To my grandparents: Mujita and “el

papito” Telésforo for giving me all the

love and courage that have helped me to
face each stage of my life” … I know
you would be proud of this

and

Teo, Clarita, Myriamcito,

Mami, Guillo, Marthica and Vivis
Thanks for teaching me how to pursue
my dreams and giving me the strength
to always continue
List of Papers

This thesis is based on the following papers, which are referred to in the text
by the Roman numerals assigned below.

I Arevalo L, Cooray V, Montaño R, and Roman F: “Modeling of

positive discharges in laboratory gaps under switching im-
pulses”. Paper presented at the XVII International Conference
on Gas Discharges and their Applications GD2008; September,
2008.

II Arevalo L, Cooray V, and Montaño R: “Breakdown effect in

long gaps under switching impulses: statistical variation”. Paper
presented at the 29th International Conference on Lightning
Protection ICLP 2008; June, 2008.

III Arevalo L, Cooray V, and Montaño R: “Breakdown times and

voltages probability calculation using a simplified numerical
methodology”. Paper presented at the 35th Grounding and 3rd
Lightning Physics Conference — GND & LPE; November,
2008.

IV Arevalo L, Cooray V, and Montaño R: “Numerical simulation

of long sparks generated by positive switching impulses.” Pub-
lished paper Journal of Electrostatics Vol 67/2 - 3 (2009) ISSN
0304-3886 pp. 228 - 234.

V Arevalo L, Cooray V, and Wu D: “Laboratory long gaps simu-

lation considering a variable corona region”. Paper presented at
the 30th International Conference on Lightning Protection
ICLP 2010; September, 2010.

VI Arevalo L, Cooray V, and Wu D: “A reliable numerical method

for the calculation of breakdown voltages”. Paper presented at
the 30th International Conference on Lightning Protection
ICLP 2010 - September, 2010.
 VII Arevalo L, Cooray V, Wu D and Jacobson B: “A new static
calculation of the streamer region for long spark gaps”. Paper
submitted to Journal of Electrostatics

VIII Arevalo L, Cooray V, Wu D and Jacobson B: “The develop-

ment of long spark gaps: Simulation including a variable strea-
mer region”. Paper submitted to IEEE Transactions on Dielec-
trics and Insulation Materials.

IX Arevalo L, and Cooray V: “Reliable model for the calculation

of negative leader discharges under switching impulses”. Paper
presented at the 36th Grounding and 4th Lightning Physics
Conference - GND & LPE; November, 2010.

X Arevalo L, and Cooray V: “A preliminary model to simulate

negative leader discharges”. Paper submitted to Journal of Ap-
plied Physics D

XI Arevalo L, and Cooray V: “On the interception of lightning

flashes by power transmission lines”. Accepted for publication
by Journal of Electrostatics.

A resume paper of this manuscript was presented at the Nordic

Insulation Symposium NORD – ISS; June, 2009.

XII Arevalo L, and Cooray V. “The mesh method in lightning pro-

tection Standards — Revisited” Journal of Electrostatics Vol 68
Issue 4, August, 2010, pp. 311 - 314.

A resume paper of this manuscript was presented at 30th Interna-

tional Conference on Lightning Protection ICLP 2010- Septem-
ber, 2010

XIII Arevalo L, and Cooray V: “Influence of multiple upward con-

necting leaders initiated from the same structure on the lightning
attachment process.” Paper presented at the International Sympo-
sium on Lightning Protection - SIPDA X; November, 2009.

XIV Cooray V, Fernando M, Arevalo L, and Becerra M: “Interaction

of multiple connecting leaders issued from a grounded structure
simulated using a self consistent leader inception and propagation
model”. Paper presented at 30th International Conference on
Lightning Protection ICLP 2010; September, 2010.
The author has contributed to other publications which are not incorporated
in this thesis, these are as follows:

XV Cooray V, Arevalo L, Rahman M, Dwyer J, and Rassoul H. “On

the possible origin of X-rays in long laboratory sparks”. Journal
of Atmospheric and Solar Terrestrial Physics. Vol 71 Issues 17 –
18. pp. 1890 – 1898. December, 2009.

XVI Arevalo L, Diaz O., and Mejia A. “Design and construction of

Surge Protection Devices for rural distribution transformers based
on the hygroscopic properties of sand”. International Symposium
on Lightning Protection – SIPDA IX. Brazil 2007.

XVII O Diaz, Arevalo L, and Roman F. “Understanding the corona

space charge influence on the ESD-like discharge produced by a
floating electrode in a fast impulse current generator”. Interna-
tional Symposium on Lightning Protection – SIPDA IX. Brazil,
2007.

XVIII O Diaz, Arevalo L, Cooray V, and Mejia A: “Electrical break-

down in soil under uniform background electric fields”. Paper
presented at the 36th Grounding and 4th Lightning Physics Con-
ference - GND & LPE; November, 2010.

Reprints were made with permission from the respective publishers.

Contents

1 Introduction .............................................................................................. 21
2 The behavior of positive switching impulses in long spark gaps ............ 24
2.1 The mechanism underlying positive leader discharges .................... 25
2.2 Leader inception models ................................................................... 27
3 A simplified and a complete methodology to calculate the breakdown
process in long spark gaps under positive switching impulses ......... 31
3.1 Simplified methodology for the simulation of long spark gaps ........ 31
3.1.1 The initial methodology ......................................................... 31
3.1.2 The statistical variations......................................................... 33
3.1.3 Comparison of the equations for two different leaders .......... 36
3.2 Full version of the long spark calculation for a variable streamer
region ................................................................................................ 40
3.2.1 Methodology .......................................................................... 41
3.2.2 Results of the simulations ...................................................... 46
4 Validation of the methodology through ultra high voltage laboratory
experiments .............................................................................................. 50
4.1 Experimental set-up .......................................................................... 50
4.1.1 Rod-plane air gaps ................................................................. 51
4.1.2 Sphere-plane configuration .................................................... 53
4.1.3 Variable distance to the wall .................................................. 54
5 Negative leader discharges under switching impulses ............................. 57
5.1 Mechanism behind negative leader discharges ................................. 57
5.2 Negative leader models..................................................................... 58
5.3 A preliminary model to simulate negative leader discharges under
switching impulses............................................................................ 59
5.3.1 “Pilot streamer system” .......................................................... 61
5.3.2 “Negative leader phase” ......................................................... 62
5.4 Application of the methodology ....................................................... 63
6 The lightning attachment process and applications to IEC standards ...... 66
6.1 Lightning protection models and standards ...................................... 66
6.2 Lightning attachment model ............................................................. 67
6.2.1 The methodology ................................................................... 68
6.2.2 Applications to the lightning attachment process .................. 70
6.3 Multiple connecting leaders from a grounded structure ................... 75
7 Conclusions.............................................................................................. 78
8 Future Work ............................................................................................. 81
Svensk sammanfattning ................................................................................ 83
Acknowledgements ....................................................................................... 85
References ..................................................................................................... 87
Nomenclature

