress or damage to the equipment [14].
Therefore, what makes the study of grounding grids under transient conditions very im-
portant, is the fact that usually higher impedance values are found compared to power
frequency [15].

Analytical methods to calculate the grounding impedance of different electrode arrange-


ments during transient conditions are described in the next sections. These methods
are:

1. The Transmission Line Method (TLM)

2. The Circuit Method

3. The Electromagnetic Method

4. The Hybrid Method (combination of 1, 2 and 3)

By using these methods, the broad frequency behaviour of a grounding system can be
modelled as a single frequency dependent impedance or as a frequency dependent impedance
matrix [16].

3.2. Transmission Line Method


The Transmission Line Method (TLM) is a differential numerical method which can be
implemented in time and frequency domain. For this method, the Maxwell equations
are solved using the theory of the transmission line. To model a grounding structure
using the TLM method, the ground structure is modelled as an n number of lumped π
circuits consisting of (R, L, C and G). For the case, of only one simple horizontal con-
ductor as shown in Figure 3.1, the conductor is divided into n smaller segments of equal
distance (`). The maximum appropriate length of each electrode section is depending
on the maximum frequency of the impulse injected to it.
3. Methods to Study the Transient Behaviour of Grounding Systems 15

Figure 3.1: Transmission line model. Adapted from [17]

Voltage and current distribution along the electrodes must satisfy the telegraphy equa-
tions (3.1), (3.2).
− ∂u(l
∂l
,t )
= R I (l , t ) + L ∂I∂t
(l ,t )
(3.1)

− ∂I ∂l
(l ,t )
= GU (l , t ) +C ∂U∂t(l ,t ) (3.2)
This method treats the current of the stroke as a wave, where the characteristics of the
wave such as, wave propagation constant and surge impedance are calculated in terms
of the per unit length of the series impedance (Z) and shunt admittance (Y) .

Z = R + j ωL Y = G + j ωC (3.3)
where G is the per-unit length conductance (S) of the grounding system, C is the capaci-
tance (F) and L is the inductance (H).
For a conductor with a relatively short length and large cross-section the internal resis-
tance and inductance are significantly smaller compared to the external self-inductance.
Thus, a simplification that only takes the self-inductance into account is made [18].
Therefore, for a horizontal buried grounding wire, the relationship between the soil prop-
erties and the grounding elements are defined as shown in equations (3.4-3.6).

π
Gi = G j = h ³ ´i (3.4)
2l
ρ ln p −1
2ad

C i = C j = G i ρ(²0 ²r ) (3.5)

µ
³ ´
Li = L j = 2π l n 2l
a −1 (3.6)

Where ρ is the soil resistivity and ²r is the relative permittivity of the soil, d is the burial
length and a the radius.
3. Methods to Study the Transient Behaviour of Grounding Systems 16

When a lightning impulse is conducted through a grounding wire, the current will create
an outward time varying electric field throughout the wire. The amount of time and
current that will be dissipated into the soil will depend on the electrical characteristics
of the soil. In equation (3.7), the relationship between the electric field and the current
density which was dissipated into the soil is provided. The linear behaviour between
current density and electric field is only valid below the soil critical breakdown value,
denoted E c [18].

E = j soi l · ρ (3.7)
Finally, an algorithm based on the transmission line method, could be described with
the following steps [19]:
a. Determine the incident voltages in all segments in regard to the injected impulse;

b. Compute the voltages, currents and fields associated to the segments of interest;

c. Compute reflected voltages for all segments;

d. Use of boundary conditions for the segments located in the extremities of the cal-
culation domain;

e. Determine the new incident voltages for the following iteration;

3.3. Circuit Method


Another method for modelling grounding grids is the Circuit method. The main steps of
this method are [20]:
• Divide the grounding system into n segments

• For each segment create an equivalent lump circuit and determine its parameters,
such as mutual and self inductance, capacitance and conductance

• Solve its nodal equations based on Kirchoff’s laws


Until now, many researchers have used different circuits for the transient analysis of
grounding grids. The first one was developed in 1983 by A. P. Meliopoulos [21], which
was only based on resistive elements and is shown in Figure 3.2. One other model based
on the circuit approach is presented in [22]. Here, the authors modelled the circuit which
represents a segment of a grounding grid buried in depth h as a cascade of π circuits as
shown in Figure 3.3.

Figure 3.2: Circuit representation of a short length conductor. Adapted from [21]
3. Methods to Study the Transient Behaviour of Grounding Systems 17

Figure 3.3: Circuit representation for each segment. Adapted from [22]

In Figure 3.3, the resistance and self inductance are connected in series and are used
to determine the current, while the conductance and the capacitance are connected in
parallel, representing the losses to the ground. The most influential parameters of the
circuit are the inductance and the conductance, while the resistance and capacitance
are usually neglected since they are very small. The transverse conductance of the con-
ductor can be determined using equation (3.8).

π·l
G =ρ ³ ´ (3.8)
2·l
ρ s ·l n r

Where, ρ is the soil resistivity, r is the radius of the conductor and l its length.

The self-inductance can be determined using equation (3.9) and is used as a means to
represent the induced magnetic field along the conductor, which is caused from the cur-
rent flowing through the grounding grid.
³ ´
L = 2 · 10−7 · l · ln 2·l
r −1 (3.9)

Finally, the mutual inductance as seen in equations (3.10), is used to represent the mag-
netic coupling of two conductors length l1 and l2.
Z Z
µ ∂l i ·∂l j
L i j = 4·π · ∂ (3.10)
ij
li lj

Where d i j is the distance between conductors i and j and µ the magnetic permeability
of the medium.
The advantage of this method is that can be directly modelled in time domain (TD), thus
it is easier to intergrate with a time domain software tool. Its main drawback is, that it is
difficult to model it in a way that anticipates the surge propagation delay [23].

3.4. Electromagnetic Method


The Electromagnetic (EM) method, is considered to be the most rigorous when mod-
elling grounding grids for transient studies. It is solely based on Maxwell equations and
has the least neglects possible [24], [25]. There are two techniques used in literature
for modelling using the electromagnetic approach, namely the Finite Element Method
3. Methods to Study the Transient Behaviour of Grounding Systems 18

(FEM) and the Method of Moments (MoM). One of the first models that was created us-
ing the EM approach based on MoM was first discussed by Grcev in 1986. In this method,
all grounding conductors were divided in n segments with length (`). The problem was
initially solved in frequency domain (FD) and by using Fourier transformation the prob-
lem can be also represented in time domain (TD). This method begins by calculating the
electric field that is exerted outwards of the conductor. Because of the current (I) and
the charge (q) along each segment of the electrode, the electric field can be expressed in
terms of the magnetic vector potential A x and the electric scalar potential φ as shown in
equation (3.11) [26].


E (x) = − ∂x φ(x) − j ωA x (x) (3.11)
Where the magnetic vector potential and electrical scalar potential are also depended on
Green’s function and by making use of Sommerfeld integrals the effect of lossy ground is
taken into account and a generalized impedance matrix can be constructed [27]:
    
U1 Z11 Z12 Z13 ··· Z1m I1
U   Z Z22 Z23 ··· Z2m  I 
 2   21  2
U   Z Z32 Z33 Z3m 
···   I3 
 
 3  =  31 (3.12)
 .   . .. .. ..  
.. .
 
 .   .
 .   . . . . .   .. 
Un Zn1 Zn2 Zn3 ··· Znn In

In equation (3.12), the equivalent impedances between points m and n that are calcu-
lated as, the resulting voltage in node m from a current injection in node n, supposing
no other feeding currents exist in the grounding system. The diagonal impedances are
the result of a current injection on that node which results on a voltage, assuming there
are no other feeding currents. The dimensions of the matrix are equal to the number of
feeding points that are connected to the system above ground. If those impedances are
calculated then the grounding system can be fully interfaced to the system above ground
[1].

3.5. Hybrid Method


This thesis is implemented using the XGSLab software tool, which is based on the Hybrid
method. Therefore, a more thoroughly explanation of this method is provided compared
to the other methods.

The Hybrid method for transient analysis of grounding systems was firstly presented by
Dawalibi in 1986 [28]. The name "Hybrid" results from the fact that this approach is a
combination of the previously discussed methods. This method is very useful because it
is considered rigorous and adaptable as it allows assimilating additional external param-
eters, such as, electromotive forces, currents, and impedances. The hybrid method, can
be used to solve any mesh of conductors in an arbitrary way in 3D space. This method
can be described with the following steps: The grounding system is divided into many
smaller segments and by using the electromagnetic and transmission line theory the cir-
cuit parameters are solved. Finally, the circuit theory is used to distinguish the affiliation
3. Methods to Study the Transient Behaviour of Grounding Systems 19

between parameters such as voltages and currents [29].

The implemented method is derived directly from Maxwell equations. However, Maxwell
equations can be rewritten into the equations of Helmholtz by making use of the vector
potential (A) and scalar potential (V).

∆ Ȧ − γ̇2 Ȧ = −µ J˙
(
q̇ (3.13)
∆V̇ − γ̇2 V̇ = − ²̇
where:

− γ̇ = j ωµ(σ + j ω²): corresponds to the propagation coefficient of the medium


p

− q̇: corresponds to the charge of the sources

− J˙: corresponds to the current density distribution of the sources

By solving equations (3.13) for sources with a linear current charge density distribution,
the vector potential and the scalar potential are shown respectively in equation (3.14).

µ R ˙ e −γ̇r
(
Ȧ = 4π L I r d l
1
R e −γ̇r (3.14)
V̇ = 4π²̇ L q̇ r d l

Maxwell equations provide the following well known relation between electric field and
scalar and vector potentials at any point as:

Ė = −∇V̇ − j ω Ȧ (3.15)
The electric field and the vector potential on the conductor surface are parallel to its axis,
therefore only the magnitude of the vectors needs to be taken into account. So equation
(3.15) becomes:

∂V̇
Ė = ∂l − j ω Ȧ (3.16)
Furthermore, the tangential electric field on the surface of a conductor, by considering
its self impedance, is equal to:

Ė = ż I˙ (3.17)
By combining equation (3.16) with equation (3.17), the following first order differential
equation is derived:

ż I˙ + j ω Ȧ + ∂dV̇l = 0 (3.18)
Equation (3.18), was derived from Maxwell equations and remains always valid. How-
ever, for practical cases, this equation can only be translated in an arithmetical way. Each
segment of the conductor/system consists of a starting point (a) and an ending point (b).
By integrating each segment and by replacing equation (3.14) into equation (3.18), the
following linear equation is derived.
3. Methods to Study the Transient Behaviour of Grounding Systems 20

Żi I˙i + M˙i j I˙j + ˙ j − Wa ˙− i j ) J˙j = 0


X X
(Wb−i (3.19)
j 6=i

Where the coefficients M i j , Wb−i j , Wa−i j are calculated as:


Z bZ b
j ωµ e −γ̇r
M˙i j = 4π r d li d l j (3.20)
a a
Z b
˙ j= ρ̇ e −γ̇r ¯
¯
Wb−i 4πl r b (3.21)
a
Z b
˙ j= ρ̇ e −γ̇r ¯
¯
Wa−i 4πl r a (3.22)
a

In equation (3.19), Ż is the self impedance of each segment, Ṁ and Ẇ are the partial mu-
tual coupling and the partial potential coefficients between each segment respectively.
As for I˙ and J˙, these quantities represent the longitudinal and leakage current respec-
tively. By solving the equation (3.19) for each segment, Maxwell equations are reduced
to a linear system, where if the propagation delay is taken into account while solving for
the coefficients of the system, the resulting model will be considered a full wave hybrid
[29].
Every segment is represented as a simplified “T” equivalent circuit as seen in Figure 3.4.

Figure 3.4: Equivalent circuit of each segment. Adapted from[29]

The unknown parameters as seen on the circuit are:

• Iin and Iout: Input and Output currents

• J: Leakage current

• V: Potential of the middle point


3. Methods to Study the Transient Behaviour of Grounding Systems 21

Therefore, the resulting linear system (3.23) is a system of arrays that includes all seg-
ments. 
 V =W ·J

E z = −(Z + M ) · I − E e (3.23)


J = A · I + Je
The solution of system (3.23) will provide, current distributions, potentials and leakage
currents along the structure by also taking into consideration the effect of other possible
sources. From the outcome, it is feasible to determine other meaningful distributions,
like: electric and magnetic fields, ground surface potentials (touch & step Voltages).

The Hybrid model is applicable in frequency domain (FD) and also in time domain (TD)
by making use of the direct and inverse Fourier transformation (eq. (3.24)).

S˙n e j 2πn f t
X
s(t ) = (3.24)
n=−∞

As seen from equation (3.24), s(t) is a superposition of different frequencies. Further-


more, it should also be noted that, the upper limit of the summation is limited to a value
N which depends on the frequencies of the inserted transient. However, there is also the
possibility that only the peak values of the time domain are of importance and a good
calculation of these values can be approximated easier, using the frequency domain ap-
proach and an equivalent single frequency input.

3.5.1. Applied software: XGSLab Ver 9.3.1.1


XGSLab is a powerful software tool developed by SINT Ingegneria which includes a vari-
ety of modules for electromagnetic simulation, grounding and lightning protection sys-
tems, which considers the Hybrid method. The different models that are included in
XGSLab are:

• GSA - Grounding System Analysis

• GSA_FD - Grounding System Analysis in Frequency Domain

• XGSA_FD - Over and Underground System Analysis in Frequency Domain

• XGSA_TD - Over and Underground System Analysis in Time Domain

• NETS - Network Solver

The models used for this thesis to be realised were the XGSA_FD and the XGSA_TD.
Those two models are similar, but the later one extends to time domain simulations as
well. XGSA_FD can only solve using a single frequency at a time, in the bandwidth of 0 to
100 MHz. Therefore, XGSA_TD was used, as it can be used to solve for multiple frequen-
cies at a time while considering the same solver as XGSA_FD. The main assumptions for
the XGSA_FD module are:
3. Methods to Study the Transient Behaviour of Grounding Systems 22

Table 3.1: Aspects taken into account in the used software

Characteristics
Resistive Coupling Yes
Capacitive Coupling Yes
Self Impedance Yes
Inductive Coupling Yes
Soil Parameters ρ, ² = f (ω)
Propagation Law e −γr /r

To quickly present one of the software capabilities, a typical grounding grid of 60x60 with
5 meshes and 4 vertical conductors on the edges was considered. The soil resistivity
was set equal to 100 Ω and the permittivity equal to 10. The two simulated cases were
for a lightning discharge at the middle and at the corner of the grid. In Figure 3.5, the
impedance of the grounding grid for both situations is shown. The result observed is
that the impedance for a lightning discharge at a corner of the grid is higher compared
to the middle. Finally, the GPR for both cases is shown in Figure 3.6, for a lightning
discharge of 10kA (1.2/50 µs).

35 25

30
20

25
Impedance [Ω]

Impedance [Ω]

15
20

15
10

10

5
5

0 0
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Impedance of the grounding grid as function of b) Impedance of the grounding grid as function of
frequency, corner energization frequency, middle energization

Figure 3.5: Impedance of the grounding grid as function of the frequency. a) For a lightning discharge at the
corner of the grid, b) For a lightning discharge at the middle of the grid
3. Methods to Study the Transient Behaviour of Grounding Systems 23

4500 3500

4000
3000

3500
2500
3000

2000
GPR [V]

GPR [V]
2500

2000 1500

1500
1000
1000

500
500

0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time [μs] Time [μs]

a) GPR for an energization at the corner b) GPR for an energization at the middle

Figure 3.6: GPR of the grounding grid. a) For a lightning discharge at the corner of the grid, b) For a lightning
discharge at the middle of the grid

As expected, the GPR at the corner is higher, since the result is proportional to the impedance.
The different calculated impedance value is related to the different injection point, which
was already mentioned that has a significant role. To elaborate more on it, even for a sim-
ple conductor, its impedance value can be changed considerably according to, where the
earth termination system is connected. Figure 3.7, illustrates this effect for two different
situations, a grounding conductor and a grounding grid.

Figure 3.7: Influence of the earth termination connection to the structure impedance to ground: a) Horizontal
conductor, b) Grounding grid-energization at the center. Adapted from [3]

For the horizontal conductor seen in Figure 3.7 a). If the current injection (energiza-
tion) occurs at point A the result obtained will be twice of how much it would be for
an energization at point B. Because, if the energization occurs at point B the current sees
two parallel impedances with very similar values, and therefore, the amplitude is halved.
Similarly, for a grounding grid shown in Figure 3.7 b), if the energization takes place at
the center (point B) this will result to an impedance which is about 1/3- 1/4 of the value,
should the energization where to happen to a corner of the grid.

In the next chapters, using the Hybrid method of XGSLab, complex grounding structures
are modelled to study their broad frequency behaviour and to include them in the EMT-
P/ATP software tool for transient simulations.
4
Broad Frequency Behaviour of
Grounding Systems

4.1. Introduction
In this chapter, the different effects that influence the impedance of grounding struc-
tures are discussed. It is already widely known that the impedance of a grounding struc-
ture varies for different electrical parameters of the soil and for different frequencies.
However, there are other factors that may influence the result which are usually disre-
garded in most studies, such as: the frequency dependency of soil and soil ionization.

4.2. Frequency dependency of soil and soil ionization


The frequency dependency of the electrical parameters of the soil has been a subject for
research for over a century. Often, for transient studies, the soil would be modelled as a
medium which has constant parameters (ρ, ²r , µ) [30]. These parameters are:

• ρ: is the soil resistivity

• µ: is the soil permeability

• ²r : is the soil’s relative permittivity

Soil resistivity could be easily estimated with four-pin measurements, (e.g. the Schlum-
berger and Wenner method) at low voltage gradients, low current densities, and rela-
tively low frequencies (around power frequency). [31]. Therefore, for most studies, the
soil resistivity is usually assumed to be equal to the measured value or equal to 100-1000
Ωm, if measurement data is not available. The permeability µ of the soil is assumed to
be equal to the permeability of air. Finally, the value of the relative permittivity (²r ) is
usually chosen from between 4 and 20 depending on the moisture content of the soil
[32]. In Table 4.1, typical values of permittivity for various types of soil can be observed.