A A constant that depends on the electrode configura-

tion in Rizk model (equation 2.1)
A, b Constants required to calculate the final jump dis-
tance (equation 5.1)
ai [m] The distance between the streamers from Goelian
and Lalande’s model
ai [m] The radius of the leader segment i at a set time in
Gallimberti’s leader model (equation 3.11)
a0i [m] The initial radii of the leader for Gallimberti’s
leader model equation
dt [s] An infinitesimal time interval
dl [m] The length of each segment of the leader channel
Ei [V/m] The electric field at the start of the leader
Esc [V/m] The electric field in the streamer zone
E [V/m] The potential gradient of the fully developed leader
channel
ELi [V/m] The potential gradient across the leader at segment
i
H [m] The height of the grounded structure under consid-
eration of the electro-geometrical method. (equa-
tion 5.4)
H [m] The height of the cloud (assumed to be equal to
4000 m) for Cooray et al’s downward leader (equa-
tion 5.3)
hf [m] The height of the conductor using the protection
cone method
hg [m] The height of the shielding wire using the protec-
tive cone method
Ip [kA] The peak magnitude of the prospective lightning
current
IL [A] The current injected into the leader channel
K A geometric constant required to calculate the
streamer charge (appears in Chapters 3 and 6)
k A constant that depends on the electrode shape and
atmospheric conditions used in equation (3.4)
K The total number of segments of the leader in Gal-
limberti’s equation (appears in Chapters 3 and 6)
LL(t) [m] The length of the leader channel at time t
L [m] The length of the parallel streamers used in Goe-
lian and Lalande’s approximation
Ll [m] The leader channel length
N The number of parallel streamers assumed by Goe-
lian and Lalande’s model
N [particles/m3] The molecular density
ni [particles/m3] The density of neutral molecules at a particular
time.
n0i [particles/m3] The initial density of neutral molecules
P0 [Pa] Atmospheric pressure at 20 degrees C and at sea
level
P [Pa] Pressure
pi(t) The probability distribution for the inception of
corona
pe(t) The probability distribution of a primary free elec-
tron being in the vicinity of the electrode tip
Q [C] The total electric charge of any streamer burst
Q [C] The electric charge
Qs [C] The critical electric charge of the streamer and the
stem zone
qL [C/m] The charge per unit length required to transform
the streamer into a new leader segment
R [m] The radius of the sphere in the rolling sphere me-
thod
R A constant that depends on the electrode configura-
tion in the Rizk equation (eq. 2.1)
S [m] The final jump distance
T [s] Time
T [K] Temperature
Tfront [s] The front time of the switching impulse voltage
waveform
UC [V] The continuous leader inception voltage
Uc [V] A constant that depends on the electrode configura-
tion in the Rizk model
U1 [V] The potential profile after the first leader has ad-
vanced
Utip [V] The potential at the tip of the leader channel
U1 [V] The voltage induced by the downward leader de-
fined by the Rizk equation (eq. 2.1)
Vmax [V] The maximum voltage applied in the switching
impulse
x0 The relationship between the leader velocity v and
the leader time constant using the Rizk leader
equation (eq. 3.2)
z0 [m] The height of the leader tip above the ground for
the Cooray equation (eq. 5.3)
[·] The angle of protection defined by the protective
angle method
The factor function of the return stroke current
related to the transmission line voltage in the elec-
tro-geometrical method
li [m] The length of the new segment by which the leader
channel has advanced
L(t+dt) [m] The total length of the leader channel at the time
t + dt
UL [V] The potential drop across the segment of the leader
0 [F/m] The permittivity of vacuum
The ratio between the heat capacity at constant
pressure and the heat capacity at constant volume
[º] The angle of deviation determining the tortuous
behavior of the leader channel
Abbreviations

FEM Finite element method

LTE Gallimberti’s local thermoequilibrium equation

CSM Charge simulation method

NC Negative streamer

PL Pilot leader

SL Space leader

NL Negative leader
1 Introduction

“The greatest enemy of knowledge is not ignorance;

it is the illusion of knowledge.”

Stephen Hawking

For some years the insulation and lightning protection designs of high and
extra high voltage equipment have been based on experimental tests and
have made use of equations derived from experiments. Global population
growth and the increase in industrialization throughout the world have in-
creased the demand for electricity, which has led to a search for energy
sources in remote places. In consequence, greater demands are imposed on
the network of electrical power systems, and the need to transport electricity
long distances more efficiently has increased. For all of the above reasons, it
has become necessary to develop and design equipment for operation at elec-
tric field levels higher than ever reached before, and at which the equations
developed in the past are no longer applicable. Thus, it has become neces-
sary to understand the physics behind the discharge phenomena to make it
possible to face these new challenges and to bring about technological ad-
vance. As a result of this, several experimental and numerical scientific
works have been dedicated to the study of discharges in long spark gaps
under different voltage stresses, electrode configurations and polarities.

Experimental studies have been used to develop empirical and physical

models based on experimental formulations and, together, these have led to
improvements in the electrical design of high voltage apparatus and insula-
tion distances, but they cannot take into account factors associated with fun-
damental physics and/or the behavior of materials. The physical models are
used to describe the discharge phenomena and to predict the behaviour of the
observable properties of lightning and laboratory discharges. However, be-
cause of the complex simulations necessary to reproduce real cases, they are
not in widespread use in the engineering of practical applications. Addition-
ally, there are still many unanswered questions about the theory of discharge
development, both for laboratory tests and lightning physics, and the pa-
rameters required to fit the models.