24
4. Broad Frequency Behaviour of Grounding Systems 25

Table 4.1: Typical values of soil permittivity

Material ²r
Dry sand 4
Ice 4
Coal 4-5
Concrete 6
Saturated sand 20
Water 80

However, for transient studies, the previously mentioned secondary effects become con-
siderably more important. A case in point is where grounding electrodes conduct a light-
ning current to ground. Lightning is a transient phenomenon with a typical frequency
range of 80 kHz to 1 MHz, therefore the result might be different in comparison to the
predicted result with constant soil parameters. Experimental results from many scien-
tists have demonstrated that, both soil resistivity and permittivity are decrease signifi-
cantly while the frequency increases. Moreover, the values of the relative permittivity of
soil have proven to be much larger over this wide range of frequency in comparison to
the ones assumed in typical studies [32]. Many renowned scientists have come up with
different models that contribute to understanding the effect of frequency dependency of
soil. An overview of those models is given in [30], [32] and they are also included in the
Appendix A.

Finally, there is also the phenomenon of soil ionization, which occurs when the am-
plitude of an injected current exceeds the soil’s threshold and, as a result, the effective
diameter of the conductor that conducts the current increases [33].

In the following sections, the behaviour of grounding systems is extensively studied, con-
sidering all the previously mentioned effects except of soil ionization, as it is not imple-
mented directly on XGSLab.

4.3. Behaviour of grounding systems when conducting light-


ning currents
It is very clear that the electrical parameters of the soil are influenced by the frequency.
Furthermore, it is widely known that the impedance value of a grounding structure is
dependent on the electrical parameters of the soil and to the applied frequency. There-
fore, as a first step, grounding systems with constant soil values are considered. It is also
worth mentioning that, the software XGSLab includes many of the frequency dependent
models discussed in Appendix A. Hence, as a second step, grounding structures at fre-
quency dependent soils are considered. Finally, soil ionization will not be considered in
any of the simulations, as it is not directly implemented in the XGSLab software tool and
because of time limitations.
4. Broad Frequency Behaviour of Grounding Systems 26

4.3.1. Grounding systems with constant soil parameters


In this section, four different grounding structures are modelled using XGSLab, in order
to observe the effect of the soil parameters on the grounding structures.The grounding
structures include: an vertical conductor, a horizontal conductor, a ring conductor and
a grounding grid. The grounding structures along with the spot where the energization
takes place can be viewed in Figure 4.1.

Figure 4.1: Grounding structures modelled: a) Vertical Conductor, b) Horizontal Conductor c) Ring
Conductor, d) Grounding Grid

The energization for all cases was chosen at 1A. In this way, the effect on both the impedance
and the GPR of the structure can be shown, in the same plot, since the impedance will
be equal to the GPR. The frequency and the electrical parameters of the soil have a great
impact on the impedance of the grounding structures. In the next subsections, numer-
ous simulations are performed to understand their impact on the impedance and on the
GPR of the grounding structures.

4.3.1.1. Vertical Conductor


The vertical conductor that was used in the simulations is made of copper and has a
length of 9 metres. In Figure 4.2, its impedance is plotted as a function of the soil resis-
tivity for three different frequencies This plot was made for two different permittivities,
1 and 8 and are seen respectively in 4.2 a) and b). The simulations actually shows that
similar differences would occur for any permittivity that is equal to 2 or higher. However,
4. Broad Frequency Behaviour of Grounding Systems 27

8 was selected to show that high values of permittivity do not play a critical role in low
frequencies.

700 700
f=100 Hz
f=100 Hz
f=1 MHz
f=1 MHz
600 f=10 MHz 600
f=10 MHz

500 500
Impedance [Ω]

Impedance Ω]
400 400

300 300

200 200

100 100

0 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Soil Resistivity [Ωm] Soil Resistivity [Ωm]

a) Plot for permittivity ²r = 1 b) Plot for permittivity ²r = 8

Figure 4.2: The effect of soil resistivity on the impedance of the vertical conductor

It can be clearly observed from Figure 4.2 a), that the effect of the soil resistivity on the
vertical rod impedance is linear in the frequencies up to 1MHz. However, for frequencies
as high as 10 MHz, the impedance of the vertical conductor no longer increases linearly
and it starts dropping slowly for soil resistivities higher than 2000 Ωm. In Figure 4.2 b),
the impedance does not increase linearly already from 1 MHz. It is also worth mention-
ing that, the impedance does not start to drop for a soil resistivity higher than 2000 Ωm
as in Figure 4.2 a), but actually starts to increase. This means that the permittivity also
plays a significant role in the impedance value of the structure, which is shown in more
detail in the next section of this document.

To further investigate the influence of permittivity on the vertical conductor’s impedance


four simulations were performed with different soil resistivities that vary with frequency.
The four soil resistivities that were chosen are 100 Ωm, 500 Ωm, 5000 Ωm and 10000 Ωm
and can be seen respectively in Figures 4.3 a), b), c) and d). This makes it easier, to realize
the effect of permittivity in high conductive, medium conductive and poor conductive
soils. The selected frequency range was from 10Hz to 10MHz. However, the plots are
scaled, since higher differences occur for higher frequencies.
4. Broad Frequency Behaviour of Grounding Systems 28

70 150
εr=5 εr=5

εr=10 εr=10
60 εr=25 εr=25

εr=50 εr=50

εr=80 εr=80
50 100
Impedance [Ω]

Impedance [Ω]
40

30 50

20

10 0
103 104 105 106 107 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Plot for soil resistivity equal to 100 Ωm b) Plot for soil resistivity equal to 500 Ωm

700 1400
εr=5
εr=10
600 1200
εr=5
ε=25
εr=10 εr=50
500 εr=25 1000 εr=80
Impedance [Ω]

Impedance [Ω]
εr=50
εr=80
400 800

300 600

200 400

100 200

0 0
103 104 105 106 107 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

c) Plot for soil resistivity equal to 5000 Ωm d) Plot for soil resistivity equal to 10000 Ωm

Figure 4.3: The effect of permittivity on the impedance of the vertical conductor as function of the frequency

In Figure 4.3 a) and b), where the conductivity of the soil is higher, it can be observed
that, for low values of permittivity the impedance of the vertical conductor is increasing
steadily, while for permittivities equal to 50 and above there are many oscillations and
the impedance of the vertical conductor is considerably smaller. On the other hand, for
soils of poor conductivity such as in Figure 4.3 c) and d) oscillations for all permitivi-
ties occur at high frequencies and the impedance varies considerably (decreasing and
increasing) for all permittivity levels. Another interesting observation in Figure 4.3 c)
and d), is the fact that in the high frequency region, the impedance of a higher resistive
soils can become substantially smaller compared to a lower resistive soil. To explain this
result, the equivalent circuit of a vertical conductor that was mentioned in section 2.4
was considered. The impedance of the equivalent circuit can be calculated by taking
the parallel combination of the capacitor with the resistor and then adding the inductor.
Therefore, to verify the result of Figure 4.3, the circuit shown in Figure 2.6, was simu-
lated using MATLAB. The results can be seen in Figure 4.4, where the impedance and
the phase angle of the vertical conductor were plotted as function of the frequency for
three different soils with a relative permittivity equal to 5. Finally, it is once more worth
4. Broad Frequency Behaviour of Grounding Systems 29

pointing out, that this circuit can only be used for a preliminary analysis as its result for
high frequencies can considerably underestimate the impedance value.

90 90

80 80

70
70

Phase Angle [Deg]


60
Impedance [Ω]

60
50
50
40
40
30

30
20

20 10

10 0
102 103 104 105 106 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

a) Impedance - soil resistivity= 100 Ωm b) Phase angle - soil resistivity= 100 Ωm

74 60

72
50
70
Phase Angle [Deg]

68 40
Impedance [Ω]

66
30
64

62 20

60
10
58

56 0
102 103 104 105 106 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

c) Impedance - soil resistivity= 500 Ωm d) Phase angle - soil resistivity= 500 Ωm

600 60

40
500

20
Phase Angle [Deg]

400
Impedance [Ω]

0
300
-20

200
-40

100
-60

0 -80
102 103 104 105 106 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

e) Impedance - soil resistivity= 5000 Ωm f) Phase angle - soil resistivity= 5000 Ωm

Figure 4.4: Impedance and phase angle as function of frequency for different soil resistivities
4. Broad Frequency Behaviour of Grounding Systems 30

By looking at the phase angle in Figure 4.4, it is evident that for a low soil resistivity, such
as in Figure 4.4 b), the phase angle is resistive for lower values of frequency and then be-
comes inductive. That is the reason the increase in the impedance is observed. However,
by observing the phase angle in Figure 4.4 f) once again, it can be seen that the behaviour
at low frequencies remains resistive, while in high frequency it becomes capacitive. For
this reason the impedance drops with the frequency, as was previously observed in Fig-
ure 4.3 c) and d). Finally, due to the fact that permittivity plays a significant role in the
high frequency region, a simulation at 10 MHz as function of the permittivity was also
conducted. The simulation was repeated for three different types of soil conductivity:
good, medium and poor. In Figure 4.5, the results of the simulation can be observed.

Figure 4.5: Impact of the permittivity on the vertical conductor impedance at high frequency (10 MHz)

From Figure 4.5, it can be easily deduced that soils with good conductivity are not influ-
enced as much as soils with poor conductivity, for different permittivities. Furthermore,
it can be observed that, a soil of medium conductivity decreases rapidly for a permittiv-
ity value of 30 or lower. Finally, for a poor conductive soil it can be observed that, the
impedance value is greatly influenced by the permittivity and that in some cases it can
result in an even lower impedance value than in soils of better conductivity.

4.3.1.2. Horizontal Conductor


The horizontal conductor is made of copper, it is buried 1 metre below the ground and
has a length of 10 metres. In Figure 4.6, the impedance of the horizontal conductor is
plotted as a function of its soil resistivity for three different frequencies. This plot was
made for two different permittivities, 1 and 8 and are seen respectively in 4.6a and 4.6b.
4. Broad Frequency Behaviour of Grounding Systems 31

700 700
f=100 Hz f=100 Hz
f=1 MHz f=1 MHz
600 f=10 MHz 600 f=10 MHz

500 500

Impedance [Ω]
Impedance Ω]

400 400

300 300

200 200

100 100

0 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Soil Resistivity [Ωm] Soil Resistivity [Ωm]

a) Plot for permittivity ²r = 1 b) Plot for permittivity ²r = 8

Figure 4.6: The effect of soil resistivity on the impedance of the horizontal conductor

It can be observed from Figure 4.6, that the effect of the soil resistivity on a horizontal
conductor has exactly the same result as a vertical conductor shown in Figure 4.2. The
same simulations to investigate the effect of permittivity for a grounding grid were re-
peated. The subsequent plots for three different soil resistivities can be seen in Figure
4.7 a), b), c) and d).

70 140
εr=5 εr=5
εr=10 εr=10
60 εr=25 120 εr=25
εr=50 εr=50
εr=80 εr=80
50 100
Impedance [Ω]

Impedance [Ω]

40 80

30 60

20 40

10 20
103 104 105 106 107 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Plot for soil resistivity equal to 100 Ωm b) Plot for soil resistivity equal to 500 Ωm
4. Broad Frequency Behaviour of Grounding Systems 32

1400
εr=5
εr=10
1200
εr=25
εr=50
1000
εr=80

Impedance [Ω]
800

600

400

200

0
103 104 105 106 107
Frequency [Hz]

c) Plot for soil resistivity equal to 5000 Ωm d) Plot for soil resistivity equal to 10000 Ωm

Figure 4.7: The effect of permittivity on the impedance of the horizontal conductor as function of the
frequency

The results in Figure 4.8, follow almost the same behaviour as with the vertical conductor
and there are no noteworthy differences. Once again, due to the fact that permittivity
plays a greater role in the high frequency region a simulation at 10 MHz as function of
the permittivity was also conducted.The simulation was performed once more for soils
of good conductivity, medium and poor. In Figure 4.8, the results of the simulation can
be observed. The results are also expected to follow the exactly the same behaviour as
with the vertical conductor since no differences were observed in the previous figures.

400
ρE= 100 Ωm
350 ρE =500 Ωm
ρE =5000 Ωm
300
Impedance [Ω]

250

200

150

100

50

0
0 20 40 60 80 100
Relative Permittivity

Figure 4.8: Impact of the permittivity on the impedance of the horizontal conductor at high frequency (10
MHz)

4.3.1.3. Ring Conductor


The ring conductor is also made of copper, it is buried 1 metre below the ground and has
a radius of 10 metres. In Figure 4.2, the impedance of the ring conductor is plotted as a
4. Broad Frequency Behaviour of Grounding Systems 33

function of its soil resistivity for three different frequencies. This plot was made for two
different permittivities, 1 and 8 and are seen respectively in 4.9a and 4.9b.

240
240 f= 100 Hz
f=100 Hz 220
220 f=1 MHz f=1 MHz
f=10 MHz 200 f=10 MHz
200
180
180
160
Impedance [Ω]

Impedance [Ω]
160
140
140
120
120

100 100

80 80

60 60

40 40

20 20

0 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Soil Resistivity [Ωm] Soil Resistivity [Ωm]

a) Plot for permittivity ²r = 1 b) Plot for permittivity ²r = 8

Figure 4.9: The effect of soil resistivity on the impedance of the ring conductor

In Figure 4.9, a similar behaviour as that of the vertical conductor can be observed, with
the only difference being that, in Figure 4.9 a), a small non-linearity appears in low soil
resistivities at the frequency of 1MHz compared to the vertical conductor where it was
also linear. The same simulations to investigate the effect of permittivity for a ring con-
ductor were repeated. The resulting plots for three different soil resistivities can be seen
in Figure 4.10 a), b), c) and d).

60 90
εr=5
εr=10
80 εr=5
50 εr=25
ε=10
εr=50 ε=25
70
εr=80 εr=50
40
Impedance [Ω]

Impedance [Ω]

εr=80
60

30 50

40
20

30

10
20

0 10
103 104 105 106 107 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Plot for soil resistivity equal to 100 Ωm b) Plot for soil resistivity equal to 500 Ωm
4. Broad Frequency Behaviour of Grounding Systems 34

200

180

160
εr=5
140 εr=10

Impedance [Ω]
εr=25
120
εr=50
100 εr=80

80

60

40

20

0
103 104 105 106 107
Frequency [Hz]

c) Plot for soil resistivity equal to 5000 Ωm d) Plot for soil resistivity equal to 10000 Ωm

Figure 4.10: The effect of permittivity on the impedance of the ring conductor as function of the frequency

The results in Figure 4.10 a), follow the same behaviour as with the vertical conductor.
However, for a moderate soil conductivity, as shown in Figure 4.10 b), the oscillations are
damped faster compared to the vertical conductor. Moreover, at high frequencies for a
permittivity equal to 50 and 80 the impedance increases significantly. A possible reason
for this is, that the software tool might provide inaccurate results for such high frequen-
cies. Also, for the poor conductive soils, it can be observed that, for a low permittivity
value, the impedance does not decrease, as much as in the case of the vertical conduc-
tor. Moreover, for the case in which permittivity equals 5, the impedance oscillates and,
at one point, it rises to a higher amplitude compared to its starting amplitude. Finally,
due to the fact that permittivity plays a greater role in the high frequency region, a sim-
ulation at 10 MHz as function of the permittivity was also performed. The simulation,
was performed again, for soils of good conductivity, medium and poor. In Figure 4.11,
the results of the simulation can be observed.

Figure 4.11: Impact of the permittivity on the impedance of a ring conductor at high frequency (10 MHz)
4. Broad Frequency Behaviour of Grounding Systems 35

In Figure 4.11, the same behaviour as in the case of the vertical conductor can be ob-
served. However, for the poor soil conductivity there is no oscillation in its amplitude,
which is expected since, as mentioned earlier, the oscillations are damping.

4.3.1.4. Small grounding grid


The grounding grid has a size of 15X15 metres, all its conductors are copper and it is lo-
cated 1 metre below the ground. The impedance of the grounding grid plotted as func-
tion of its soil resistivity for three different frequencies can be seen in Figure 4.12. This
plot was made for two different values of permittivity, 1 and 8, which are seen respec-
tively in Figure 4.12 a) and b).

200 200
f=100 Hz f=100 Hz
180 180 f=1 MHz
f=1 MHz
f=10 MHz f=10 MHz
160 160

140 140
Impedance [Ω]

Impedance [Ω]
120 120

100 100

80 80

60 60

40 40

20 20

0 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Soil Resistivity [Ωm] Soil Resistivity [Ωm]

a) Plot for permittivity ²r = 1 b) Plot for permittivity ²r = 8

Figure 4.12: The effect of soil resistivity on the impedance of the grounding grid

It can be clearly seen that the behaviour of the grounding grid for different soil resis-
tivities as seen in Figure 4.12, has the same behaviour as the ring conductor shown in
Figure 4.9. The same simulations to investigate the effect of permittivity for a grounding
grid were repeated. The resulting plots for three different soil resistivities can be seen in
Figure 4.13 a), b), c) and d).
4. Broad Frequency Behaviour of Grounding Systems 36

110

100 εr=5
εr=10
90
εr=25

80 εr=50

Impedance [Ω]
εr=80
70

60

50

40

30

20

10
103 104 105 106 107
Frequency [Hz]

a) Plot for soil resistivity equal to 100 Ωm b) Plot for soil resistivity equal to 500 Ωm

300

250
ε=5
ε=10
ε=25
200
Impedance [Ω]

ε=50
ε=80

150

100

50

0
103 104 105 106 107
Frequency [Hz]

c) Plot for soil resistivity equal to 5000 Ωm d) Plot for soil resistivity equal to 10000 Ωm

Figure 4.13: The effect of permittivity on the impedance of the grounding grid as function of the frequency

By looking at Figure 4.13, it can be observed that, the grounding grid follows more or less
a similar behaviour with that of the ring conductor, seen in Figure 4.10. Once again, due
to the fact that permittivity plays a greater role in the high frequency region, a simula-
tion at 10 MHz as function of the permittivity was also conducted. The simulation was
performed once more for soils of good, medium and poor conductivity. In Figure 4.14,
the results of the simulation can be observed, where the result is expected to be similar
to that of the ring conductor, shown in Figure 4.11.
4. Broad Frequency Behaviour of Grounding Systems 37

200
ρE= 100 Ωm
180
ρE =500 Ωm

160 ρE =5000 Ωm

140

Impedance [Ω]
120

100

80

60

40

20

0
0 20 40 60 80 100
Permittivity

Figure 4.14: Impact of the permittivity on the impedance of a grounding grid at high frequency (10 MHz)

4.3.2. Grounding systems with frequency dependent soil


As stated earlier in this chapter, when the frequency dependency of the soil is included,
the soil resistivity and permittivity will drop by increasing the frequency. To show the
effect of frequency dependency of soil, the model of Visacro-Alipio was considered. This
model was chosen as every model predicts similar results, with the only exception be-
ing the model of Visacro-Portela. When considering the frequency dependency of soil,
a permittivity value is set arbitrarily, considering the value of the resistivity of the soil.
Therefore, to explain the behaviour of grounding systems for a frequency dependent
model, the same vertical conductor as in section 4.3.1 is considered and the result was
compared to the result of a soil with constant parameters ρ = r ho 0 and ²r = 10. This
comparison is done for a soil of 100 Ωm, 500 Ωm, 5000 Ωm, 10000 Ωm and the result are
shown in Figure 4.15 a), b), c) and d) respectively.

70 130
=10 r
=10
r
120
r r
60
110

100
50
90

40 80

70
30
60

50
20
40

10 30
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Plot for soil resistivity equal to 100 Ωm b) Plot for soil resistivity equal to 500 Ωm
4. Broad Frequency Behaviour of Grounding Systems 38

700 1400
r
=10 r
=10

600 r 1200 r

500 1000

400 800

300 600

200 400

100 200

0 0
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

c) Plot for soil resistivity equal to 5000 Ωm d) Plot for soil resistivity equal to 10000 Ωm

Figure 4.15: Comparison of frequency dependent soil with a soil with constant soil parameters of ρ = ρ 0 and
²r = 10

From Figure 4.15 a), b), c) and d), it can be noticed that the behaviour of the vertical
conductor has not changed. It remains, mostly inductive for high conductive soils and
capacitive for poor conductive soils. Furthermore, it can also be observed that, for a soil
of 100 Ωm, the impedance remains the same, almost up to 100kHz and then decreases
very slightly. Finally, it is clear that, the poorer the conductivity of the soil, the faster the
impedance curve will begin to drop with the frequency.

4.4. Discussion
In this chapter, an extensive analysis of the behaviour of grounding conductors was con-
ducted. For this analysis, the behaviour of the grounding conductors was investigated
for constant soil parameters and for soil parameters that vary with the frequency.

For the case of the Dutch transmission system, very low soil resistivities are usually ob-
served, since 1/3 of the country is located below sea water level and there is a lot of rain-
fall, especially during the winter period. Therefore, the grounding structures in the next
chapters are modelled for constant soil parameters (ρ = 100 and ²r = 10). This decision is
based on the fact that, the models that predict the frequency dependency of soil are less
accurate for soils of low soil resistivity [32].Moreover, for low soil resistivies it was found
that the impedance decreases slightly. Therefore, quite small differences are expected,
if this is taken into account and it also enables us to have more conservative models.
However, the frequency dependency of soil, will be taken into account just for one case,
as it is interesting to show that the generated overvoltages of the grounding system will
be slightly lower. Finally, soil ionization is disregarded for all cases in accordance with
literature for low resistive soils [32].
5
Modelling Approach - Frequency
Dependent Models

5.1. Introduction
In this chapter, the assumptions and the methods that were used to develop frequency
dependent models, and which are used for transient simulations later on, are analyzed.
Furthermore, the results of the created models are validated by comparing them to the
results obtained with XGSLab. In Figure 5.1, a representation of the steps that were fol-
lowed is provided.

Data Earthing Grid

XGSLab

Vector Fitting ATP/EMTP ARMAFIT

Lightning Waveshape

ATP/EMTP

Transient Model
Results

Figure 5.1: Flowchart of the steps taken

39
5. Modelling Approach - Frequency Dependent Models 40

As shown in Figure 5.1, different grounding structures are initially modelled using XGSLab,
such as: a vertical conductor, a ring conductor, high voltage towers and a grounding grid
of a substation. Then, by performing the analysis of the grounding system in XGSLab,
a file with the impedance values for all different frequencies is obtained. Thereafter, by
using fitting techniques such as Vector Fitting (VF) and ARMAFIT equivalent Frequency
Dependent (FD) impedances are developed. Subsequently, these FD impedances are in-
cluded on the software EMTP/ATP and transient simulations that include the grounding
systems are conducted. Finally, all results are extracted and are plotted using MATLAB.

5.2. Fitting Methods


Two different methods were found that can be used to convert the output from XGSLab
to usable information in the transient software tool ATP, namely: VF and ARMAFIT. Both
methods, can be applied on the output from XGSLab and fit the frequency responses of
the structure. The accuracy obtained from VF was found to be higher than ARMAFIT.
However, ARMAFIT is faster and can also be applied to the output produced from VF
which might increase the accuracy of the equivalent generated model.

Both methods are described in the following sections.

5.2.1. Vector Fitting Method


The first option to convert the results obtained from XGSLab into the EMTP/ATP soft-
ware tool is, by applying the vector fitting (VF) method. By using the matrix fitting tool-
box provided by Sintef, the calculated frequency responses of simple grounding struc-
tures, high voltage towers and a substation will be converted into rational functions.

N
Rm
Y (s) = C (sI − A)−1 B + D + sE
X
Y (s) = s−a m + D + sE (5.1)
m=1

Where a m are the poles, R m are the residues, D and E are optional parameters. The initial
MATLAB function that is included, is the VFdriver.m which identifies models such as in
equation (5.1) by using the pole relocating vector fitting technique described in [34], [35]
and [36]. Afterwards, it stacks the upper triangle of the admittance matrix into a single
column and then it is fitted by vector fitting using a common set of poles. The user must
take into account that all input frequency samples should be positive and that the num-
ber of frequencies used are more than the fitting order selected in the function [37].

The next step is to use the MATLAB function Netgen_ATP.m. This function exports the
results to a TXT file containing RLC values. This text file can be included in EMTP soft-
ware to make simulations in time domain. However, the user should keep in mind that
this model is passive. Hence, it is not able to generate power by itself, so it should be
connected to a source. To verify the accuracy of the conversion, a simple AC source is
connected to perform a frequency scan in ATPDraw to verify that the results obtained
match the result of XGSLab. The frequency scan can either be done by:

• Drawing the circuit manually according to the output in the TXT file
5. Modelling Approach - Frequency Dependent Models 41

• Include the TXT file as a library model directly in EMTP/ATP

• Use the result of VF in combination with ARMAFIT (described below) to make the
ATP ’KIZILCAY F-DEPENDENT model

5.2.2. ARMAFIT Method


The frequency response of the modelled structures can also be fitted using ARMAFIT.
This method scans the network admittances by a rational function either in z domain or
in Laplace domain and its result can be directly used in the ATP model by using the com-
ponent called ’KIZILCAY F-DEPENDENT (KFD)’ [38]. The advantage of ARMAFIT is that
it can be more easily implemented than VF (no additional software needed). However,
based on some comparisons it seems to have a lower accuracy compared to VF, espe-
cially in more complex situations such as bigger grounding structures and higher values
of permittivity. Moreover, for grounding structures that have a negative phase angle for
their impedance, the ARMAFIT approach fails to provide an equivalent model. An expla-
nation on how to use the ARMAFIT method to get the frequency response of a network
(grounding structure) is given in the Appendix A.

5.3. Modelled Structures


In this section, the ATP models that were developed from VF and ARMAFIT are verified
by comparing their frequency response to the frequency response of the same grounding
structure that was modelled in XGSLab. The modelled grounding structures include:
a vertical conductor, a ring conductor, a high voltage tower and a substation. Finally,
after verifying these models, transient simulations will be conducted in chapter 6, with
the grounding systems included. The way to perform a frequency scan in EMTP/ATP to
verify the models produced in XGSLab can be seen in Figure 5.2.

Figure 5.2: Perform frequency scan using ATPdraw

5.3.1. Vertical Conductor


Two cases were investigated for the vertical conductor whose characteristics can be seen
in Table 5.1.
5. Modelling Approach - Frequency Dependent Models 42

Table 5.1: Characteristics of the Vertical Conductor for two cases

Characteristics Case A Case B


Length 9m 9m
Burial Depth - -
Soil Resistivity 100 Ωm 100 Ωm
Relative Permittivity 1 10
Conductor Diameter 10 mm 10 mm

The soil resistivity for both cases was chosen at 100 Ωm as a well suited generic value
of the homogenic soil in the Netherlands while the relative permittivity for case A is set
to 1 and for case B to 10. The value of case A, was a value selected to test the fitting
methods, while the value for case B was selected as a good representative value, of the
soil characteristics in the Netherlands.

5.3.1.1. Case A
To fit the frequency response of a grounding electrode with permittivity equal to 1 is
quite easy. Its frequency response using VF can be fitted with an order of approximation
(N=5). The higher the order of approximation selected, the bigger the RLC network that
will be generated. Therefore, for Case A a rather limited circuit is sufficient to generate
the expected results. For illustrative purposes, it was manually drawn in ATPDraw and
can be seen in Figure 5.3. However, this will be avoided for the other grounding struc-
tures since the circuits will be larger and it is not practical.

Figure 5.3: Equivalent frequency dependent model for a vertical conductor, case A

By making a frequency scan, using the ATP models generated from VF and ARMAFIT, it
is observed that the result in Figure 5.4 matches the result perfectly with the response of
the vertical conductor modelled in XGSLab. The values of the circuit can be found in the
Appendix B.
5. Modelling Approach - Frequency Dependent Models 43

70 40
XGSLab XGSLab
ATP VF ATP VF
35
60 ATP Armafit ATP Armafit

30

Phase Angle [Deg]


50
Impedance [Ω]

25

40 20

15
30

10

20
5

10 0
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Impedance of vertical conductor - Case A b) Angle of vertical conductor - Case A

Figure 5.4: Frequency response of a vertical conductor, comparison of XGSLab and ATP

When ARMAFIT is used to fit the result obtained from VF, the generated result is usu-
ally more accurate than applying ARMAFIT directly to the result obtained from XGSLab.
In the next sections, the ARMAFIT approach will be used in every grounding structure
possible, as it provides the option to export the result in a Kizilcay model.

5.3.1.2. Case B
To fit the frequency response of the vertical conductor studied in this case, the approxi-
mation order (number of poles) was set equal to N=15 which resulted in a bigger circuit
as previously explained seen in Figure 5.5.

Figure 5.5: Equivalent frequency dependent model for a vertical conductor, case B

The frequency scan on the software tool EMTP/ATP can be performed, either by using
the circuit shown in Figure 5.5 or by generating an equivalent kizilcay model, using the
ARMAFIT approach. The result of doing this is shown in Figure 5.6, where it is clearly
seen that the impedance curve of the grounding structure was successfully interfaced
on the EMTP/ATP software tool. The values for the circuit shown in Figure 5.5, can be
found in the Appendix B.
5. Modelling Approach - Frequency Dependent Models 44

70 40
XGSLab XGSLab
ATP ATP
35
60

30

Phase Angle [Deg]


50
Impedance [Ω]

25

40 20

15
30

10

20
5

10 0
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Impedance of vertical conductor - Case B b) Angle of vertical conductor - Case B

Figure 5.6: Frequency response of a vertical conductor, comparison of XGSLab and ATP, Case B

5.3.2. Ring Conductor


For the ring conductor, two cases are considered, one for a soil with constant parameters
(ρ = 100 Ωm and ²r = 10) and one with a frequency dependent soil. Their characteristics
can be seen in Table 5.2.

Table 5.2: Characteristics of the Ring Conductor for two cases

Characteristics Case A Case B


Length - -
Burial Depth 1m 1m
Radius 10 m 10 m
Soil Resistivity 100 Ωm 100 Ωm
Relative Permittivity 10 Frequency dependent soil
Conductor Diameter 10 mm 10 mm

5.3.2.1. Case A
For this case, in order to fit the frequency response of this ring conductor using VF, an
approximation order of N=35 was used. This results in a much larger RLC network which
is impractical of drawing. Therefore, the result can be included in a library model to
perform the frequency scan. The result of this case can be seen in Figure 5.7.
5. Modelling Approach - Frequency Dependent Models 45

70 45
XGSLab XGSLab
ATP 40 ATP
60

35
50

Phase Angle [Deg]


30
Impedance [Ω]

40 25

30 20

15
20
10

10
5

0 0
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Ring conductor impedance - Case B b) Ring conductor angle - Case B

Figure 5.7: Ring conductor, XGSLab and ATP comparison, Case B

5.3.2.2. Case B
Similarly, a ring conductor with frequency dependent soil was fitted and the result is
shown in Figure 5.8.

40 40
XGSLab XGSLab
ATP 35 ATP
35

30
30
Phase Angle [Deg]

25
Impedance [Ω]

25
20
20
15
15
10

10
5

5 0

0 -5
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Ring conductor impedance - Case B b) Ring conductor angle - Case B

Figure 5.8: Ring conductor, XGSLab and ATP comparison, Case B

Finally, to show that the soil dependency with frequency does not greatly influence the
result, both cases are plotted together in Figure 5.9.
5. Modelling Approach - Frequency Dependent Models 46

60
ρ=ρ0, εr=10
ρ=ρ(ω), ε=ε(ω)
50

40

Impedance [Ω] 30

20

10

0
101 102 103 104 105 106 107
Frequency [Hz]

Figure 5.9: Ring conductor impedance - Case B

As found in chapter 4, the impedance drop is very small, so small differences are ex-
pected when their response to lightning is simulated in the next chapter.

5.3.3. High Voltage Towers


In this section, frequency dependent tower models are built, in order to conduct simu-
lations in a transmission line model, with the grounding part of the high voltage tower
being modelled as a FD impedance. For the high voltage line, Wintrack towers were
chosen to be modelled, as their footing design is very different in comparison to lattice
towers and in most cases more complex to model. The characteristics of the towers that
were simulated can be seen in Table 5.3.

Table 5.3: Characteristics of high voltage towers

Characteristics Tower 18 Tower 19


Number of poles 48 24
Length of Poles 22.27 m 14.27 m
Soil Resistivity 100 Ωm 100 Ωm
Relative Permittivity 10 10

5.3.3.1. DTC-NDR380 Wintrack Tower 18


The base of the DTC-NDR380 Wintrack tower 18 consists of 2 circles with a radius of 6
metres. The two circles are apart from each other with a side length of 1.615 metres and
the two towers are at a distance of 3.6 metres away from each other. Finally, the tower
foundation is completed by placing 48 poles in total, 24 in each tower, spaced 15◦ apart
from each other and extending to 22.27 metres deep into the ground. The representation
of the tower can be seen in Figure 5.10.
5. Modelling Approach - Frequency Dependent Models 47

Figure 5.10: Grounding system of Wintract tower 18

Similarly, the fitting response of a high voltage tower was found. The comparison be-
tween XGSLab and ATP can be seen in Figure 5.11.

30 60
XGSLab XGSLab
ATP ATP
50
25

40
Phase Angle [Deg]

20
Impedance [Ω]

30
15
20

10
10

5
0

0 -10
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) High voltage tower impedance - Tower 18 b) High voltage tower angle - Tower 18

Figure 5.11: High voltage tower 18, XGSLab and ATP comparison

5.3.3.2. DTC-NDR380 Wintrack Tower 19


The base of tower 19 consists of 2 circles with a radius of 4 metres. The 2 circles are apart
from each other with a side length of 1.348 metres and the two towers are at a distance of
8.4 metres away from each other. Finally, the tower foundation is completed by placing
5. Modelling Approach - Frequency Dependent Models 48

24 poles in total, 12 in each tower, spaced 30◦ apart from each other and extending 14.27
metres deep into the ground.

Figure 5.12: Grounding system of Wintract tower 19

The result obtained for tower 19 is very similar to that of the high voltage tower 18. There-
fore, for this tower an equivalent EMTP/ATP frequency model is disregarded. This shows
that minor modelling adjustments, do not affect the result as much as the soil character-
istics of the location.

5.3.4. Boxtel 150kV Substation


The substation selected for this thesis is Boxtel 150kV, since a transient study for this sub-
station has already been performed in the past [39]. In Figures 5.13, 5.14 the modelled
grounding grid and an overview of the Boxtel substation are shown. The locations for
current injection are shown in Figure 5.13. These locations include, a common ground-
ing point of the transformer (location D) and three possible locations to place a surge
arrester (locations A to C).
5. Modelling Approach - Frequency Dependent Models 49

Figure 5.13: Grounding grid of Boxtel 150kV substation along with the locations of current injection
5. Modelling Approach - Frequency Dependent Models 50

C
n
io
at
m
or
nI f
l a
rn B
te
In
C

Figure 5.14: Drawing of Boxtel substation


5. Modelling Approach - Frequency Dependent Models 51

5.3.4.1. Network Approach, N-ports system


The software tool XGSLab is able to calculate the impedances of the above mentioned lo-
cations but it cannot directly compute the impedance between these locations, usually
referred to as coupling impedance. Therefore, a network approach similar to [1], [40] is
used, resulting in FD models for transient simulations (reported in chapter 6). This ap-
proach considers a system as a network of N ports, where N can be any integer number.
A port is a pair of terminals and through it a current may flow. The current may enter or
exit the network as seen in Figure 5.15. Some examples of simple 1-port systems are: a
simple resistor, a capacitor or an inductor (or a combination of the three).

Figure 5.15: Example of an 1-port system

The presented substation, is modelled as a 2-port system where one port is the trans-
former point and the other is the surge arrester point. A simple illustration of such a
system is shown in Figure 5.16.

Figure 5.16: Example of a 2-port system

The system of Figure 5.16, can be described with the system (5.2) or (5.3) [41].