21
Hence, it is necessary to continue working on the methods developed to re-
duce the number of unproved hypotheses and to validate and complement
the engineering models used. For these reasons, the aim of the work pre-
sented here was to study the breakdown process in long gaps under positive
and negative polarity and the initiation of positive leaders from grounded
structures under the influence of downward moving negative lightning
stepped leaders; a process known as the lightning attachment process.

Chapter 2 contains a brief summary of the current knowledge about positive

discharges, including, in particular, the physics and the models underlying
them. Chapter 3 and Papers I to VIII are entirely dedicated to determining
a reliable model to calculate the breakdown phenomena in long spark gaps
under positive switching impulses by means of the physics of the discharge.
The first section starts with a presentation of a simplified methodology; from
the said methodology, it was observed that one of the main parameters re-
quired in the simulation process is the characteristic of the streamer region.
Generally, this is obtained by considering the streamer region to be a geome-
trical region with a constant geometrical shape during the complete dis-
charge process. Later, to simplify the procedures, the streamer region was
approximated by a constant, K, called the geometrical constant.

When a switching voltage is applied to an electrode arrangement (i.e., a high

voltage and a grounded electrode), the background electric field changes
with time. Thus, if the background electric field is modified, the streamer
zone could extend over a larger or a smaller area. With the aim of reducing
the number of assumptions required in the calculation of long gap discharges
under switching impulses, a new model was developed and presented in
Section 3.2. In Paper VII the methodology was introduced and is compared
with other authors’ propositions; Papers V, VI and VIII present the results
of the application of this streamer calculation to a complete calculation of
long spark gaps.

With the intention of describing the electrical potential of the leader, two
alternative formulations were compared; a description of this comparison is
presented in Chapter 3 and Paper IV. Statistical delays have been included
to reproduce the time taken by the inception of corona and, based on labora-
tory tests, Gaussian approximations were found to describe the tortuous be-
haviour of the discharge channel.

Several ultra high voltage tests performed at ABB Ludvika were simulated
for two different configurations: sphere-plane and rod-plane to validate the
long spark gap methodology proposed in Chapter 3. The results are summa-
rized in Chapter 4; good accuracy was obtained for the breakdown voltage
and the time to breakdown.

22
Chapter 5 and Papers IX and X are concerned with the first attempts to
reproduce negative leader discharges under switching impulses. The existing
work on the simulation of negative switching discharges had been held up by
the lack of experimental data and by the fuzzy understanding of the physics
involved. In the scientific community, it is well known that most of the dis-
charges that occur in nature are of negative polarity and, because of their
complexity, the only way to understand them is to generate the discharges in
laboratories under controlled conditions. The waveshape applied to this
problem in laboratories is a negative switching impulse. An electrostatic
approximation of the negative leader discharge process is presented with the
intention of applying the available information to a more realistic method
than has been done previously. The simulation procedure takes into consid-
eration the physics of positive and negative discharges, considering the nega-
tive leader to be propagating towards a grounded electrode and the positive
leader to be propagating towards a rod electrode.

Most of the knowledge gained on the physics of long spark gaps under
switching impulses has been incorporated in the development of models on
lightning attachment. During recent years, Uppsala University has been
working to develop a consistent model to simulate positive upward leaders
and the process of lightning attachment. Consequently, one implementation
of a lightning attachment model is included in Chapter 6 and Papers X to
XIV, in which attention is paid to the international standards implemented to
protect structures against lightning. The standards on lightning protection
design suggest different methodologies which are based on empirical meth-
ods or equations.

Chapter 6 and Papers XIII and XIV deal with the effect of having multiple
upward leaders developing from grounded structures. Whenever an upward
leader is incepted on a structure, other upward leaders can be incepted if the
conditions of the electric field allow it. The newly incepted upward leader
can influence the background electric field and the evolution of the lightning
attachment process.

23
2 The behavior of positive switching
impulses in long spark gaps

For years the designers of external insulation for extra high voltage and ultra
high voltage transmission systems have been interested in understanding the
electrical and physical behaviour of the discharges to enable them to coordi-
nate the insulation system and the electrical protection, and to achieve the
optimal design for the equipment. A first approach to understanding the elec-
trical discharge phenomena has been made by conducting experimental stud-
ies. Furthermore, equations and numerical models have been proposed in-
corporating the experimental data obtained in the laboratory.

In 1971 Les Renardieres’ Research Group [1] was entrusted with the task of
understanding the physics underlying long spark gaps, and the different pa-
rameters involved in the breakdown process, based on an experimental pro-
gramme. The experimental work consisted of performing a systematic varia-
tion of the different conditions relating to the discharge, such as changing the
applied voltage (by introducing positive and negative switching impulses
with a range of rise times, fall times and maximum magnitude), and using
various electrode configurations and gap distances. Once the experimental
results had been obtained, the data was divided among the scientists, ena-
bling different approximations and models to be proposed [2- 4].

One of the greatest contributions of Les Renardieres’ Group was to identify

the fact that positive discharges are composed of two parts: the leader chan-
nel and the streamer zone. This research enabled them to describe the com-
plete discharge process right up to the instant at which the breakdown took
place. From the experimental data collected by Les Renardieres’ Group,
different models have been proposed. Some of them are known as engineer-
ing models, which are approximations or equations fitted to the experimental
data; these include models such as the critical radius concept of Carrara and
Thione [5], the leader intensification criterion of Petrov and Waters [6] and
the leader inception voltage concept of Rizk [7- 9]. With the evolution of the
computational software and hardware capacity, more complex numerical
models have been developed, most of which are based on the detailed com-
pendium written by Gallimberti [10] and the leader-streamer model which
was the result of collaborative work, published by Bondiou et al [11]. Some

24
of the leader inception models based on the real physics of the discharge are
Lalande’s upward leader models [12], Goelian and Lalande’s simplified
approximation [13, 14] and Becerra and Cooray’s model [15].

After the Les Renardieres’ Group work, the interest of the scientific commu-
nity focused on the application of the numerical equations and methods de-
veloped to areas such as rocket-triggered lightning, lightning protection, and
lightning attachment. Nowadays, most of the numerical models that are pre-
sented to the scientific community have been validated using the results ob-
tained by Les Renardieres’ Group.