V1 = Z11 I 1 + Z12 I 2
½
(5.2)
V2 = Z21 I 1 + Z22 I 2

I 1 = Y11 V1 + Y12 V2
½
(5.3)
I 2 = Y21 V1 + Y22 V2
More specifically, a two-port network in its matrix form, can be described in two different
ways:

a. Using an admittance matrix as seen in equation (5.4), should the current be mod-
elled as a function of the voltage.

b. Using an impedance matrix as seen in equation (5.5), should the voltage be mod-
elled as a function of the current.
5. Modelling Approach - Frequency Dependent Models 52

· ¸ · ¸· ¸
I1 Y11 Y12 V1
= (5.4)
I2 Y21 Y22 V2
· ¸ · ¸· ¸
V1 Z11 Z12 I1
= (5.5)
V2 Z21 Z22 I2
In order to solve the system, the impedance or the admittance matrix need to be calcu-
lated. The approach to calculate them is the same. Therefore, only the calculation of the
admittance matrix will be shown, since the fitting methods are applied to the admittance
values. However, the impedance curve will be plotted for the results as this is more often
seen in articles. In order to calculate the admittance matrix, first the output and then the
input is short circuited.
By short circuiting the output V2 : By short circuiting the input V1 :

• Y11 = VI 11 ¯V2 =0 Input admittance • Y12 = VI 12 ¯V1 =0 Reverse transfer admit-


¯ ¯

tance
• Y21 = VI 21 ¯V2 =0 Forward transfer
¯
• Y22 = VI 22 ¯V1 =0 Output admittance
¯
admittance

As previously stated, the substation is modelled as a two-port network and can also be
simplified using a π which is seen in Figure 5.17. Sometimes this approach is preferred
as it reduces the number of admittances that require fitting. Nevertheless, it is more ac-
curate to include the whole Y or Z matrix, as the π circuit is a valid simplification when
Y12 and Y21 are equal. Most times, their values were found to be very close or identi-
cal. However, for this case the π approach does not reduce the number of admittances
that the fitting methods need to be applied to. This is based on the fact that three dif-
ferent placements for the surge arrester are considered. A different placement each time
means: the surge arrester self admittance as well as the coupling admittance have their
values changed and need to be recalculated. Hence, considering a π approach results
in a different admittance value for the transformer point as well since it depends on the
coupling admittance (Y2 = Y22 + Y12 ). Therefore, for the substation study the whole Y-
matrix is connected to the EMTP/ATP model.

Figure 5.17: π circuit approach, 2-port system


5. Modelling Approach - Frequency Dependent Models 53

By looking at how the current is divided when entering from port-1 and from port-2 the
equation system (5.6) is obtained.

I 1 = Y1 V1 + Y2 (V1 − V2 ) ⇒ I 1 = (Y1 + Y2 )V1 − Y2 V2


½
(5.6)
I 2 = Y2 (V2 − V1 ) + Y3 V2 ⇒ I 2 = −Y2 V1 + (Y 2 + Y 3)V2
By comparing equation (5.6) with equation (5.3) the values of the π circuit can be solved.

Y11 = Y1 + Y2 Y12 = Y21 = −Y2 Y22 = Y22 + Y12 (5.7)


Finally, by rearranging the terms, the admittances of the π circuit are:
• Y1 = Y11 + Y12

• Y2 = −Y12

• Y3 = Y22 + Y12
The three different cases that are considered, each case representing a different location
for the surge arrester, can be seen in Figure 5.13. For all three cases, the self-impedance
of the transformer remains the same, as its grounding point does not change. The val-
idation of the self-impedance for the transformer point (Z22 ) is shown in Figure 5.18.
Thereafter, for every different location, the self impedance (Z11 ) of the surge arrester as
well as the coupling impedance (Z21 ) are shown in the following subsections. For all
cases the self impedances were directly fitted using the ARMAFIT approach while the
coupling impedances were fitted using VF, as ARMAFIT cannot fit admittances where
the phase angle is positive. For these three cases the grounding grid was symmetrical
so the impedance values Z21 and Z12 are identical and, therefore, for all cases only the
impedance Z21 will be shown.

Finally, as mentioned earlier, this network approach can be also extended to systems
with a larger number of ports. For example, a 3-port system would be used for two surge
arresters present in the substation (near the power transformer and at line entrance).

35 50
XGSLab XGSLab
ATP ATP
30 40

25 30
Phase Angle [Deg]
Impedance [Ω]

20 20

15 10

10 0

5 -10

0 -20
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Impedance point A b) Phase angle point A

Figure 5.18: Transformer point impedance (Z22)


5. Modelling Approach - Frequency Dependent Models 54

5.3.4.2. Surge arrester at location A


For this case, the surge arrester is placed at location A, as seen in Figure 5.13. The equiv-
alent impedances that were calculated and validated with the results of XGSLab can be
seen in Figure 5.19.

1.1 0
XGSLab XGSLab
1 ATP ATP
-10
0.9

Phase Angle [Deg]


0.8 -20
Impedance [Ω]

0.7
-30
0.6

0.5 -40

0.4
-50
0.3

0.2 -60
101 102 103 104 105 106 101 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

c) Impedance between point A and D d) Phase angle between point A and D

40 60
XGSLab XGSLab
ATP ATP
35 50

30
40
Phase Angle [Deg]
Impedance [Ω]

25
30
20
20
15

10
10

5 0

0 -10
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Impedance point D b) Phase angle point D

Figure 5.19: Substation - Case A, XGSLab and ATP comparison

5.3.4.3. Surge arrester at location B


For this case, the surge arrester is grounded at point B as seen in Figure 5.13. The equiv-
alent validated impedance curves between XGSLab and EMTP/ATP are shown in Figure
5.20.
5. Modelling Approach - Frequency Dependent Models 55

1.1 0
XGSLab XGSLab
ATP ATP
-5
1

-10

Phase Angle [Deg]


0.9
Impedance [Ω]

-15

0.8 -20

-25
0.7

-30

0.6
-35

0.5 -40
101 102 103 104 105 106 101 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

c) Impedance between points B and D d) Phase angle between points B and D

35 60
XGSLab XGSLab
ATP ATP
30 50

25 40
Phase Angle [Deg]
Impedance [Ω]

20 30

15 20

10 10

5 0

0 -10
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Impedance point D b) phase angle point D

Figure 5.20: Substation - Case B, XGSLab and ATP comparison

5.3.4.4. Surge arrester at location C


For this case, the surge arrester is grounded at location C as seen in Figure 5.13. The
equivalent impedances that were calculated and validated with the results of XGSLab
can be seen in Figure 5.21.
5. Modelling Approach - Frequency Dependent Models 56

2.2 20
XGSLab XGSLab
ATP ATP
10
2

Phase Angle [Deg]


1.8
Impedance [Ω]

-10
1.6
-20

1.4
-30

1.2
-40

1 -50
101 102 103 104 105 106 101 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

c) Impedance between point C and D d) Phase angle between point C and D

35 60
XGSLab XGSLab
ATP ATP
30 50

25 40
Phase Angle [Deg]
Impedance [Ω]

20 30

15 20

10 10

5 0

0 -10
101 102 103 104 105 106 107 101 102 103 104 105 106 107
Frequency [Hz] Frequency [Hz]

a) Impedance point D b) Phase angle point D

Figure 5.21: Substation - Case C, XGSLab and ATP comparison

5.4. Discussion
In this chapter, equivalent frequency dependent models for the grounding structures
were developed and validated.The frequency dependent models were developed by us-
ing either the VF method and/or ARMAFIT. From the two methods, ARMAFIT is easier
to use and integrate in ATPDraw (using the Kizilcay model). However, VF usually pro-
vides higher accuracy and can also fit more complex situations, whereas ARMAFIT of-
ten returns errors. In this chapter, the ARMAFIT approach was preferred for most cases
because of its simplicity and the fact that it can be more easily integrated with the EMT-
P/ATP software. VF was used in cases where ARMAFIT was provided inaccurate results
or errors. The equivalent kizilcay models developed from the ARMAFIT approach can be
found in the Appendix C.

For the grounding grid only, a slightly different approach was considered since the cou-
5. Modelling Approach - Frequency Dependent Models 57

pling impedance cannot be directly calculated using XGSLab. To calculate the relevant
impedances a 2-port network approach was considered and the grounding grid was
modelled, by connecting the calculated values of the Y-matrix, as shown in Figure 5.22.

6.25 m
V

6.25 m
U

6.25 m

1m 1m 1m
UI

I
MOV

MOV

MOV

2.2 m 2.2 m 2.2 m

Y12

BRH

Y11
F(s|z)

Y22

F(s|z)
BRH

Y21

Figure 5.22: Grounding grid as defined in the EMTP/ATP software

The blue lines represent where a connection is made by labeling the nodes accordingly.
As already mentioned, the equivalent models for the self admittances were developed
using the ARMAFIT method, while the coupling admittances were developed using VF
as ARMAFIT could not work to fit their response.
6
Transient Simulations - Results

6.1. Introduction
In this chapter, the frequency dependent models created in chapter 5 are used to per-
form transient simulations for fast-front transient overvoltages using the EMTP/ATP soft-
ware tool. First, for the simple grounding structures, such as the vertical conductor and
the ring conductor, the time domain result in ATP/EMTP is compared to the time do-
main result obtained in XGSLab. Next, different waveshapes of the Heidler function are
injected to the frequency dependent model, and are compared to a simple (equivalent)
resistor calculated using Sunde’s formulas [31]. The results are compared to demonstrate
the difference between the two methods. The values of the frequency dependent models
that were used in the simulations can be found in the Appendix C.

Furthermore, more complex simulations are then performed for a high voltage tower
footing impedance in an overhead line and for a substation. For the overhead line the
380kV/150kV combined Wintrack line Doetinchem-Niederrhein was selected, as multi-
ple measurements on these towers have been performed. For the substation the 150kV
Boxtel was selected as an extensive study had already been performed for this substation
in the past [39].

Similarly, for both cases the results are compared to a resistor value calculated at power
frequency using XGSLab. Furthermore, the systems were tested for higher soil resistivi-
ties and for smaller front times of lightning, to observe the increase in the system over-
voltages. Finally, for the substation, more cases are investigated, such as (1) determining
the right place to install a surge arrester and (2) Optimizing the topology of the ground-
ing grid to decrease the overvoltages occurring at the transformer terminal.

6.2. Vertical Conductor


To show the effect for a vertical conductor, only the frequency dependent model from
section 5.3.1.1. was only considered since this variant had a very small decrease of its

58
6. Transient Simulations - Results 59

impedance for an increased value of the relative permittivity and mainly in the very high
frequency region.

6.2.1. Time domain comparison XGSLab - EMTP/ATP


In both software tools, a time domain calculations was performed by injecting a light-
ning current. In XGSLab the calculation was performed by injecting a lightning current
(Heidler waveshape: 31kA, 1.2/75 µs). However, to approximate a similar result in EMT-
P/ATP, different values were used for the front and tail time. In EMTP/ATP, a lightning
current (Heidler 31kA, 4.2/70 µs) was injected instead. The result can be seen in Figure
6.1, where it can be noticed that a very good comparison for the GPR between the two
software packages is obtained.

105
4.5
XGSLab
4 ATP

3.5

2.5
GPR [V]

1.5

0.5

0
0 5 10 15 20 25 30

Figure 6.1: Time domain comparison between XGSLab and ATP

The values used in both cases to approximate similar results are shown in Table 6.1.

Table 6.1: Characteristics of the Heidler function for XGSLab and EMTP

Software Front time [µs ] Tail time [µs ]


XGSLab 1.2 75
EMTP/ATP 4.2 70

The reason that different times for the front peak and its tail were applied is that the
Heidler function is modelled in EMTP/ATP as defined in the IEC 62305-1. A representa-
tion of that Heidler waveshape and its characteristics are shown in Figure 2.4 while the
XGSLab has the Heidler function defined as in equation (2.1) in section 2.3.1. In the next
grounding structure (ring conductor), the Heidler function modelled in XGSLab is mod-
elled in EMTP/ATP by using the MODELS tool of the software. This enables the user to
6. Transient Simulations - Results 60

inject the exact same current waveshape and obtain a better comparison between the
two software tools.

6.2.2. EMTP/ATP comparison of frequency dependent model and resistor


A calculation was performed to show the difference between using a resistor and an FD
model to represent the grounding system. A resistor with a value for the vertical con-
ductor using Sunde’s equation was calculated and it is compared with the FD model
developed in chapter 5. The equivalent equation (6.1) given by Sunde in [31] is:

ρ
³ ´
R = 2πL l n 8L
d − 1 (6.1)

Where,

• ρ is the soil resistivity

• L is the length of the conductor

• d is the diameter of the conductor

Based on the values from Table 5.1 a resistor value of 12.7 Ω is determined, which is also
similar to the value calculated by XGSLab at power frequency. In Figure 6.2 a) and b), the
resulting GPR for the resistor value and the frequency dependent model are seen for two
different lightning waveshapes.

105 105
6 6
FD model FD model
Resistor Resistor
5 5

4 4
GPR [V]

GPR [V]

3 3

2 2

1 1

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30

a) Case A - front time 4.2 µs b) Case B - front time 1.2 µs

Figure 6.2: Ground potential rise for the vertical conductor for different lightning waveshapes

In Figure 6.2 a), for a lightning waveshape of 31 kA (4.2/70 µs), it can be observed that,
the GPR for both the FD model and the resistor are very close. Therefore, a resistor with
a value of 12.7 Ω would slightly underestimate the generated overvoltages for this case.
However, for a steeper lightning waveshape with a front time of 1.2 µs, the impedance
will be higher and so the resulted GPR. This is presented in Figure 6.2 b).
6. Transient Simulations - Results 61

6.3. Ring Conductor


For the ring conductor the FD models developed in section 5.3.2 are investigated in a
similar manner as with the vertical conductor. However, in this case, there is an extra FD
model developed with frequency dependency of soil included.

6.3.1. Time domain comparison: XGSLab and EMTP/ATP


In order to perform a better comparison between the software tools, the Heidler function
was modelled on MODELS of EMTP/ATP. The equivalent model is shown in Figure 6.3,
while the code can be found in the Appendix E.

MODEL
heidler
I

Figure 6.3: Heidler function as defined in EMTP/ATP

By injecting a lightning waveshape in both software tools (Heidler waveshape: 31kA,


1.2/75 µs) the voltage response for the ring conductor is obtained in Figure 6.4. For the
rest of this chapter this comparison will not be included as it is time consuming and
impractical since all other simulations are carried out using EMTP/ATP.

Figure 6.4: Time domain comparison between XGSLab and ATP

The standard lightning waveshape of the EMTP/ATP will be considered for the next sim-
ulations, so that similar front and tail times for all the grounding structures can be seen
through this report.
6. Transient Simulations - Results 62

6.3.2. EMTP/ATP comparison of frequency dependent model and resistor


Similar to the vertical conductor the FD models of Case A and Case B created in section
5.3.2 are compared with an equivalent resistor calculated using Sunde’s equation (6.2).
Case A considers a ring conductor with constant soil parameters, while Case B a ring
conductor with frequency dependent soil.

ρ
³ ´
R = 2π2 D l n 8D 4D
d +ln h (6.2)

Where,

• ρ is the soil resistivity

• D is the diameter of the circle (2*radius)

• d is the diameter of the conductor

• h is the burial depth

The resistance using as input the values from Table 5.2 in chapter 5 was calculated at 3.56
Ω. The GPR comparison between the resistor and the FD models can be seen in Figure
6.5 a) and b).

105 105
4 4
Case A Case A
3.5 Case B 3.5 Case B
Resistor Resistor

3 3

2.5 2.5
GPR [V]

GPR [V]

2 2

1.5 1.5

1 1

0.5 0.5

0 0
0 5 10 15 20 25 0 5 10 15 20 25

a) Case A - front time 4.2 µs b) Case B - front time 1.2 µs

Figure 6.5: Ground potential rise for the ring conductor for different lightning waveshapes

Figure 6.5, clearly shows that for a ring conductor the impedance value has a bigger ef-
fect on the resulting structure voltage. This was expected as the resistor value of the ring
conductor at power frequency is much lower than that of the vertical conductor. There-
fore, greater differences were expected. Finally, the FD model that considers a frequency
dependent soil does not differ significantly from the one with the constant soil parame-
ters which was already hypothesized in chapter 4.
6. Transient Simulations - Results 63

6.4. High Voltage Towers - Doetinchem line


6.4.1. EMTP/ATP comparison of frequency dependent model and resistor
In this section, the tower footing of the 380kV/150kV combined overhead line Doetinchem-
Niederrhein is compared for two different cases: a tower footing resistor and a tower
footing impedance (modelled as Kizilcay-component).

For this a network of 7 Wintrack towers is simulated, connected with conductors over
a span of 400 metres in between. The overhead lines were modelled using the Berg-
eron model at a frequency equal to 120 kHz. The tower structure (surge impedance) was
modelled by connecting 7 Bergeron models at similar frequency, with a total height of
37.5 metres and its travelling speed, is close to the speed of light and equal to 3 · 105 .

Finally, the towers are connected from their structure to the ground, by connecting the
structure to a resistor or an FD model as described in section 5.3.3.1. The resistance
value of the tower at power frequency was calculated at 1 Ω. The simulated model is
presented in Figure 6.6.

Figure 6.6: Transmission line system of Doetinchem - 380kV, 7 Wintrack towers connected

The lightning current (Heidler waveshape of 31kA, 4.2/70 µs) is injected into the middle
tower (tower 19) and the generated overvoltages for towers 19 to 22 are shown in Figure
6.6. For tower 19, the same lightning waveshape is also injected directly into the FD
model.
6. Transient Simulations - Results 64

104 104
7 2
FD model FD model
Resistor Resistor
6 Ground system 1.5

5
Tower overvoltages [V]

Tower overvoltages [V]


1

4
0.5
3

0
2

-0.5
1

0 -1
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

a) Overvoltage at Wintrack tower 19 b) Overvoltage at Wintrack tower 20

10000 10000
FD model FD model
8000 Resistor 8000 Resistor

6000 6000
Tower overvoltages [V]

Tower overvoltages [V]

4000 4000

2000 2000
2.8 m
0 0

-2000 -2000

-4000 -4000

-6000 -6000

-8000 -8000
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

c) Overvoltage at Wintrack tower 21 d) Overvoltage at Wintrack tower 22

Figure 6.7: Overvoltages in a 380kV/150kV transmission line system, for an lightning waveshape of 31kA
(4.2/70 µs)

As seen in Figure 6.7, the generated overvoltages when the tower footing is represented
as a FD impedance are considerably higher for every tower compared to, when repre-
sented by a resistor. The difference of injecting the lightning to the whole tower struc-
ture is that the generated overvoltages decrease slightly and the rise time for the voltage
occurs slightly later because of the extra time needed for the wave to travel. That time
can be calculated as the total distance divided by the speed (37.5m/3 · 105 m/s).

Similarly, with the vertical and the ring conductor by decreasing the front time to 1.2 µs
the generated overvoltages are shown in Figure 6.8. However, for this case, it is compared
with the previous waveshape with with 4.2 µs front time, since the resistor has a very
small value at power frequency and the differences are extremely high, especially for the
adjacent towers.
6. Transient Simulations - Results 65

104 104
11 11

8 8
Tower overvoltages [V]

Tower overvoltages [V]


6 6

4 4

2 2

0 0

-2 -2

-4 -4
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

a) Overvoltage at Wintrack tower 19 b) Overvoltage at Wintrack tower 20

104 104
4 4

3 3

2 2
Tower overvoltages [V]

Tower overvoltages [V]

1 1

0 0

-1 -1

-2 -2

-3 -3

-4 -4
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

c) Overvoltage at Wintrack tower 21 d) Overvoltage at Wintrack tower 22

Figure 6.8: Overvoltages in a 380kV/150kV transmission line system,for an lightning waveshape of 31kA
(4.2/70 µs)

As expected, the resultant overvoltages shown in Figure 6.8 are considerably higher, in
particular the overvoltages seen for the adjacent towers are now about three times higher.

6.4.2. Effect of different tower footing impedance values


The application of a Kizilcay-component for the tower footing impedance enables the
user to easily adjust its value by adjusting the gain. This can be considered applicable for
similar tower footings in different soil types although this is not strictly accurate as soils
of high soil resistivity behave differently to soils of low soil resistivity for high frequencies.
Therefore, using this approach, an approximation for different soil resistivities (gains) is
shown in Figure 6.9 a), b), c) and d) for a lightning discharge (Heidler 31kA, 4.2/70 µs).
6. Transient Simulations - Results 66

105 104
5 20
Gain=1 Gain=1
4.5 Gain=0.5 Gain=0.5
Gain=0.25 Gain=0.25
4 Gain=0.1 15 Gain=0.1

3.5
Tower overvoltages [V]

Tower overvoltages [V]


3 10

2.5

2 5

1.5

1 0

0.5

0 -5
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

a) Overvoltage at Wintrack tower 19 b) Overvoltage at Wintrack tower 20

104 104
10 10
Gain=1 Gain=1
Gain=0.5 Gain=0.5
8 Gain=0.25 8 Gain=0.25
Gain=0.1 Gain=0.1

6 6
Tower overvoltages [V]

Tower overvoltages [V]

4 4

2 2

0 0

-2 -2

-4 -4
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

c) Overvoltage at Wintrack tower 21 d) Overvoltage at Wintrack tower 22

Figure 6.9: Overvoltages in a 380kV/150kV transmission line system, for different soil resistivities

From Figure 6.9, it is obvious that the amplitude for the tower overvoltages increases
dramatically for higher soil resistivities and lasts for a longer time. This effect can be
easily observed in Figure 6.9 c), where with the use of 0.1 gain, it can be clearly seen that
the voltage peaks last longer. This is especially important for insulation coordination
studies, as a backflash over to the insulator could have occurred and with the grounding
system being modelled as a resistor, it could have not been correctly predicted.

Finally, a worst case scenario is presented, where the Kizilcay admittances have a gain of
0.1 and a steep lightning waveshape (Heidler 31kA, 1.2/70 µs) is injected. The generated
overvoltages for all towers are shown in Figure 6.10.
6. Transient Simulations - Results 67

105
10
Tower 19
Tower 20
8 Tower 21
Tower 22

Tower overvoltages [V]


6

-2
0 2.5 5 7.5 10 12.5 15

Figure 6.10: Tower overvoltages, worst case scenario

6.5. Boxtel150kV Substation


6.5.1. Introduction
In the previous chapters, the behaviour of grounding structures was investigated step-
wise. In this section, the influence of the grounding structures inside a substation is
determined. Generally, direct strokes at the substation terminals are never taken into ac-
count for transient simulations, as it is presumed that the substation is always protected.
Therefore, a lightning strike at an adjacent tower is considered causing backflashover, as
well as a lightning strike directly at phase A (shielding failure).

For the protection of the primary equipment inside a substation, especially for the power
transformer, surge arresters are installed at specific points. The proper placement of the
surge arresters is crucial, since the distance directly affects the protection performance
of the arrester. This is especially the case for power transformers, since the connection
to a transformer behaves as an un-terminated end. Therefore, the surges will reflect to
practically double values. In order to reduce the overvoltages at the transformer ter-
minal, the placement of the surge arrester should be as close as possible to the power
transformer [42]. Thus, this argument will initially be proved, and thereafter with the
surge arrester in a fixed place, further results will be shown.

Finally, ways to optimize the topology of the substation grounding grid, in order to re-
duce the overvoltages at the power transformer terminal are investigated.

6.5.2. Model description


The substation model, was modelled on the electromagnetic transient program (EMT-
P/ATP) as shown in Figure 6.11. To determine the optimal placement for a surge arrester,
three possible locations were considered which are shown in Table 6.2. As discussed in
section 5.4, a different placement for a surge arrester will result in different self and cou-
6. Transient Simulations - Results 68

pling impedance. The different self and coupling impedances that were calculated in
section 5.3.4 are plotted together in Figure 6.12 for better comparison purposes. The
transmission line was modelled using Bergerons model at a frequency of 400kHZ, while
the body of the tower was modelled by its surge impedance in a similar manner as de-
scribed in section 6.4. The transformer was modelled as a simple capacitor to ground,
in accordance with IEC and the Nota Isolatiecoördinatie [43], [44], with a value of 1.5 nF.
The modelling of the transformer as a single capacitor according to recent literature is
valid but it is a simplified model [45], and thus provides lower accuracy. However, the
current modelling of the transformer will still provide good enough results to show the
differences occurring by including the grounding grid in a broad frequency range. Fi-
nally, the surge arresters are modelled according to, the specifications as used by TenneT
and described in the Nota isolatie coördinatie [39], [44].

Figure 6.11: Substation network as defined in the EMTP software tool

Table 6.2: Distance of surger arrester from transformer

Surge Arrester Distance


Location A 26.57 m
Location B 14.85 m
Location C 6.25 m
6. Transient Simulations - Results 69

25 2.2
Case A Case A
Case B 2 Case B
Case C Case C
20 1.8

1.6
Impedance [Ω]

Impedance [Ω]
15 1.4

1.2

10 1

0.8

5 0.6

0.4

0 0.2
101 102 103 104 105 106 101 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

a) Self impedance for the three different locations b) Coupling impedance for the three different locations

Figure 6.12: Impedance value for different placement of surge arrester

As shown in Figure 6.12 a), the impedance for Case B is the smallest since it represents a
location at the middle of the grounding grid. Thus, based solely on imepedance values,
Case B would result in the smallest GPR as shown in section 3.5.1. However, the lowest
overvoltages are expected to occur for Case C for the reasons already explained at the be-
ginning of this section. This therefore means that this methodology is vitally important
but can only draw significant conclusions when defining an appropriate placement to
install a second surge arrester. This is because the placement above and below ground
has a considerable effect on the overvoltages at the transformer terminal. Currently, the
Boxtel substation has a second surge arrester installed at its entrance, which makes it
interesting to determine whether it is optimal or whether placing it in the middle would
lower the incoming surge. However, it will not be part of this report due to time limita-
tions.

6.5.3. Calculation results direct strike


6.5.3.1. Different surge arrester locations
The overvoltages for the three different locations of the surge arrester are shown in Figure
6.13, for the situation in which a lightning current of 10 kA (Heidler waveshape, 1.2/50
µs) hits phase A of tower 1 as seen in Figure 6.11. This situation represents a shielding
failure situation, where the current amplitude is considered to be conservative, based on
the tower designs as mentioned in the Nota Isolatiecoördinatie.
6. Transient Simulations - Results 70

105 105
12 12
Location A Location A
10 Location B 10 Location B
Location C Location C

Overvoltages at surge arrester [V]


8 8
Overvoltages at Transformer [V]

6 6

4 4

2 2

0 0

-2 -2

-4 -4

-6 -6
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

a) Overvoltages at transformer b) Overvoltages at surge arrester

Figure 6.13: Overvoltages at transformer and surge arrester for a discharge current of 10 kA and 1.2/50 µs
directly at phase A - for surge arrester at different positions

Figure 6.13, shows that a surge arrester in ’Location C’, closest to the transformer results
in the lowest overvoltages which is as expected according to the theory and literature.

6.5.3.2. Different grounding structure options


Next, the position of the surge arrester is fixed to location C while the representation
of the grounding structure varies. A lightning current of 20 kA (Heidler waveshape, and
1.2/50 µs) again strikes phase A of tower 1. The overvoltages at the transformer and surge
arrester are shown in Figure 6.14, for three different cases: 1) Applying a resistive value
of 1.08 Ω calculated at power frequency, 2) for a soil resistivity of 100 Ωm, 3) for a soil
resistivity of 1000 Ωm.

105 105
14 10
106 105
1.3
8.8
12
1.25 8.7
Overvoltages at surge arrester [V]

10 8.6
Overvoltages at Transformer [V]

1.2
8.5
8 1.15 5 8.4
1.8 1.82 1.84 1.92 1.925 1.93
6 10-6 10-6

2 0

-2

-4 -5
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

a) Overvoltages at transformer b) Overvoltages at surge arrester

Figure 6.14: Overvoltages at transformer and surge arrester for a discharge current of 20 kA and 1.2/50 µs-
comparison between different soils
6. Transient Simulations - Results 71

From Figure 6.14 it can be observed that the soil resistivity plays an important role. More
specifically, when the grounding system is considered to be a simple resistance, the over-
voltages remain the lowest, however still comparable to the situation where the ground-
ing structure is considered as an impedance calculated at 100 Ωm. More noticeable dif-
ferences are present for the situation with a soil resistivity of 1000 Ωm.

6.5.4. Calculation results backflashover


6.5.4.1. Different surge arrester locations
Similarly, for the three different locations for the surge arrester, a lightning strike (Hei-
dler 100kA, 1.2/50 µs) occurring at the overhead shielding wire is also considered. The
generated overvoltages for the three cases are shown in Figure 6.15.

105 105
6 6
Location A Location A
Location B Location B
5 5
Location C Location C

Overvoltages at surge arrester [V]


Overvoltages at Transformer [V]

4 4

3 3

2 2

1 1

0 0

-1 -1

-2 -2
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

a) Overvoltages at transformer b) Overvoltages at surge arrester

Figure 6.15: Overvoltages at transformer and surge arrester for a discharge current of 100 kA and 1.2/50 µs at
shielding wires - for surge arrester at different positions

From Figure 6.15, it is observed that the lower overvoltages are found when the surge
arrester is closer to the power transformer (Case C).

6.5.4.2. Different grounding structure options


Moreover, a lightning strike to tower 1 was simulated for two different cases. The first
case takes the impedances of the tower and the grounding grid of the substation into ac-
count as frequency dependent models, where for the tower the ring conductor created in
section 5.4.2 represents its grounding system. In the second case, the grounding struc-
tures are modelled as resistors equivalent to impedance value at power frequency (3.5 Ω
and 1.08 Ω for the tower and substation respectively). The frequency dependent tower
footing is based on the impedance of a ring conductor (as presented in section 5.4.2). A
lightning current of 100 kA (Heidler, 1.2/50 µs) is applied.
6. Transient Simulations - Results 72

105 105
6 6
FD models FD models
5 Resistor 5 Resistor

4 4

Overvoltages at surge arrester [V]


Overvoltages at Transformer [V]

3 3

2 2

1 1

0 0

-1 -1

-2 -2

-3 -3

-4 -4
0 2.5 5 7.5 10 12.5 15 0 2.5 5 7.5 10 12.5 15

a) Overvoltages at transformer b) Overvoltages at surge arrester

Figure 6.16: Overvoltages at transformer and surge arrester for a discharge current of 100 kA and 1.2/50 µs, at
the shielding wires

Figure 6.16 shows that the overvoltages at both the transformer and the surge arrester
are significantly higher (double) for the case where all grounding structures are consid-
ered frequency dependent (FD). Furthermore, it also shows that the impedance of the
towers close to the substation have a great impact on the generated overvoltages at the
transformer terminal.

6.6. Improve the topology of the grounding grid to reduce over-


voltages
In this final section, possible ways to optimize the grounding grid topology of the substa-
tion (to reduce overvoltages at the transformer terminal) are explored. A simple solution
seems to be to place vertical conductors near the transformer terminal, so that they dis-
sipate the current flowing there directly into remote ground.

6.6.1. Addition of vertical electrode(s) at power transformer


6.6.1.1. Two electrodes of 15m
To study the effect of adding vertical electrodes, the grounding grid is extended with two
long vertical conductors of 15 metres in length are placed near the transformer terminal.
The diameter of the vertical conductors is selected to be 10mm which is equal to the rest
of the conductors of the grounding grid. The impedance of the grounding grid at the
transformer point (self impedance - Z22) is shown for two different soil resistivities, 100
Ωm and 1000 Ωm is shown in Figure 6.17 a) and b) respectively. The upper frequency is
set to 2 MHz, so that the effects of the grounding grid in the frequency region of lightning
can be more easily seen.
6. Transient Simulations - Results 73

25 80
Standard Standard
Extended Extended
70
20

60
Impedance [Ω]

Impedance [Ω]
15
50

40
10

30

5
20

0 10
101 102 103 104 105 106 101 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

a) Self impedance Z22 - soil resistivity 100 Ωm b) Self impedance Z22 - soil resistivity 1000 Ωm

Figure 6.17: Transformer point impedance - for additional conductors in the grounding grid

Figure 6.17 a) shows that for a soil of 100 Ωm the impedance at the transformer point is
decreasing with the frequency by placing two vertical conductors near the transformer
grounding point. However, in Figure 6.17 b), where the soil resistivity is 1000 Ωm the
impedance is not really influenced up to frequencies of 0.5 MHz.

6.6.1.2. Effective length of an electrode


In addition to the previous section, the effect of different lengths of a single vertical elec-
trodes was studied to determine the most effective length of a vertical electrode, as in
[32], [46]. However, in this work, the results will be shown in a broad frequency range. In
[46] this was done for one value of frequency that could represent a first stroke and for
another frequency that could represent subsequent strokes with the result shown as a
function of the conductor’s length. The results obtained here are exactly the same, how-
ever, an even better comparison could have been made if the frequency values they used
were known.

Figure 6.18 a) and b) show the effect of applying different electrode lengths for two elec-
trodes in a soil with a resistivity of 100 Ωm and 1000 Ωm respectively with a relative
permittivity of 10.
6. Transient Simulations - Results 74

30 200
5 5
6 180 6
7 7
25 8 8
160
9 9
10 10
11 140
Impedance [Ω]

Impedance [Ω]
11
20 12 12
13 120 13
14 14
15 15
100
15 30
40
80

60
10

40

5 20
102 103 104 105 106 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

a) Effect of length in a soil with 100 Ωm resistivity b) Effect of length in a soil with 1000 Ωm resistivity

Figure 6.18: Effect of different lengths of vertical conductors - for different soil resistivities

Figure 6.18 a), shows that longer electrodes result in lower impedance values for fre-
quencies in the region of 100-300 kHz. However, from 300kHz upwards, the 9 metre long
conductor has lower impedance. The changes from 9 to 13 metres in length are very
small, so it can be said that the most effective length for a soil of 100Ωm is 9 metres
which is in agreement with the result in [32]. From Figure 6.18 b), the effective length
for a conductor in a soil of 1000Ωm will be determined. As shown in Figure 6.18 b), the
impedance of a vertical conductor of 15 metres in length is much higher (about 80 Ω)
than the impedance of the grounding grid (10 Ω). However, for vertical conductors of
longer length, the impedance decreases considerably. For changes from 30 to 40 me-
tres, the impedance in the higher frequency region (600kHz-10 MHz) is not affected so
greatly and also exceeds the value of the 15 metre length conductor. Therefore, it can
be said that the most effective length for a soil of 1000 Ωm is approximately 30 metres
for the first stroke and 15 metres for subsequent strokes, which is also in agreement with
the literature in [32], [46]. These results provide a very important insight into how future
grounding grids are built since it appears that for soils of higher soil resistivity, longer
vertical conductors will need to be installed. However, the effective length will need to
be determined for the different soil types as, after a specific threshold, longer conductors
no longer help in decreasing the impedance.

Based on these results, the self impedance Z22 (transformer point) was adjusted for ver-
tical conductors of 9 metres in lengths for the soil of 100 Ωm and 30 metres for the soil
of 1000 Ωm. The result of this is presented in Figure 6.19.
6. Transient Simulations - Results 75

25 80
Standar Standar
Extended 70 Extended
Optimized Optimized
20
60
Impedance [Ω]

Impedance [Ω]
50
15

40

10
30

20
5
10

0 0
102 103 104 105 106 102 103 104 105 106
Frequency [Hz] Frequency [Hz]

a) Self impedance Z22 - soil resistivity 100 Ωm b) Self impedance Z22 - soil resistivity 1000 Ωm

Figure 6.19: Transformer point impedance - for additional vertical conductors with effective length in the
grounding grid

As shown in Figure 6.19, the impedance has decreased significantly, especially for case
b). This shows that for soils of higher soil resistivity, longer vertical conductors need
to be placed compared with soils of low resistivity. The grounding grid can be further
optimized by placing different kinds of conductors or by changing the diameter of the
conductors.