As far as the consideration of the breakdown behaviour under switching

impulses is concerned, the industrial sector is still performing a variety of
experimental tests to facilitate the design of equipment and substations, be-
cause the theoretical methods developed to date are incapable of predicting
breakdown voltages and times to breakdown under different electrode con-
figurations. In addition, the available methods are time-consuming and they
have several input parameters to which the models are highly sensitive and
which need to be changed for each different spark gap configuration.

The lightning attachment models [14, 15] have been validated first through
the laboratory work of the Les Renardieres’ Group [1] and subsequently
through the experimental work conducted with rocket-triggered lightning
[14], but they have assumed several parameters that also need to be vali-
dated.

2.1 The mechanism underlying positive leader

discharges

The mechanism underlying long air-gap discharges was the first attempt to
provide a theoretical explanation for the physical observations and was
based on the application of switching impulses to a high voltage electrode-
plane arrangement by Les Renardieres’ Group [1, 2] and the physical expla-
nation of the data presented in the references [10, 16, 17]. The process starts
with the appearance of a bunch of filaments called streamers or the streamer
zone at a point in time, designated t1, developing from the high voltage elec-
trode. Continuing the analogy of the bunch of filaments, it is possible to
observe a region that is illuminated in the root area of the streamer, which
reveals the path along which the leader channel forms. The charge located in
the streamer zone reduces the local electric field over the tip of the high
voltage electrode. However, the reduction of the electric field brought about

25
by the streamer charge is compensated for by an increase in the voltage at
the high voltage electrode, which increases the electric field. Then it is fol-
lowed by a period characterised by the absence of light, a so-called dark
period.

At t2, as shown in Figure 1, the simultaneous development of a leader chan-

nel and a second streamer takes place. The first leader segment is called the
“stem” and it is determined by the following two criteria [10]:
- The combined current of the first corona streamers passing through
the root of the first corona burst will cause heating of the gas. If the
temperature rises to 1500 K, the detachment of negative ions will
take place producing an increase of electrons in the medium, and
thus increasing the plasma conductivity. As a result of this, this zone
becomes highly conductive and increases the electric field at the
high voltage electrode.
- The electric field at the extreme of the stem (the closest point to the
grounded electrode) should be high enough to allow the formation of
another stem and a second streamer.

This transition from streamer to leader is called unstable leader inception,

and it is brought about when the total charge Q in the second streamer is
equal to or larger than about 1 C [10]. After the time t2, a system composed
of a leader channel and a streamer zone at the head of this leader channel
starts to evolve; this stage is called the “Leader – Streamer” stage. The evo-
lution of this “Leader – Streamer” is highly influenced by the rise time of
the applied switching impulse. When the rise time is fast, the propagation of
the discharge is continuous, and when it is very slow, the propagation is
discontinuous. When the energy available at the front of the leader tip is high
enough to sustain the thermalization of the channel and the creation of new
leader segments, the leader starts to propagate continuously with streamers
developing at its tip. This condition defines the stable leader inception.

During the propagation, the leader advances with a velocity of 104 m/s and it
is supplied at its end by an average current of 0.5 to 1 A [3, 17].

The final stage of the leader propagation or the “final jump” is produced
when the streamer zone of the leader reaches the opposite electrode. In this
stage the leader is accelerated from a velocity of the order of cm/µs to m/µs.

26
Figure 1. A streak image and sketch of the development of the positive leader dis-
charge in a laboratory long air gap with a 10 meter gap, tfront = 500 µs and Vmax = 2.5
MV (adapted from [17] with permission of Elsevier).

2.2 Leader inception models

Based on the experimental data and the previous physical details analyzed by
Gallimberti [10, 17] and Les Renardieres’ Group [2], different equations and
numerical models have been developed to calculate the leader inception
voltage. The models have been validated with the measurements from the
Les Renardieres’ Group and good agreement has been obtained. The models
and the equations obtained for the propagation have been extrapolated for
use on lightning and, nowadays, are well-known concepts used in the design
of the lightning protection of grounded structures.

The knowledge gained about the mechanism of the discharge in air has fa-
cilitated the improvement of engineering models for the prediction of break-
down in gaps, such as the Golde’s criterion [18, 19], and the models of
Lemke [20], Alexandrov [21], Jones [22], Hutzel [23], and Bazelian [24].
These models do not enter into detail on physical aspects of the phenome-
non, but they utilize a large amount of information, qualitative and quantita-
tive, gained on the characterization of each one of the phases of the dis-
charge mechanism. A brief description of them can be found in Arevalo’s
work [25].

Other models tried to approach the physics of the phenomena through labo-
ratory experiments, by identifying equations for the leader channel and the
streamer region, such as those of Carrara and Thione [5], Rizk [8-9, 26] and
Dellera and Garbagnati [27- 28], and Petrov and Watters’ criteria [6].

A concept widely known as “the critical radius concept” was proposed by

Carrara and Thione [5] and is now in widespread use. The model was built

27
up from observations on rod-plane and conductor-plane configurations under
positive impulses with a critical front. The model has been extensively ap-
plied to phase-to-ground and phase-to-phase configurations, finding good
agreement between the computed and the experimental results. Nevertheless,
the study extended by the Les Renardieres’ Group [3] for the same configu-
rations with other voltage waveforms found that the value of the critical ra-
dius changes for different values of the time to crest for the applied voltage.

The critical radius principle was first used on lightning studies by Eriksson
[29]. Since then, it has generally been applied to compute the leader incep-
tion conditions in rods, masts, power lines [7, 8, 26-28, 30, 31] and in build-
ings [32]. As a result, any sharp point on a structure, such as the tips of
lightning rods, corners or edges, is rounded off to the critical radius to mi-
nimize the probability of a lightning strike, and it is assumed that a stable
leader is initiated when the electric field on the surface is equal to the critical
streamer inception electric field (about 3 MV/m).

A more physically realistic model was presented by Rizk [8, 26], whose
model introduces the criteria for leader inception and the breakdown voltage
of different electrode configurations under positive switching impulses with
a critical time to crest. It takes into account the conditions of the leader and
the conditions for the final jump. The zone containing the streamer and stem
are characterized by the critical charge Qs and the lengths of the leader chan-
nel and the air gap. This condition is satisfied under impulse voltages that
increase, keeping the potential at the leader tip approximately constant as the
leader propagates in the gap. The model results in expressions for the conti-
nuous leader inception voltage UC for different gap configurations:
Uc (2.1)
UC
A
1
R
where A, Uc and R are magnitudes that depend on the electrode configura-
tion. The model shows a good agreement with the experimental results. The
application of this model to air density effects is reported in the work of
Rizk [26, 9], where it is proposed that a continuous positive leader is in-
itiated from any grounded structure of height h when the voltage U1 induced
by the downward leader and/or the thundercloud at a height h in the absence
of the structure is equal to:
Uc (2.2)
U1
A
1
h

28
The values of the constants A and Uc have been determined for rods, tall
masts and transmission lines [9]. The results of this model are based on la-
boratory data pertinent to switching impulses and it is doubtful whether they
are also valid in the case of electric fields generated by stepped leaders. Moreo-
ver, the results are valid for isolated conductors which make its application
more restricted.

Another model was proposed by Dellera and Garbagnatti. This model in-
volves the simulation of the propagation of positive and negative leaders
from data based on electrostatic calculations. The model is known as the
leader progression model. In this case, the downward leader is modeled by a
line with a given charge density, which is a function of the prospective re-
turn stroke current, while the upward leader is represented by line charge
segments of constant density. The attachment of the downward leader to a
grounded structure was computed from the calculated striking distance and
the relative velocity ratio of both leaders. The criterion for leader inception
used is the critical radius concept [5].

Petrov and Waters [6] have developed a leader inception criterion, which
assumes that when the streamers from the electrode extend beyond a critical
length equal to 0.7 m, positive upward leaders can develop. The electric field
over the streamer zone must exceed a critical value of 500 kV/m. However,
laboratory experiments with long sparks in the rod-plane configuration have
shown that the critical streamer length for unstable leader inception is equal
to about 3 m [4]. This latter criterion was used by Akyuz and Cooray [33] to
study lightning attachment to Franklin rods. The model of Petrov and Waters
was extended in research conducted by D’Alessandro and co-workers [34]
by assuming that the stable propagation of the incepted leader is reached if
the rate of change of the potential induced by the downward leader at the
rod’s tip is larger than 6 kV/ s.

Other physical models delve deeply into the particles’ behavior, like Bon-
diou and Gallimberti’s model [11], which introduced the basic physics prin-
ciples, and other authors’ work, like that of Lalande [12], Goelian et al [13],
Castellani [35], and Becerra and Cooray [15], where modifications were
made to the model introduced by Gallimberti [10] and in which lightning
attachment processes or the breakdown in long gaps were calculated. The
assumptions underlying the streamer calculation are different in each model,
as described in the following paragraphs:

Bondiou and Gallimberti’s model [11] calculates the charge generated by the
streamer formation in terms of estimating the total number of electrons that
have left the gap by reaching the high voltage electrode. The calculation
takes into account the attachment process. However, as a simplification, the

29
charge is assumed to come from a single filament, therefore the charge for
the total streamer area is estimated by multiplying the charge from a single
streamer by a branching factor and by the number of filaments. In this pro-
cedure, the constants used for the first streamer and all subsequent streamers
are the same.

A simplified model for the simulation of positive sparks was developed by

Goelian and Lalande’s model [12, 13] calculates the charge associated with
the advancement of the streamer-leader by multiplying the area formed by
the potential distribution immediately before and after the streamer forma-
tion by a geometrical constant. For the calculation, the streamer area is as-
sumed to be comprised of N parallel streamers of length L, located at a dis-
tance ai from each other. The charge simulation method is then used to eva-
luate the electrical charge from the area covered by the leader between the
potentials at two different instants of time and the representation of the N
linear streamers.

The Becerra and Cooray streamer approximation [15, 36, 37] assumes that
the area covered by the streamer zone required for the simulation process is
conical. The charge accumulated is calculated by means of the charge simu-
lation method. In the most recent version, Becerra and Cooray [36] proposed
a simplified method in which the area accumulated between the potential of
two consecutive leader segments is proportional to the streamer charge. The
proportional constant used to calculate the charge is called the geometrical
constant K.

The last three methodologies mentioned above presumed that the leader ad-
vances by segments and each segment is characterized by the thermo-
equilibrium equation of Gallimberti [10]. The charge per unit length required
to achieve the transition to a new leader segment in the case of the models of
Bondiou and Gallimberti [11], Lalande’s [12] and Becerra and Cooray [15]
is derived from thermodynamic analysis of the transition region where the
streamer converges on the leader tip [10].

30
3 A simplified and a complete methodology
to calculate the breakdown process in long
spark gaps under positive switching
impulses

A new and simplified methodology for the calculation of breakdown volt-

ages and the time to breakdown in long spark gaps is presented in Papers I
and IV. It is based on the previously described physical methodologies and
the well-known “streamer-leader” model [11]. Papers II and III concern the
application of the methodology mentioned.

After analyzing the advantages and disadvantages of the different methods it

was determined that the most critical part of the calculation is that for the
streamer. Therefore, two electrostatic approximations for calculating the
streamer region are presented in Paper VII and the implementation of one
of them to the long spark gap methodology is presented in Papers V, VI and
VIII.

3.1 Simplified methodology for the simulation of long

spark gaps

The simplified model has been modified, mainly in the calculation of the
leader channel where two different models were implemented and compared.
For all of the cases simulated, the model used the finite element method
software COMSOL® to calculate the electric field and the potential distribu-
tion between the high voltage electrode and the grounded electrode. For the
input of the physical equations and routines, Matlab ® was employed.

3.1.1 The initial methodology

In Paper I, a non-probabilistic methodology is presented. The calculation is

divided into two main parts, one for the streamer zone and the other for the
leader channel. The streamer zone was modeled using the geometrical ap-

31
proximation of Lalande [12] (see equation 3.1). The leader channel was
modelled using Rizk’s equation [7], which is an approximation taken from
long gap experiments [2].

The streamer charge is calculated using the following equation:

(3.1)

where (U2 - U1) is equivalent to the potential change attributable to the

streamer charge region and K is a geometrical factor introduced by Lalande
[12, 13].

In contrast to previous methodologies, the model supposes that the ionization

process is generated in the zone between the leader and the front of the
streamer, which is in direct contrast to the presumption made in the model of
Bondiou and Gallimberti [11], which supposes that the ionization process
only takes place in the head of the streamers. In Figure 2, dashed lines can be
seen representing the area assumed to be the new region after the time
(t+dt).