By looking at Figure 6.14, it can be observed that, when the grounding grid is modelled
as FD impedances in a soil of 100 Ωm the generated overvoltages are very close to when
modelled as a simple resistor at power frequency, while at 1000 Ωm they are considerably
higher. Therefore, optimizing a grounding grid at 100 Ωm will not decrease the overvolt-
ages seen at the transformer terminal, thus this will be done only for the case of 1000
Ωm. With the newly acquired impedances, the model is updated and new calculations
have been performed. The result of the voltages at the power transformer terminals is
shown in Figure 6.20.
6. Transient Simulations - Results 76

105
14 106
1.26 Normal
12 Optimized
1.24
1.22
10

Overvoltages at Transformer [V]


1.2

8 1.18
1.84 1.86 1.88
10-6
6

-2

-4
0 2.5 5 7.5 10 12.5 15

Figure 6.20: Voltage decrease in the transformer terminal, by optimizing the grounding grid

In Figure 6.20, the overvoltages at the transformer terminal are compared, for the up-
dated and the previous state of the grounding grid. It can be seen that the overvoltages
have been reduced by approximately 7kV.
7
Conclusions
In this thesis, electromagnetic transient studies were performed within a broad frequency
range by taking into account the impedances of grounding structures in the simulations.
So far, grounding grids have mostly been modelled by a simple resistance which is not
entirely correct as it does not take into account the behaviour of grounding systems.
Grounding systems behave differently at various frequency ranges and are affected by
both the geometry of the grounding structure and the characteristics of the surround-
ing soil. At low frequencies, it was observed that the behaviour of grounding systems is
purely resistive, while for higher frequencies it is:

• inductive when the ground system is buried in a high conductive soil

• capacitive when the ground system is buried in a poor conductive soil

The importance of modelling grounding systems and including them in the simulations
has been thoroughly explained during this thesis. One important observation is that
the generated overvoltages are proportional to the soil resistivity. For highly conduct-
ing soils, the overvoltages are much lower compared to those of poor conductive soils.
However, in the latter, soil ionization and frequency dependency of the soil have a more
significant role and act in favour of decreasing the generated overvoltages to some ex-
tent.

Lightning is a high frequency phenomenon and therefore for transient simulations, all
models should be frequency dependent in order to obtain more accurate results. To
create frequency dependent models for the grounding structures studied in this thesis,
two methods were used on the output of XGSLab: the VF method and the ARMAFIT
method, the former providing results with greater precision and the latter being easier
and quicker to implement.

Using the above-mentioned methods, FD models were developed for: 1) a vertical con-
ductor, 2) a ring conductor, 3) footing of a Wintrack tower and 4) to calculate the relevant

77
7. Conclusions 78

impedances in a substation grounding grid. To calculate the impedances at the substa-


tion, a network approach of 2-ports is used, as XGSLab cannot directly calculate cou-
pling impedance. By using these FD models, the importance of creating high frequency
models for transient studies was initially proved. Further to that, the optimal position to
install a surge arrester was determined. This position was found to be the one closest to
the power transformer as it greatly helps in reducing voltages from reflections. Further-
more, using the same approach previously mentioned in Chapter 5, the optimal position
to install a second surge arrester can also be determined.

Finally, in order to reduce the overvoltages at the transformer’s terminal by optimising


the grounding grid, two vertical conductors were placed near the transformer’s ground-
ing point. This resulted in a significant reduction in the self impedance (Z22). Nonethe-
less, it was discovered that for higher soil resistivities, longer vertical conductors are
needed, so that the impedance and subsequently the overvoltages will decrease at the
transformer’s terminal.

Recommendations for future work


This thesis demonstrated that the broad frequency behaviour of grounding grids can
be affected by many different parameters. Moreover, it was proved that when perform-
ing transient simulations, the grounding system is a variant that should not be ignored.
However, there are still many things that could be further explored. Some of them are:

1. Elaborate in more detail on the effect of soil ionization. During this thesis, this was
the only effect that was not elaborated in detail, as it is not directly modelled in the
available software tools and it was not an important effect to consider since only
high conductive soils were investigated [32].

2. Model more grounding systems and have an available database in order to per-
form accurate studies more quickly when needed. This can be difficult to achieve
for a grounding grid of a substation, since the position where the surge arresters
and transformers are grounded will most likely be different for different substa-
tions. During this thesis, a 2-port network was considered to determine the opti-
mal place to install a surge arrester. This technique can also be easily extended to
a 3-port system in order to determine the optimal placement for a second surge
arrester.

3. Inject different lightning wave shapes to view the different generated overvoltages
in the system.
References
[1] M. Popov, L. Grcev, H. Kr. Høidalen, B. Gustavsen, and V. Terzija. Investigation of
the overvoltage and fast transient phenomena on transformer terminals by taking
into account the grounding effects. IEEE Transactions on Industry Applications Vol.
51, no. 6, November/December 2015.

[2] J. D. Schoene. Direct and nearby lightning strike interaction with test power distri-
bution lines. University of Florida, PhD dissertation, 2007.

[3] V. Cooray, M. Fernando, S. Visacro, and L. Grcev. Lightning protection. IET Power
and Energy Series 58, 2010.

[4] Y. Baba and V. A. Rakov. Electromagnetic computation methods for lightning surge
protection studies. IEEE Press, Vol. 116, 2016.

[5] R. H. Golde. Lightning. vol 1, 2 London, U.K: Academic, 1977.

[6] K. Elrodesly. Comparison between heidler function and the pulse function for mod-
eling the lightning return-stroke current. Ryerson University, January 2010.

[7] F. Heidler. Analytische blitzstromfunktion zur lemp-berechnung. Proc. 18th Int.


Conf. Lightn. Protec., Munich, Germany pp. 63-66, September 1985.

[8] C. A. Nucci, F. Rachidi, M. V. Ianoz, and C. Mazzetti. Lightning-induced voltages on


overhead lines. IEEE Trans. Electromagn. Compat. Vol. 35 no. 1, pp 75-86, February
1983.

[9] A. De Conti and S. Visacro. Analytical representation of single- and double-peaked


lightning current waveforms. IEEE Transactions on Electromagnetic Compatibility,
2007.

[10] Technical Brochure 063. Guide to procedures for estimating the lightning perfor-
mance of transmission lines. Cigre, 1991.

[11] IEC 62305-1. Protection against lightning - part 1: General principles. Nederlands
Elekrotechnisch Comite, 2010.

[12] L. Grcev and M. Popov. On high frequency circuit equivalents of a vertical ground
rod. IEEE Transactions on Power Delivery, Vol. 20, no. 2, pp.1598-1603, April 2005.

[13] A. Manunza. Frequency-dependent equivalent circuit of grounding grids for


electro-magnetic transient simulations. International Journal of Electrical Power
& Energy Systems, Vol. 116, 2020.

79
REFERENCES 80

[14] D. S Gazzana, A. S. Bretas, G. A. D. Dias, and M. Tello. Comparative analysis of emc


methodologies applied on transients studies of impulsive grounding systems. IEEE
International Conference, January 2010.

[15] I. F. Gonos. Experimental study of transient behaviour of grounding grids using


scale models. Measurement Science and Technology, Vol. 17, 2006.

[16] A. Ek Mghairbi. Assessment of earthing systems and enchantment of their perfor-


mance. Cardiff University, 2012.

[17] M. I. Lorentzou, N. D. Hatziargyriou, and B. C. Papadias. Time domain analysis of


grounding electrodes impulse response. IEEE Transactions on Power Delivery, Vol.
18 no. 2, 2003.

[18] V. Steinsland, L. H. Sivertsen, E. Cimpan, and S. Zhang. A new approach to include


complex grounding system in lightning transient studies and emi evaluations. 6th
IEEE International Conference on High Voltage Engineering and Application, Vol. 17,
2019.

[19] D. S. Gassana, A. S. Bretas, G. A. D. Dias, and M. Tello. Transient response of ground-


ing electrode with emphasis on the transmission line modeling method (tlm). 30th
International Conference on Lightning Protection, 2010.

[20] Y. Liu. Transient response of grounding systems caused by lightning modelling and
experiments, phd thesis. Acta Universitatis Upsaliensis, 2004.

[21] A. P. Meliopoulos and M. G. Moharam. Transient analysis of grounding systems.


IEEE Trans. Power Apparatus and Systems, PAS-102, no. 2, 1983.

[22] V. T. Kontoargyri, I. F. Gonos, F. V. Topalis, and I. A. Stathopoulos. Transient be-


haviour of a horizontal grounding grid under impulse current. National Technical
University of Athens, Vol. 14, no. 3, 2003.

[23] A. F. Otero, J. Cidras, and J. L. Alamo. Frequency-dependent grounding system cal-


culation by means of a conventional nodal analysis technique. IEEE Trans. Power
Apparatus and Systems, Vol. 14, no. 3, 1999.

[24] L. Grcev. Computation of grounding system tramient impedance, ph.d. disserta-


tion,. University of Zagreb, Croatia, 1986.

[25] A. Bayadi T. Rouibah, N. Harid, and K. Kerroum. Transient electromagnetic field


radiated by grounding systems caused by lightning strike in a dissipative half-space.
Cogent Engineering, 2017.

[26] L. Grcev and V. Arnautovski. Grounding systems modeling for high frequencies and
transients: Some fundamental considerations. IEEE Bologna PowerTech Confer-
ence, 2003.

[27] L.Grcev. Computer analysis of transient voltages in large grounding systems. IEEE
Transactions on Power Delivery, Vol. 11, 1996.
REFERENCES 81

[28] F. Dawalibi. Electromagnetic fields generated by overhead and buried short con-
ductors part 2 - ground networks. IEEE Transactions on Power Delivery, Vol. PWRD-
1, no. 4, 1986.

[29] J. Meppelink, R. Andolfato, and D. Cuccarollo. Calculation of lightning effects in the


frequency domain with a program based on hybrid methods. International Collo-
quium on Lightning and Power Systems, 2016.

[30] D. Cavka, N. Mora, and F. Rachidi. A comparison of frequency-dependent soil mod-


els: Application to the analysis of grounding systems. IEEE Transactions on Electro-
magnetic Compatibility, Vol. 56, pp 177-187, 2014.

[31] KEMA Nederland B.V. Guide for transmission line grounding: A roadmap for design,
testing, and remediation:part i—theory book. EPRI, Palo Alto, CA: 2007. 1013900,
December 2007.

[32] S. Visacro, W. Chisholm, and M. Haddad. Impact of soil-parameter frequency de-


pendence on the response of grounding electrodes and on the lightning perfor-
mance of electrical systems. Cigre, Vol. 56, pp 177-187, 2019.

[33] E.P. Nicolopoulou, F.E. Asimakopoulou, I.F. Gonos, and I.A. Stathopulos. Compar-
ison of equivalent circuit models for the simulation of soil ionization. Elsevier Vol.
113 no. 8, pp 180-187, 2014.

[34] B. Gustavsen and A. Semlyen. Rational approximation of frequency domain re-


sponses by vector fitting. IEEE Transaction on Power Delivery, vol. 14, no. 3, pp.
1052-1061, July 1999.

[35] B. Gustavsen. Improving the pole relocating properties of vector fitting. IEEE Trans-
action on Power Delivery, Vol. 21, no. 3, pp. 1587-1592, July 2006.

[36] T. Dhaene D. Deschrijver, M. Mrozowski and D. De Zutter. Macromodeling of mul-


tiport systems using a fast implementation of the vector fitting method. IEEE Mi-
crowave and Wireless Components Letters, Vol. 18, no. 6, pp. 383-385, June 2008.

[37] Gustaven B. Matrix fitting toolbox: User’s guide and reference. Sintef Energy Re-
search, 2009.

[38] T. Noda and N. Nagaoka. Development of arma models for a transient calculation
using linearized least-squares method. Trans. IEE of Japan, Vol. 114-B, no. 4, pp.
396-402, 1994.

[39] K. Velitsikakis. Substation insulation coordination 150 & 110 kv study for the per-
formance against lightning overvoltages. TenneT TSO B.V., March 2018.

[40] S. Papenheim, P. Malicki, N. Pfeifer, and M. Kizilcay. Lightning overvoltage analy-


sis of a mixed hvdc transmission line with consideration of grounding system and
different sheath-bonding types. University of Siegen, Germany, 2019.
REFERENCES 82

[41] H. B. Gatland. Electronic engineering applications of two-port networks. Elsevier


Ltd, Pergamon Press, 1976.

[42] V. Vita, C. A. Christodoulou, and L. Ekonomou. Lightning performance of medium


voltage distribution networks. Springer Nature Switzerland AG, 2020.

[43] IEC. Npr-iec/tr 60071-4. Nederlands Elektrotechnisch Comité (NEC), July 2014.

[44] C.S Engelbrecht, P.L.J. Hesen, C.S. Suurman, I. Tannemaat, and E.P Evertz. Nota
isolatiecoordinatie voor 380, 220, 150, 110 en 50 kv elektriciteitsnetten in nederland.
DNV Kema Energy & Sustainability, July 2013.

[45] L. Rouco, X. M. L.-Fendandez, C. A.Mario, and H. Gago. Fast front transients in


transformer connected to gas insulated substations: (white+black) box models and
tdsf monitoring. e-ARWtr2016 transformers, May 2016.

[46] R. Alipio and S. Visacro. Impulse efficiency of grounding electrodes: Effect of


frequency-dependent soil parameters. IEEE Transactions on Power Delivery, Vol.
29, no. 2, April 2014.

[47] J. H. Scott, R. D. Carroll, and M. HADDAD. Dielectric constant and electrical con-
ductivity of moist rock from laboratory measurements. Sensor and Simulation Note,
Vol. 116, August 1964.

[48] J. H. Scott, R. D. Carroll, and M. Haddad. Electrical and magnetic properties of rock
and soil,” theoretical notes. Geological Survey, no. 18, 1966.

[49] F. H. Silveira, S. Visacro, H. B. Ferreira, and A. De Conti. Lightning-induced volt-


ages calculated with the hybrid electromagnetic model considering frequency-
dependent soil parameters. International Symposium on Lightning Protection, Vol.
7-11, 2013.
Appendix A - Models of Frequency
Dependency of Soil

Frequency dependence results by Smith Rose


The first scientist to discover the soil dependence with the frequency was Smith-Rose. In
1933 he presented his work, where he measured the conductivity and permittivity, using
frequencies in the range of 100kHz-10MHz for different soil samples, which were taken
from different sites within the United Kingdom [32].

The basis of this method is to measure the capacitance and resistance, for the collected
soil samples. Afterwards, from the values measured and by considering the capacitance
of the place the soil is stored with air as its dielectric, the effective conductivity and di-
electric constant for all different soil samples were estimated. What was observed in the
results was, that when the frequency was increasing, the soil conductivity would increase
as well. The increase was more noticeable in soils with less moisture content. Further-
more, some of the samples were tested using low frequencies down to the range of 50 Hz
and the results showed that the conductivity would still decrease with a decrease in the
frequency. Finally, according to the author there were some limitations to the method,
but the magnitude of the soil parameters could be correctly predicted [32].

Soil frequency dependency results by Scott et al.


Afterwards, in 1964 Scott et al. from laboratory and measurements done on the field,
using frequencies from 100 Hz-1 MHz for a different range of soil samples with different
water content. Using the results obtained, Scott made plots for the permittivity, conduc-
tivity and permeability which were dependent on frequency and the percentage of the
water content [30], [47].
Furthermore, based on previous measurements along with some additional measure-
ments and by using math techniques, such as: curve-fitting techniques, equations for
the estimation of the conductivity and the relative permittivity were developed, which
can be seen below [48].

σ( f , σ0 ) = 10K (7.1)

2
K = 0.028 + 1.098l og 10 (σ0 ) − 0.068l og 10 ( f ) + 0.036l og 10 (σ0 )
(7.2)
2
−0.046l og 10 ( f )l og 10 (σ0 ) + 0.018l og 10 (f )

²r ( f , σ0 ) = 10D (7.3)

83
2
D = 5.491 + 0.946l og 10 (σ0 ) − 1.097l og 10 ( f ) + 0.069l og 10 (σ0 )
(7.4)
2
−0.114l og 10 ( f )l og 10 (σ0 ) + 0.067l og 10 (f )
Where:

• f is the frequency in Hz

• σ0 is the conductivity at 100 Hz in mS/m

Finally, Scott et al. observed that, there was some statistical dispersion in his results for
the frequency dependence of the soil parameters. However, there was no proposal on
how to correct it.

Soil frequency dependency results by Longmire and Smith


The first semi-theoretical model which could describe the frequency dependence of soil
was proposed by Longmire and Smith. Their result was based on the experimental re-
sults obtained from Scott et al. Moreover, Smith and Longmire presented and expanded
the curve-fit model for frequencies up to 1012 Hz.
Their equations were derived from an RC model for the soil, where the results were for-
mulated as function of frequency and water content [32].

10
ω/β 2
σ(ω) = σ0 + ²0 c i βi ( i ) 2
X
(7.5)
1+(ω/βi )
i =1

10
ci
²r (ω) = ²+ +
X
(7.6)
1+(ω/βi )2
i =1

Where,

• σ is the soil conductivity

• σ0 the soil conductivity at low frequency

• ²r is the relative permittivity of the soil

• ω is the frequency multiplied with 2π

• ²+ = is the high frequency limit for the dielectric constant,which is equal to 5

Finally, the coefficient values of c n are seen in Table 7.1 and βi = (w/10)1.28 10i −1 , with w
be the water content of the soil [30].

84
Table 7.1: c i coefficients

i cn
1 3.4 · 106
2 2.74 · 105
3 2.58 · 104
4 3.38 · 103
5 5.26 · 102
6 1.33 · 102
7 27.2
8 12.5
9 4.8
10 2.17

Soil frequency dependency results by Messier


Another representation which was also based on Scott’s work was done by Messier [30].
The expressions that were proposed by Messier for the soil conductivity and permittivity
are seen in equations 7.7 and 7.8 respectively. Where σ0 is the soil conductivity at low
frequency, ²+ is the permittivity at higher frequencies and ω is the angular frequency.
³ q ´
σ(ω) = σ0 · 1 + 2ω²
σ0
+
(7.7)
q
²+
³ ´
2σ0
²r (ω) = ²0 · 1 + ω² +
(7.8)

Soil frequency dependency results by Visacro and Portela


More recently, in 1987, Visacro and Portela presented their results for the frequency
dependence of the soil. Their experiment was completed in three different types of
soil, using frequencies from 40 Hz to 2 MHz. Finally, from the value of the measured
impedance, the soil resistivity and permittivity were calculated for each different fre-
quency and an empiric formula for both the soil resistivity and permittivity which de-
pends on frequency can be seen on equations 7.9 and 7.10.
³ ´−0.535
²r ( f ) = 2.34 · 106 1
σ0 · f −0.597 (7.9)

f 0.072
³ ´
σ( f ) = σ0 100 (7.10)

Soil frequency dependency results by Visacro and Alipio


Visacro afterwards did more research alone and in cooperation with others about how to
determine the frequency dependence of the soil. However, only the most recent one is
mentioned here, which is the one made in cooperation with Alipio in 2011. The biggest
advantage of their proposal in comparison to the previous ones mentioned is, that it was

85
carried out on the field. Based on the results obtained, Visacro and Alipio developed
expressions to calculate the soil permittivity and the soil conductivity as seen in equa-
tions 7.11 and 7.12 respectively. These equations can predict the soil dependency in the
frequency range of 100Hz to 4 MHz. However, the permittivity does not yield accurate
results for frequencies less than 10 kHz [30], [32], [49].