Rizk’s equation is applied, for the calculation of the leader channel, which
established that the voltage at the tip of the leader is a function of the leader
length Ll, the electric field in the streamer zone Esc, the final quasi-stationary
leader gradient E and the relation x0, which is the relation between the lead-
er velocity v and the leader time constant

Ll ( t )

U tip Ll E x0 E ln
Esc Esc E
e x0 (3.2)
E E

The advancement of the leader channel is calculated using the equation

Q(i ) (3.3)
lL(i )
qL

where qL is the charge per unit length required to cause the transformation of
the streamer located in the active region in front of the already formed leader
channel into a new segment of the leader. The magnitude of qL is based on
the observations made by Les Renardieres’ Group [2]

The methodology presented in Paper I makes a simplification of the calcula-

tion of the background potential for each time step. The potential is calcu-
lated only once with FEM in per unit scale, and for the next time-step, the

32
curve obtained is scaled up as the potential increases in the high voltage
electrode.

Figure 2. Representation of the potential distribution variation for two different time
steps. The background potentials are represented by green and blue lines, the leader
potential drop using Rizk’s equation (3.2) and the streamer zone as a constant elec-
tric field of 500 [kV/m]. The charge necessary for the next time step corresponds to
the dashed line.

The methodology was compared with the experimental results presented by

Les Renardieres’ Group [1], and with the results presented by Lalande [12],
Goelian et al [13], and Becerra and Cooray [15]. Good agreement was ob-
served between the existing models and the proposed methodology.

3.1.2 The statistical variations

The time at which an electrical discharge is initiated is uncertain and depen-

dent on many unknowns, which introduces one source of statistical depen-
dence. Another source is introduced by the tortuous nature of the path taken
by the discharge. One statistically dependent source of time delay associated
with breakdown corresponds to the combined effect of every elementary
process that the discharge needs to develop before breakdown can occur, and
the other arises from the random nature of the path the discharge takes to
reach the ground. In Paper II, the statistical variations mentioned here are
incorporated into the model. A brief summary of Paper II is presented in the
following paragraph.

The first source of time delay corresponds to the random nature of the coro-
na inception, which requires a primary free electron in the vicinity of the
electrode tip which is capable of initiating an ionization process. The proba-

33
bility distribution for inception was taken from the critical volume region of
corona proposed by Baldo [16]:
t
pi t pe (t ) exp pe t dt (3.4)
0

pe t k V (t ) t (3.5)
1
Ti (3.6)
2 k V (t )

In the above equations pi(t) is the probability distribution for the inception of
corona, pe(t) is probability distribution corresponding to a primary free elec-
tron being in the vicinity of the electrode tip, V(t) is the variation of the po-
tential with time, and k is a constant that is determined by the electrode
shape and atmospheric conditions, which was calculated from measurements
conducted by Les Renardieres’ Group [2].

Initially, to represent the random behavior of the leader channel, experimen-

tal measurements of the tortuosity of lightning discharges made by different
researchers were used [3, 38-39]. A Gaussian distribution with a mean value
of 15 degrees and a standard deviation of 11 degrees was selected to
represent the tortuous path. For every segment of the leader, a random selec-
tion of the angle from the Gaussian distribution was calculated, therefore
each resultant segment has a magnitude dl and an angle , as shown in Pa-
per II.

The advance of each successive leader segment was calculated from the
following equations:
Q (3.7)
l
qL

l l cos (3.8)

Lt dt LL (t ) ll (3.9)

34
 TORTUOUS LEADER CHANNEL VS GAP DISTANCE
10

DEVIATION IN GROUND PLANE[m]

4

-2

-4

-6

-8

-10
0 1 2 3 4 5 6 7 8 9 10
GAP DISTANCE [m]

Figure 3. Tortuous leader channel obtained in one of the simulations after the geo-
metric projection of the vector dl onto the straight path, where the potential distribu-
tion was solved using COMSOL ®

With the intention of evaluating the methodology including the statistical

variations, a constant level test was simulated to obtain the 50% breakdown
voltage and the 50% time to breakdown. A comparison was made and the
results are presented for a tortuous leader path and a straight path. As shown
in Paper II, the breakdown voltage is not affected by the tortuous behavior
of the leader channel, however, the time to breakdown changes with the tor-
tuosity of the channel. Table 1 is constructed from the data presented in Pa-
per II; it can be seen that the channel tortuosity changes the time to break-
down of the discharge.

50% Time to Breakdown

Calculated value
Waveform
Measured
Value Straight % Tortuous %
path error channel error

500/10000 µs 552 µs 579.36 µs 4.9% 609 µs 11%

1800/10000 µs 980 µs 787.7 µs 19% 809.4 µs 17%

2200/10000 µs 1183 µs 943.3 µs 3% 953.4 µs 19%

Table 1. Comparison of the time to breakdown using a straight discharge channel

and a tortuous discharge channel

The percentage error is smaller when a straight channel was used for the
simulations than when a tortuous path was assumed. This high percentage

35
error arises because the normal distribution used for the tortuous path is
based on lightning measurements owing to the lack of information available
on tortuous leader channels in long spark gaps under switching impulses.

Analyzing the effect of the leader direction, it can be observed that the
change of direction generates a reduction in the leader velocity. Therefore,
the breakdown might be delayed, i.e., the time to breakdown of the discharge
is increased.

3.1.3 Comparison of the equations for two different leaders

With the aim of improving the model by making a more physically realistic
approximation, two different leader models were implemented in the simpli-
fied methodology and the behavior of the models was compared. Papers III
and IV present the results obtained and the comparison of the models, re-
spectively.

The first leader criterion used was the Rizk engineering approximation [7]
assuming tortuous and non-tortuous leader paths. Gallimberti’s concept of
“local thermodynamic equilibrium” [10] was implemented to introduce a
more physically realistic approach to the voltage drop along the leader chan-
nel.

With Gallimberti’s approach the leader was decomposed into elementary

segments of length dl, temperature T, pressure P and with a molecular densi-
ty n that is uniform along the channel. In this instance, the potential drop
UL in the segment i will be:

(3.10)

where dli is the length of the segment i and ELi is the potential gradient of the
segment of length i.