²r ( f ) = 7.6 · 103 · f −0.4 + 1.3 f ≥ 10k H z (7.11)


³ h ³ ´0.73 ih i0.65 ´
σ( f ) = σ0 · 1 + 1.2 · 10−6 · σ10 f − 100 (7.12)

Comparison of the previously discussed methods


In this section, a comparison of the previously discussed methods is given. To make
the comparison, plots of the resistivity which is equal to ρ = 1/σ and the permittivity ²r
for different low frequency soil resistivities in a frequency range of 100 Hz to 4 MHz are
given. The results for soil a resitivity of 300 Ωm, 1 kΩm and 3 kΩm in Figures 7.1, 7.2
and 7.3.

(a) Soil resistivity-models comparison for ρ 0 = 300 Ωm (b) Soil permittivity-models comparison for ρ 0 = 300 Ωm

Figure 7.1: Plots of resistivity and permittivity, where low frequency resistivity is 300 Ωm. Adapted from [32]

(a) Soil resistivity-models comparison for ρ 0 = 1000 Ωm (b) Soil permittivity-models comparison for ρ 0 = 1000 Ωm

Figure 7.2: Plots of resistivity and permittivity, where low frequency resistivity is 1000 Ωm. Adapted from [32]

86
(a) Soil resistivity-models comparison for ρ 0 = 3000 Ωm (b) Soil permittivity-models comparison for ρ 0 = 3000 Ωm

Figure 7.3: Plots of resistivity and permittivity, where low frequency resistivity is 3000 Ωm. Adapted from [32]

The resistivity and permittivity were expressed as functions of the low-frequency resis-
tivity and the permittivity of vacuum respectively. From the figures it is clear that, there
is a strong variation of the resistivity and the permittivity when the frequency changes.
Both parameters are decreasing significantly by increasing the frequency, regardless of
the low-frequency resistivity value. However, higher variations between the models are
found in the range of high frequencies for low-resistivity soils [30]. Finally, it is also be
observed that, all the models predict overall similar responses, expect from the model
made by Visacro-Portela which differs noticeably.

87
Appendix B - ARMAFIT
Instructions
ARMAFIT: instructions
Armafit is a program that allows to obtain the coefficient of rational function in s of an
admittance Y = I/V. This method is useful since the result can be inputted on the ATP-
draw model called ’KIZILCAY F-DEPENDENT branch’.

Step 1 - Making an AFT file


In order to produce the punch file that will be used in the Kizilcay model, you have to
give as input the following:

• Frequency samples (not in logarithmic scale)

• Impedance Amplitude or Admittance Amplitude in [Ω]

• Impedance Angle or Admittance Angle in [Deg]

This input data can be either provided from software such as XGSLab or by doing a fre-
quency scan on the network you are interested in ATPDraw.

This data along with a header must be written in a txt file which its extention name
should be changed to AFT afterwards. An example along with the header is given bel-
low

KIZILCAY F-DEPENDENT { keyword for the KIZILCAY F-DEPENDENT use


C – fitting parameters ——————————————————-
S { ’S’: Laplace s-rational model, ’Z’: z-rational (ARMA) model (*1)
0.35 -1. { admittance values at freq = 0 and infinity (*2)
1 11 { min and max orders (*3)
2. 0 { permitted error in %, and max number of iteration (*4)
151 1. { number of samples, and pre-defined gain (*5)
C – frequency response ——————————————————-
C frequency [Hz], amplitude [S], phase angle [deg]
0.10000E+02 0.24553E+00 -0.51240E+02
0.10471E+02 0.23873E+00 -0.52502E+02
0.10965E+02 0.23185E+00 -0.53746E+02
0.11482E+02 0.22492E+00 -0.54970E+02 (Samples should be 151)

Step 2- Generate the punch file

88
• Place the AFT file you want to fit on the same folder as ARMAFIT

• Open the cmd and use the command ’cd’ to enter the folder you have your file and
ARMAFIT

• Type on the cmd window "armafit filename.AFT"

• Include the .pch file on a Kizilcay model

Step 3- Ploting result


The user should know that the Kizilcay model is an admittance. Therefore, should you
connect it to a voltage source and you plot the current, then you will find:

• The admittance, if you used ARMAFIT with the admittance amplitude and angle
as an input

• The impedance, if you used ARMAFIT with the impedance amplitude and angle as
an input

However, it is suggested that you use the admittance as input, since you will need to
invert the result every time and in bigger networks result in problems.

89
Appendix C - Vector Fitting Values
Below are the values of the circuits in Chapter 5 as calculated from VF

Vertical Conductor - Case A

Table 7.2: Values of the circuit of the vertical conductor - Case A

Values Node R [Ω ] L [H] C [F]


A 9.9185 · 101 1.9035 · 10−4 -
B 1.0481 · 102 3.5604 · 10−3 -
C 1.7474 · 101 6.0876 · 10−3 -
D 2.2335 · 104 1.0224 · 102 -
E −3.4831 · 102 −6.0738 · 10−1 -
F - - 2 · 10−10
G 5 · 105 - -

Vertical Conductor - Case B

Table 7.3: Values of the circuit of the vertical conductor - Case B

Values Node R [Ω ] L [H] C [F]


A −8.5838 · 101 −5.2096 · 100 -
B −1.8721 · 103 1.2404 · 10−1 -
C 1.1830 · 102 2.1472 · 100 -
D −3.5435 · 101 −1.6346 · 10−1 -
E 3.4051 · 101 1.3033 · 10−1 -
F −1.6730 · 101 −1.7960 · 10−2 -
G 1.7052 · 101 1.5920 · 10−2 -
H 2.3022 · 101 8.0491 · 10−3 -
I 4.5497 · 101 6.8059 · 10−3 -
J −3.9430 · 101 −3.2051 · 10−3 -
K 5.2080 · 101 2.0475 · 10−3 -
L −6.5161 · 101 −3.9804 · 10−4 -
M 3.7073 · 101 5.8955 · 10−5 -
N - - 2 · 10−10
O 5 · 105 - -
A1 −8.5838 · 101 - -
A2 - - −1.5794 · 10−1
B1 2.7475 · 103 - -
B2 - - 1.3212 · 10−5

90
Appendix D - Kizilcay Models
Vertical Conductor - Case A

KIZILCAY F-DEPENDENT 4 1.00000E+000 S


1.000000000000000000E+0001.000000000000000000E+000
1.362421169414656200E+0001.818054732201670700E+001
6.249404897045473900E-0068.611163729011517300E-005
6.953288658577087400E-0132.944363087594944900E-011
7.142360135156163500E-0217.647424126471417600E-019

Vertical Conductor - Case B

KIZILCAY F-DEPENDENT 9 1.00000E+000 S


1.350000000000000000E+0011.000000000000000000E+000
1.008648586963731000E+0021.350216044978634300E+003
1.771187367568134000E-0022.360629491712762700E-001
2.615121741228613400E-0083.980809003097076000E-007
1.222980451783392300E-0142.412233703815065100E-013
1.905649248602318300E-0215.763844415530122100E-020
1.725290523730636200E-0285.834959747753746400E-027
9.089183269712191400E-0363.993508988432726400E-034
2.448208703707579000E-0431.213901585133038500E-041
3.492792250599407900E-0512.274292075003583900E-049
Ring Conductor - Case A

KIZILCAY F-DEPENDENT 6 1.00000E+000 S


2.500000000000000000E+0001.000000000000000000E+000
2.076925987051899400E+0016.924388806481724100E+001
3.548972740330867400E-0052.826832467325552200E-004
1.225209900970284600E-0111.885753325895342400E-010
8.712166950724867900E-0192.790901751109595600E-017
9.412088942588518800E-0278.006585340064774400E-025
2.746767448740478700E-0391.005983871984653800E-033
Ring Conductor - Case B

KIZILCAY F-DEPENDENT 9 1.00000E+000 S


3.499999999999999800E-0011.000000000000000000E+000
1.061522666189714300E-0053.860229156038621400E-005
1.427354359460134300E-0111.287070293807923600E-010
3.698171279706770500E-0186.546616551156517400E-017

91
2.417583135570145800E-0257.871296711154099600E-024
7.336738002961983700E-0332.870388258472351400E-031
2.794676849652340500E-0409.715710016186888800E-039
3.824468284484517300E-0482.564516319602538100E-046
3.656795636691971000E-0561.286282561343343900E-054
4.055764261812632400E-0642.943892684025963000E-062
High Voltage Tower

KIZILCAY F-DEPENDENT 6 1.00000E+000 S


1.000000000000000000E+0001.000000000000000000E+000
4.183405391002558300E-0067.458373213939631200E-006
2.251064409316320400E-0121.050758000276657300E-011
1.941298093195391300E-0192.618222852614562900E-018
3.548065222169458500E-0271.060924839419587400E-025
4.997250917452424400E-0358.876056780801268700E-034
5.745828005344821600E-0432.220832043599549900E-041

92
Appendix E - Heidler Function as
Defined in MODELS
Heidler function modelled on MODELS of EMTP based on Equation (2.1).

MODEL HEIDLER

DATA Ip, k, t1, t2

OUTPUT i

VAR i

−− HEIDLER Function – Equation (2.1) at Section 2.3.1


FUNCTION HEIDLER(time,Ip,k,t1,t2):=((Ip/(exp(-(t1/t2)*((k*t2/t1))
**(1/k))))*exp(-(time/t2))*((time/t1)**(k))/(1+((time/t1))**(k)))

EXEC
i := heidler(t,Ip,k,t1,t2)

ENDEXEC
ENDMODEL

93
Appendix-F
Authors: I. Tannemaat + M. Kouros + K. Velitsikakis

Double peaked waveshape

Abstract — To perform lightning performance calculations on power systems an accurate


representation of the lightning waveshape is required. Over the years many different
waveshapes have been proposed. A good approximation was given by Heidler, as this function
showed continuous time derivatives. However, in order to represent subsequent strokes, at
least two functions of the same type are needed, which results in a waveshape that loses it
concavity at the front. Two waveshapes have been proposed that do not suffer of these
limitations, being the Cigre and the double peaked waveshape. Currently, only the CIGRE
waveshape is standard available in the ATP-EMTP software. In this paper/presentation a way to
apply the double peaked waveshape in ATP-EMTP is described using the MODELS function.
This is done with the summation of 7 Heidler functions, each with 4 different parameters per
function. However, this approach results in a Double Peaked waveshape with constant
characteristics which the user cannot directly edit. For adjustments the user will need to
manually change the parameters of the Heidler function and that may result in a very different
waveshape. Therefore, a routine was developed that takes the normalized characteristics of the
double peaked waveshape, using an iterative process to fit the output to the desired
characteristics.

Introduction
For the representation of double peaked waveshapes it is proposed to use the sum of 7 Heidler-
functions with different characteristics, and fit the result to the waveshape acquired by the
measurements in [1]. To create a progressive build up four waveforms with increasingly higher front
times and gradually steeper rates of rise are applied. Then the maximum steepness (Sm) and the first
peak value (Ipk1) can be approximated by adding a fifth function. The tail time (T50) and the second
peak (Ipk2) are determined by adding the sixth and seventh function. This gradual build-up is
graphically presented in Figure 1. For practical reasons all graphs are presented with positive
waveshapes.

50
1 to 4
1 to 5
40
1 to 6
Current [pu]

1 to 7
30

20

10

0
0 10 20 30 40 50 60
Time [us]
Figure 1 – Double peaked waveshape build-up from 7 different Heidler functions (dataset MCS)

The expression for the Heidler-function is shown in equation (1).

𝑖(𝑡) = ∑𝑚 𝑛𝑘 𝑛𝑘
𝑘=1(𝐼0𝑘 /𝜂𝑘 )𝑒𝑥𝑝(−𝑡/𝜏2𝑘 ){(𝑡/𝜏1𝑘 ) /[1 + (𝑡/𝜏1𝑘 ) ]} (1)

With

𝜂𝑘 = 𝑒𝑥𝑝[−(𝜏1𝑘 /𝜏2𝑘 )(𝑛𝑘 𝑡2𝑘 /𝑡1𝑘 )1/𝑛𝑘 ] (2)

Where
𝑛𝑘 = exponent controlling the steepness
𝑡 = time in µs
𝐼0𝑘 = controls the amplitude, value in kA
𝜏1𝑘 = front-time constant, µs
𝜏2𝑘 = decay-time constant, µs

The waveshape is characterized by several parameters, as observed in Figure 2. The parameters for
the amplitudes of the first peak is, Ipk1, and for the second peak is, Ipk2, both in kA. T10 and T30
correspond to the time necessary for the current to rise from 10% and 30% to 90% of the first peak
current value, both in µs. The corresponding related steepnesses are characterized by S10 and S30
respectively, both in kA/µs. Sm represents the maximum steepness of the front wave in kA/µs. Finally
the parameter that describes the time necessary for the current to decay to 50% of the second peak is
given by T50, in µs. The last mentioned parameter is not presented in the figure.

50
Ip2
Ip1
40
0.9Ip1

30
Current [kA]

S30
S10
20
0.3Ip1
10 Sm
0.1Ip1
T30
0
T10
0 2 4 6 8 10 12 14 16 18 20 22 24
Time [µs]
Figure 2 – Double peaked waveshape and its main characteristics (dataset MCS)

Waveshape characteristics
The values that shall be applied in the Heidler-functions are given per function for two different
datasets from literature [1], and result in the waveshapes presented in Figure 3. For each dataset a
different amplitude is applicable, but to be able to compare the waveshapes they are presented in per
unit.
1

0,75
Current [pu]

0,5

0,25
MCS dataset
MSS dataset
0
0 10 20 30 40 50 60
Time [us]
Figure 3 – Double peaked waveshape and its main characteristics (dataset MCS)

Table 1 – Parameters used to synthesize first negative stroke currents


MCS dataset MSS dataset
I0k nk τ1k τ2k I0k nk τ1k τ2k
k (kA) (µs) (µs) (kA) (µs) (µs)
1 6 2 3 76 3 2 3 76
2 5 3 3.5 10 4.5 3 3.5 25
3 5 5 4.8 30 3 5 5.2 20
4 8 9 6 26 3.8 7 6 60
5 16.5 30 7 23.2 13.6 44 6.6 60
6 17 2 70 200 11 2 100 600
7 12 14 12 26 5.7 15 11.7 48.5

Scaling the waveshape for study purpose


A method to scale the waveshape is presented in paper [2], this enables the user to apply some
variations of the double peaked waveshape in simulations and perform detailed studies with different
lightning current amplitudes and waveshapes. The presented approach is however limited, as also
mentioned by the authors.

Modelling the waveshape


To enable the user to have higher flexibility in the application of the double peaked waveshape an
ATP-EMTP routine was written in MODELS. The user is able to enter some specific data for the
waveshape, and the routine then uses a sub-model to auto-fit the waveshape to hit the desired values
of the peaks in an iterative way. It should be noted that for practical reasons the user is not able to
enter the main characteristics of the waveshape (as described in figure 2) but other input parameters
defining the waveshape, as mentioned below.

Input parameters
The following input parameters need to be defined as input for the model to work. The values for Ipk2
and Tpk2 are mandatory and shall always be provided by the user. For Ipk1, Tpk1 and Sm also a value of
'-1' can be applied. Then the routine will use automatically defined values based on the main
waveshape. For practical reasons the user is able to enter a factor for T 50 (e.g. from 0 to 9999) to
adjust the tail time.

Ipk1 (A) Tpk1 (s)


Ipk2 (A) Tpk2 (s)
Sm (kA/µs) T50 (factor)
Main waveshape
As presented in Figure 3 there are two waveshapes defined from lightning measurements. To be able
to auto-fit the waveshape the routine needs to be able to find two peaks. Since the waveshape for the
MCS dataset has more distinctive peaks, this waveshape was used as basis.

Implementation in ATPDraw
To implement the model in ATPDraw the user needs to include a default model. In this model place
the code under 'edit' and edit local definitions with currents in 'A' times in 's' and steepness in 'kA/µs'.
Then, connect the model with TACS (current source). To check the output a resistor can be connected
and the current can be measured.

Figure 4 – ATPDraw model

Description of routine
The routine to plot the waveshape based on the user input makes use of the following steps: some
initial parameters are defined in the initialization phase, then a subroutine produces the initial
waveshape and returns the relevant values that can be compared to the user defined parameters. The
parameters are changed iteratively to have a better fit. The final parameters are then used to plot the
waveshape. More detailed information is presented below.

Basic set up of a model and submodel


A model consists of two main parts. The initialization procedure (INIT) describes the initial point of
operation of the model before the simulation is started. The execution procedure (EXEC) describes
how the operation of the model is updated at each successive instant (timestep) of the simulation.

A submodel is a full model of its own, except for the fact that it is declared and used inside another
model. Once it is declared, it can be used in that model in as many use instances as required, each
one maintaining its own separate state of operation [3].

Table 2 – Model using locally-defined submodel


MODEL component -- defining a model
INIT
CONST ...
DATA ...
INPUT ...
VAR ...
OUTPUT ...
MODEL controller -- defining a local submodel
INIT
...
ENDINIT
EXEC
...
ENDEXEC
ENDMODEL
ENDINIT

EXEC
...
USE controller AS control_1 -- using the submodel
ENDUSE
...
ENDEXEC
ENDMODEL

Initial values
The waveshape is based on the unified standard waveshape for the MCS dataset. The initial values
are scaled based on the user parameters for the second peak, where the value for Ipk2 scales the
amplitude of the main waveshape and Tpk2 scales the waveshape along the time-axis. This takes place
in the INIT part on lines 168 to 174.