Gallimberti’s model gives the evolution of the internal electric field EL as a

function of the current injected, calculated under the supposition that the
conductivity of the leader channel is essentially controlled by electronic col-
lisions between neutral molecules and electrons accelerated in the electric
field EL. It supposes that all the injected current IL is used to extend the lead-
er and the expansion is made with a constant mass. The formulation describ-
ing this hypothesis results in the following set of equations:

a02i n0i dli ai2 ni dli (3.11)

36
 p0 d a2
EL I L
1 dt
(3.12)

where ai is the radius of a leader segment at a particular instant in time, ni is

the density of the neutral molecules at the same instant and a0i and n0i are the
initial conditions required for the leader formation. P0 is atmospheric pres-
sure, the ratio between the specific heats at constant pressure and volume,
d( a2) is the variation in the cross-section of the leader, and EL IL is the pow-
er injected into the channel during a time step dt.

From equation (3.11), it is possible to calculate the channel section for the
next time step as a function of time at the time t, and the internal electric
field and the charge IL dt
a2 t dt a2 t
1
E I dt
(3.13)
p L L
0
As the mass from the molecules is constant, the density can be written as:
a2 t (3.14)
n t dt n t 2
a t dt

And, using the hypothesis that the ratio EL/n is constant, the internal electric
field as a function of time will be given by:

n t dt (3.15)
E L t dt EL t
nt

Thus, it is feasible to calculate the time evolution of the internal electric field
for each segment and the potential drop along the leader channel

k
UL E Li dli (3.16)
i 1

where k is the total number of segments.

As the goal was to improve the calculation by adopting a more realistic ap-
proach than had previously been taken, in Papers III and IV, the electric
field in the gap was recalculated for every time step, taking into account the
new leader segment and the voltage applied to the high voltage electrode.
The models were compared by conducting three different tests: one was a
deterministic simulation performed with Rizk’s equation and the “Local
Thermoequilibrium, LTE” equation of Gallimberti’s, presented in Paper IV.
Another was a statistical simulation including an up-and-down test, pre-
sented in Paper III, and a constant level test for different peak voltages for a
positive switching impulse waveform 500/10000 µs applied to the rod elec-

37
trode which is discussed in Paper IV. The simulations were made using the
Les Renardieres’ Group laboratory configuration [1]. The characteristics of
the setup are a 10 m point-plate with a 10 mm radius cone at the tip, forming
the point.

In Paper IV and Figure 4, the results for a deterministic calculation are pre-
sented, i.e., without statistical variations and by making a simplified calcula-
tion of the streamer zone, the breakdown conditions using the LTE equation
are fulfilled faster than using Rizk’s equation. It is important to mention here
that Becerra and Cooray [15] simulated the long gap discharges using the
same set-up as Les Renardiers’ group [1] using their own model, which is
also based on the physics described by Gallimberti [10]. They found good
agreement with experimental data, however, in their calculation, they eva-
luated the charge in the streamer zone by using the charge simulation me-
thod for each time step. This procedure was not used in the present engineer-
ing application because the aim of this part of the research was to evaluate
and analyze a simplified calculation avoiding a large number of computa-
tions.

Figure 4. Numerical simulation of the propagation of the positive discharge in a

point-plate laboratory configuration. Comparison between Rizk’s equation and the
LTE equation. From left to right and top to bottom, the figures presented are: a
streak image of the leader channel, the internal electric field of the leader channel,
and the velocity and current of the leader channel. Reproduced from Paper IV.

The resulting accuracy of the engineering model was improved by using the
“LTE equation of Gallimberti”, in conjunction with the statistical distribu-
tions representing the corona inception and the tortuous leader channel. The

38
results showed that the mean value of the leader current and the velocity
agree best with the experimental results, as can be observed in Figure 5.

From the results obtained with the statistical simulations performed with the
LTE equation of Gallimberti’s and the Rizk equation, it is possible to con-
clude that the time to breakdown and the leader characteristics are quite
close between both leader criteria. It is important to outline that Rizk’s equa-
tion was deduced from Les Renardieres’ Group results [1, 3], therefore it is
only to be expected that good agreement would be obtained with this equa-
tion. Meanwhile, Gallimberti’s LTE equation, together with the other ap-
proximations used in the procedure, need to have statistical variations incor-
porated for the path taken by the discharge to give the correct results.

Figure 5. Numerical simulation of the propagation of the positive discharge in a

laboratory using Gallimberti’s LTE equation with statistical delays incorporated to
represent the inception of the streamer and the path of the discharge. Taken from
Paper IV.

As an application of the model, an up-and-down test and a constant level test

were simulated. The results are presented in Papers III and IV, respectively.
The breakdown probability at different voltage levels and different wave-
forms is computed, but unfortunately the values obtained cannot be com-
pared with experimental data since the experimental data available in the
literature is sparse.

The contributions made up to this point to improving the calculation can be

found in Papers III and IV and can be summarized as follows:

- The leader channel was simulated to obtain a more physical ap-

proach using Gallimberti’s thermoequilibrium equations [10].

39
 - Two statistical variations were implemented, one for the inception of
the corona, and the other to describe the tortuous nature of the leader
channel.
- For every leader segment, the background electric field was recalcu-
lated, in contrast to the method used in Papers I and II and dis-
cussed here on pages 31 to 36.

3.2 Full version of the long spark calculation for a

variable streamer region

To improve the methodology proposed for the calculation of long sparks,

one must analyze the sensitivity of the input parameters to calculate the
streamer region and estimate the effect of the simplified method used to
compute the streamer zone. Authors have assumed certain physical factors,
like the distribution of the electrical charge, to simulate the streamer region
of the discharge and, in general, this has been achieved by representing it as
a region with a constant geometrical shape. Different geometries have been
used, such as cones [15] or several parallel filaments [12, 13] to determine
which best reproduces the experimental data.

However, when a switching voltage is applied to an electrode arrangement

(comprised of a high voltage and a grounded electrode), the background
electric field varies with time. Thus, if the background electric field is mod-
ified, the streamer zone could cover a different volume.