Subroutine to retrieve main parameters


To be able to fit the curve the deviation between the desired characteristic values and the actual
waveshape values needs to be determined multiple times. For this a subroutine is defined, which is
found on lines 24 to 111. The subroutine is called twice in the EXEC part, in lines 187 to 197 and lines
215 to 225.

This subroutine needs a separate definition of the Heidler function (line 40+41). Then it produces a
waveshape for the latest waveshape characteristics and stores the values in a dedicated matrix (line
50 to 56). Based on this calculated waveshape the time and amplitude of both peaks are looked up in
predefined time windows (lines 60 to 76). Finally all relevant times and values for steepness are
defined. It should be noted that for practical reasons more characteristic values are retrieved than
needed for the applied fitting.

Fitting of waveshape
A subroutine can only be called from the EXEC part of the code, however it is only needed for
initialisation. Therefore the fitting routine is placed in an IF-statement that enables this part of the
routine to run during the first time step. This can be done without any problems since the waveshape
always starts with a value equal to zero. Therefore the part of the code from lines 184 to 237 is
actually a quasi-INIT code.

The FOR loop in this part of the routine defines the number of iterations needed to get to an
acceptable result. It is now set to 5 times, which gives acceptable results, but can be also changed by
the user. In general the result will be more accurate if the number increases. The calculation time will
also increase a bit, but more important is that there is a memory limit that should not be exceeded for
the model to run. For every iteration step the subroutine is called two times. Every subroutine is then
performed under a unique identification number and uses a part of the work memory. For this the
subroutine requires a counter, which is identified in line 185 and applied for calling the subroutine in
lines 187 and 215.

To be able to change the parameters and have a fitted waveshape the subroutine is called before any
change is done. First the routine changes the parameters of Heidler-function 5 (lines 199 to 213),
which mainly defines the first peak amplitude, the maximum steepness and time of occurrence of the
first peak. Then Heidler-function 7 is changed (lines 227 to 234), which mainly defines the second
peak amplitude. This last step is necessary since every other change may influence the maximum
peak amplitude.

Plotting the waveshape


When the waveshape is fitted it can be plotted in the program. This is done in lines 240 to 248. This
part of the routine also takes a possible user defined delay for the start of the plot into account.

Check on input parameters


Several checks on the input parameters are performed. The code for this part of the routine can be
found in lines 138 to 166.
Results and comparison
Figure 5 shows the model output for the two datasets with parameters from Table 3, compared to the
waveshapes as mentioned in [1]. It can be seen that the shape for the MCS dataset has a perfect
match, since the model-waveshape is based on this dataset. Some differences can be seen for the
comparison with the MSS dataset, however the defined points match quite well.

Table 3 – Parameters used for comparison to reference dataset


Ipeak1 Ipeak2 T1pk T2pk T50 SM Tstart
MCS 40065 A 45284 A 8.25E-6 s 1.375E-5 s 1 20.13 kA/µs 0
MSS 27696 A 31000 A 7.80E-6 s 1.360E-5 s 2 23.84 kA/µs 0

Figure 5 – Comparison of model output with waveshapes from [1]


In the model the time and amplitude of the second peak can be changed as required, without almost
no limitations. Examples for variation of the amplitude of the current (up to 300kA) and time of the
peak (up to 3x) is presented in Figure 6 and Figure 7. Other values in these examples have been set
to '-1' for automatic determination.

The values for the time and amplitude of the first peak can be changed within some limits. For the
amplitude it is recommended to stay within a range of 95% to 105% and for the time a value of 90% to
110% is recommended. Figure 8 and Figure 9 show the waveshapes with application of these ranges.
In both cases it can be noted that the maximum steepness is kept to the initial steepness. This might
lead to a deformed waveshape, which can (partly) be corrected by also changing the values for the
maximum steepness.

To influence the decay of the tail the factor of T50 can be changed. Any value greater than 0 can be
used (e.g. 0.01 to 9999). Figure 10 shows the waveshape for values of 0.2, 1, 5, 50 and 500. Use of
this factor has an influence on the time and peak of the second peak, for which the routine corrects as
much as possible. Figure 11 zooms in on the second peak. It is noted that the amplitude remains
equal but the peak occurs sooner for small factors.

The maximum steepness can also be changed. For this it is recommended to stay within a value
range of 80% to 120%. Figure 12 (full waveshape) and Figure 13 (zoom) show the effect of a value of
80% and 150%. For the latter it is noted that the waveshape is slightly deformed. This can (partly) be
corrected by changing the time for the first peak.
Figure 6 – Variation of second peak amplitude Figure 7 – Variation of second peak time

Figure 8 – Variation of first peak amplitude Figure 9 – Variation of first peak time

Figure 10 – Variation of factor T50 Figure 11 – Zoom on influence of factor T50 on peak

Figure 12 – Variation of maximum steepness Figure 13 – Zoom on maximum steepness


References

[1] Analytical Representation of Single- and Double-Peaked Lightning Current Waveforms, A. De


Conti, S. Visacro, IEEE Transactions on Electromagnetic Compatibility, Vol. 49, No. 2, May
2007
[2] Adjustment of current waveform parameters for first lightning strokes: Representation by
Heidler functions, A.J. Oliveira, 2017 International Symposium on Lightning Protection (XIV
SIPDA)
[3] Users guide to models in ATP, Laurent Dubé, April 1996
Annex A - Code

1 MODEL Double_Peak
2
3 DATA Ipeak1
4 Ipeak2
5 T1pk
6 T2pk
7 T50
8 SMAX
9 Tstart
10
11 OUTPUT IOUT
12
13 VAR N,L,O,P,IOUT,temp1,temp2,delta,average,cnt,IMpk2,Tdiff1,Tdiff2,polarity
14 Ipeaks[1..2], times[1..3], FIT[1..7], I1[1..7], I2[1..7],
15 I3[1..7], IK[1..7], T1K[1..7], T2K[1..7], P[1..7],
16 PCurrents[1..4], Ptime[1..8], PcharI[1..2], PcharT[1..2], temp1
17 stp[1..2], difference[1..4], alfa, beta,
18
19
20 -- HEIDLER Function ------------------------------------------------------------
21 FUNCTION HEIDLER(a,x,c,n,t1,t2):=((c/(exp(-(t1/t2)*((n*t2/t1))
22 **(1/n))))*exp(-(a/t2))*((a/t1)**(n))/(1+((a/t1))**(n)))
23
24 --- Define submodel to get parameters of waveshape as often as user wants ------
25 MODEL GPARAM
26 DATA IKsub[1..7], T1Ksub[1..7], T2Ksub[1..7], Psub[1..7],
27 Timessub[1..3]
28 OUTPUT Currents[1..4], time[1..8], charI[1..2], charT[1..2],
29 CONST Length { val: 401 } -- limited length
30 VAR M, K, checka, nIpeak1,
31 nT1pk, pk2, pk3, Checkb[1..7],
32 charI[1..2], charT[1..2], I[1..7], pk1f3[1..7],
33 Currents[1..4], Difference[1..7],time[1..8], pk1,
34 pk1f[1..7], pk1f2[1..7], Windowpk[1..2],
35 Itotal1[1..Length], peaks_values[1..3],
36 Itotal2[1..Length], Itotal3[1..Length],
37 Ratio_t2, Usertimestep,
38
39 -- HEIDLER Function ----------------------------------------------------
40 FUNCTION HEIDLER(a,x,c,n,t1,t2):=((c/(exp(-(t1/t2)*((n*t2/t1))
41 **(1/n))))*exp(-(a/t2))*((a/t1)**(n))/(1+((a/t1))**(n)))
42
43 INIT
44 time[1..8] := [0 0 0 0 0 0 0 0]
45 Ratio_t2 := Timessub[2]*78431.3725
46 Usertimestep := 5E-8*Ratio_t2
47
48 ENDINIT
49 EXEC
50 FOR M:=1 TO Length-1 DO -- Call Heidler function for standard plot ----
51 FOR K:=1 TO 7 DO
52 I[K]:=HEIDLER(M*Usertimestep,K,IKsub[K],Psub[K],T1Ksub[K],T2Ksub[K])
53 ENDFOR
54 Itotal1[M] := I[1]+I[2]+I[3]+I[4]+I[5]+I[6]+I[7]
55 Itotal2[M] := I[1]+I[2]+I[3]+I[4]
56 ENDFOR
57
58 Windowpk[1] := trunc(Timessub[1]/Usertimestep)
59 Windowpk[2] := trunc(Timessub[2]/Usertimestep)
60 ------- Routine to find peaks and times ----------------------------
61 checka := 0
62 FOR M:= windowpk[1] - 20 TO windowpk[1]+20 DO -- search in given window
63 IF abs(Itotal1[M]) > checka THEN
64 checka := abs(Itotal1[M])
65 charI[1] := Itotal1[M]
66 charT[1] := M*Usertimestep
67 ENDIF
68 ENDFOR
69 checka := 0
70 FOR M:= windowpk[2] - 20 TO windowpk[2]+20 DO -- search in given window
71 IF abs(Itotal1[M]) > checka THEN
72 checka := abs(Itotal1[M])
73 charI[2] := Itotal1[M]
74 charT[2] := M*Usertimestep
75 ENDIF
76 ENDFOR ----- End of routine to find both peaks and times ----------
77
78 Currents[1..4]:=[0.1*CharI[1] 0.3*CharI[1] 0.9*CharI[1] 0.5*charI[2]]
79 Checkb[1..7] :=[9999 9999 9999 9999 0 0 0]
80
81 FOR M:=1 TO Length-2 DO
82 FOR N:=1 TO 4 DO
83 Difference[N] := abs(Currents[N] - Itotal1[M])
84 ENDFOR
85 Difference[5] := abs(Itotal1[M+1]-Itotal1[M])
86 Difference[6] := abs(Itotal2[M+1]-Itotal2[M])
87 IF Difference[1] < checkb[1] THEN
88 checkb[1] := difference[1]
89 time[1] := M*Usertimestep -- time for 10% value
90 ENDIF
91 IF Difference[2] < Checkb[2] THEN
92 checkb[2] := difference[2]
93 time[2] := M*Usertimestep -- time for 30% value
94 ENDIF
95 IF Difference[3] < checkb[3] and M < 165 THEN
96 Checkb[3] := difference[3]
97 time[3] := M*Usertimestep -- time for 90% value
98 ENDIF
99 IF Difference[5] > Checkb[5] and M < 165 THEN
100 checkb[5] := difference[5]
101 ENDIF
102 time[5] := Checkb[5] *0.02 / ratio_t2 -- maximum steepness total in A/us
103 IF Difference[6] > Checkb[6] THEN
104 checkb[6] := difference[6]
105 ENDIF
106 time[7] := Checkb[6] *0.02 / ratio_t2 -- maximum steepness 1 to 4 in kA/us
107 ENDFOR
108 time[6] := abs(1E-9*(Currents[2]-Currents[1])/(time[2]-time[1])) --steepness T10-T30 kA/us
109 ENDEXEC
110 ENDMODEL
111 --- End of submodel to get parameters of waveshape as often as user wants ------
112
113
114 INIT
115 stp[1..2] := [Tstart SMAX]
116 times[1..3] := [T1pk T2pk T50]
117 Ipeaks[1..2] := [abs(Ipeak1) abs(Ipeak2)]
118
119 ---------- set to standard values if not specified by user -------------------------------
120 IF Ipeak1 = -1 THEN -- standard value for Ipeak1 if uses gives -1 ------------
121 Ipeaks[1] := Ipeaks[2]*0.88521
122 ENDIF
123 IF Ipeak2 < 0 THEN -- check polarity of current waveshape -------------------
124 polarity :=-1
125 ELSE
126 polarity :=1
127 ENDIF
128 IF times[1] = -1 THEN -- standard value for time of first peak if uses gives: -1
129 times[1] := times[2]*0.59927
130 ENDIF
131 IF times[3] = -1 THEN -- set to standard value 1 if user gives -1 -------------
132 times[3] := 1
133 ENDIF
134 IF stp[1] = -1 THEN -- start at 0 if user gives -1 for Tstart ----------------
135 stp[1] := 0
136 ENDIF
137
138 ----------------Value of Second peak must always be greater than the first------
139 IF abs(Ipeaks[2]) <= abs(Ipeaks[1]) THEN
140 ERROR
141 write('The value of the second peak must be always greater than the first')
142 write('If you are having trouble specifying a correct value for Ipeak1, use -1 ')
143 STOP
144 ENDIF
145 ---------------- both peaks must be negative of positive -----------------------
146 IF Ipeaks[2]*Ipeaks[1] < 0 THEN
147 ERROR
148 write('Both peaks must be of the same polarity (positive or negative)')
149 write('If you are having trouble specifying current values use -1 for Ipeak1 ')
150 STOP
151 ENDIF
152 ----------------Value of Second peak must always be greater than the first------
153 IF abs(Ipeaks[2]) >= abs(Ipeaks[1]*1.25) THEN
154 ERROR
155 write('The maximum difference of current peak amplitudes is 25%')
156 write('If you are having trouble specifying a correct value for Ipeak1, use -1 ')
157 write('or change the limits manually in MODEL ')
158 STOP
159 ENDIF
160 --------------- T1pk<T2pk and times should be in us---------------------------
161 IF times[2]<=times[1] THEN
162 ERROR
163 write('Please check the time values of the peaks, time 2nd > time 1st')
164 write('Time of the peaks should be in us')
165 STOP
166 ENDIF
167
168 ------- Set first values as determined for two datasets Visacro ------------
169 IK[1..7] :=[0.13250*Ipeaks[2] 0.11041*Ipeaks[2] 0.11041*Ipeaks[2]
170 0.17666*Ipeaks[2] 0.36437*1.12976*Ipeaks[1] 0.37541*Ipeaks[2] --IK[5] based on Ipeak1
171 0.26499*Ipeaks[2]]
172 P[1..7] :=[2.0 3.0 5.0 9.0 30.0 2.0 14.0]
173 T1K[1..7]:=[3.0 3.5 4.8 6.0 7.00 70. 12.0]*times[2]*0.072727
174 T2K[1..7]:=[76. 10. 30. 26. 23.2 200 26.0*times[3]]*times[2]*0.072727
175
176 IF stp[2] = -1 THEN -- standard value for max steepness if user gives -1
177 stp[2] := 20.18/(72727.27*times[2])*(Ipeaks[2]/45284) -- value scaled with second peak
178 ENDIF
179
180 ENDINIT
181
182 EXEC
183
184 IF t<1*timestep THEN -- run this part only once --------------------------
185 FOR cnt :=1 TO 5 DO -- repeat FOR loop several times ---------------------
186
187 USE GPARAM as GP1[cnt] ---------- Use submodel to get parameters -------
188 DATA IKsub[1..7] :=IK[1..7]
189 T1Ksub[1..7] :=T1K[1..7]
190 T2Ksub[1..7] :=T2K[1..7]
191 Psub[1..7] :=P[1..7]
192 Timessub[1..3] :=Times[1..3]
193 OUTPUT pCurrents[1..4]:=Currents[1..4]
194 Ptime[1..8] :=time[1..8]
195 PCharI[1..2] :=charI[1..2]
196 PCharT[1..2] :=charT[1..2]
197 ENDUSE ----------------------- End use of submodel -------------------
198
199 ------------------Change current function 5 ------------------------------------
200 FOR L:=1 TO 5 DO
201 FIT[L] := heidler(times[1],L,IK[L],P[L],T1K[L],T2K[L])
202 ENDFOR
203 temp1 := FIT[1]+FIT[2]+FIT[3]+FIT[4]
204 Delta:=(1-((PCharI[1]-temp1)/(Ipeaks[1]-temp1))) -- correct the first peak amplitude
205 IK[5]:= IK[5]*(1+Delta)
206
207 Delta:=(1-(Ptime[5]/stp[2])) -- correct maximum steepness -------
208 P[5]:=P[5]*(1+Delta)
209
210 Delta:=(1-(Pchart[1]/Times[1])) -- correct time of first peak ------
211 T1k[5]:=T1k[5]*(1+Delta)
212 T2k[5]:=T2k[5]*(1+Delta)
213 ------------- End change current function 5 ----------------------------
214
215 USE GPARAM as GP2[cnt] ---------- Use submodel to get parameters -------
216 DATA IKsub[1..7] :=IK[1..7]
217 T1Ksub[1..7] :=T1K[1..7]
218 T2Ksub[1..7] :=T2K[1..7]
219 Psub[1..7] :=P[1..7]
220 Timessub[1..3] :=Times[1..3]
221 OUTPUT pCurrents[1..4]:=Currents[1..4]
222 Ptime[1..8] :=time[1..8]
223 PCharI[1..2] :=charI[1..2]
224 PCharT[1..2] :=charT[1..2]
225 ENDUSE ----------------------- End use of submodel -------------------
226
227 ------------------Change current function 7 ------------------------------------
228 FOR L:=1 TO 6 DO
229 FIT[L] := heidler(times[2],L,IK[L],P[L],T1K[L],T2K[L])
230 ENDFOR
231 temp1 := FIT[1]+FIT[2]+FIT[3]+FIT[4]+FIT[5]+FIT[6]
232 Delta:=(1-((PCharI[2]-temp1)/(Ipeaks[2]-temp1)))
233 IK[7]:= IK[7]*(1+(Delta))
234 ------------- End change current function 7 ----------------------------
235
236 ENDFOR ------ end of FOR loop
237 ENDIF ------ end of part that runs only once ------------------------------
238
239
240 ------------------Output current -----------------------------------------------
241 IF t>=stp[1]+timestep THEN -- start after Tstart --------------
242 FOR L:=1 TO 7 DO
243 FIT[L] := heidler(t-stp[1],L,IK[L],P[L],T1K[L],T2K[L])
244 ENDFOR
245 IOUT := polarity*(FIT[1]+FIT[2]+FIT[3]+FIT[4]+FIT[5]+FIT[6]+FIT[7])
246 ELSE
247 IOUT:=0
248 ENDIF
249
250 ENDEXEC
251 ENDMODEL

You might also like