With the aim of reducing the number of assumptions required in the calcula-
tion of long gap discharges under switching impulses, a new means of
representing the streamer region was presented in Paper VII, and practical
applications of this model to the long-gap calculation methodology were
made in Papers V, VI and VIII. This model considers a variable streamer
zone that changes with the electric field as was described in Paper VII. The
three dimensional region that fulfills the streamer condition is identified for
each time step, and the charge accumulated in that region is then calculated.
The only assumption made in the calculation is the minimum electric field of
the streamer region.

The leader channel was modeled using the local thermoequilibrium model of
Gallimberti [10], with the streamer zone being located in front of the leader
and the statistical variations arising from the inception and tortuous nature of
the leader channel being included. For comparison purposes, different gap
configurations [1, 11, 40, 41] were simulated and their results are presented

40
in Papers VI and VIII. The outputs of the model agreed well with pub-
lished experimental data.

3.2.1 Methodology

A detailed description of the calculation procedure can be read in Papers VI

and VIII. Here, instead, brief explanations of the novel implementations
introduced in the calculation are described in the following paragraphs,
where reference is made to the relevant publications.

3.2.1.1 Streamer Region

Two methodologies were implemented to calculate the streamer region, and

they are discussed and compared in Paper VII; only a brief description is
given here

Procedure 1

The calculation was divided up into four specific steps:

1. Once the avalanche has developed into a space charge cloud of posi-
tive ions, the local field may be high enough to start a new avalanche
closer to the cathode. Whether or not this process continues until the
cathode is reached depends on the applied electric field strength. The
lowest value at which stable propagation appears to be possible is
called the stability field, Est [10]. Its characteristic value in ambient air
is between 400 and 500 kV/m [10, 11]. It has been shown from the
theory that the field inside the streamer channel, is equal to the stabili-
ty field [10, 42-43].
As the stability field is equal to the field inside the streamer channel,
the volumetric region between the tip of the leader channel/or the high
voltage electrode and the ground point is divided into several layers.
For each layer, the area where the field is equal to or higher than the
stabilization electric field (assumed here as 450 kV/m) is identified.

2. In this streamer region, it is assumed that the streamers propagate

along the electric field lines; the drop in the potential along the elec-
tric field line to the boundary of the region is equal to 450 kV/m. As is
illustrated in Figure 6, streamers are assumed to propagate along the
electric field lines, and the drop in the potential along the line is
equivalent to the propagation electric field.

41
Figure 6. The geometry of the streamer region at a specific instant in time; the ar-
rows indicate the direction of the electric field vector. The dotted red lines represent
the layers for which the charge is calculated.

3. The streamer region is then divided into several layers and the charge
is calculated for each layer. This calculation requires that the direction
of the electric field vector be taken into consideration, including at the
edge of the region.
Using Gauss’s theorem, the streamer charge located in each layer Qi is
calculated, and the total charge accumulated in the volume for each
time step, Qtotal is computed. Once the new charge has been found, the
source voltage increases and the calculation to identify the new strea-
mer region is redone.

Qi 0 E dS (3.17)
s
n
Qtotal Qi (3.18)
i 1

Procedure 2

1. In this procedure, as in the first one discussed above, it is presumed

that the streamers propagate along the electric field lines, and there-
fore, for each electric field line from the high voltage electrode, the
respective background potential is determined, and for each potential
profile, the distance at which the background potential satisfies the
condition that the stabilization electric field Estab is equal to
450 kV/m is found.
2. The charge is calculated using Gauss’s theorem, as used in the first
procedure in equations (3.17 and 3.18).

42
For all the calculations carried out, procedure 1 was used, because, as re-
vealed by the comparison made in Paper VII, there are no important differ-
ences between the two procedures.

Figures 7, 8 and 9 demonstrate how the streamer zone changes at different

points in time; the calculations presented have been carried out for the confi-
guration used by Les Renardieres’ Group [2], assuming a conical tip under a
1850 kV switching voltage impulse (500/10000 s), and a gap space of
10 meters. Figure 7 represents the instant at which the charge started to ac-
cumulate in the region between the tip and the plane. This electric charge is
located next to the tip of the leader channel, and the tip is moving away from
the high voltage electrode. As can be observed, the geometrical shape of the
streamer charge takes a form that is approximately conical (dashed lines).

Figure 7. Streamer region in a long gap of 10 meters. Results obtained for a simula-
tion assuming a conical tip-plane geometry. The high voltage rod is located 10 m
from the grounded plane, and the latter is defined as being located at the position
defined as 0 m. The small picture corresponds to a photograph taken of the same
experiment.

However, when the leader channel has propagated some distance towards the
grounded plate electrode, the streamer region changes shape from the conical
one illustrated above to that presented in Figure 8. Once the final jump con-
dition has been satisfied, the results showed that the geometrical shape of the
streamer zone becomes easier to describe again, being more like a cylinder
than the initial cone or the amorphous intermediate geometrical shape (see
Figure 9).

43
Figure 8. The streamer region in a 10 meter long gap, for a gap comprised of a rod
and plane. This simulation was made when part of the leader channel had moved
towards the grounded electrode. The small picture corresponds to a photograph
taken of the same experiment.

Figure 9. Streamer region in a 10 meter long gap. The results obtained are for a
simulation assuming a conical tip and a plane. The grounded plane is located at the
position marked 0 m, and the high voltage rod is located 10 m away from it. The
small picture corresponds to a photograph taken of the same experiment, and it illu-
strates the leader channel and the streamer region.

The photographic results presented in a previous paper [17] are reproduced

in Figure 10; they show a similar time-dependent behavior to that obtained
in the simulations: Initially a conical region was observed, then a hyperbolo-
id and, finally, the charge extends into a shape that is roughly cylindrical in
form.

Figure 10. Streak image of the development of a positive leader discharge in a long
air gap in a laboratory. The air gap was 10 meters, tfront = 500 µs and Vmax = 2.5 MV.
This figure has been published elsewhere [19] and has been reproduced with the
permission of Elsevier.

44
The change in the shape of the streamer region under a switch-like voltage
impulse in the simulations is caused by the incrementation of the voltage
applied to the high voltage electrode. When the voltage source is increased,
the electric field rises, and the region that will fulfill the criterion for a
streamer region grows. This result reveals that the frequently made assump-
tion of a constant geometry for the streamer region during an incrementation
of the electric field disagrees with the real physical process of the discharge.

3.2.1.2 Tortuous nature of the path

Electrical discharges are characterized by: their random nature, taking tor-
tuous paths and having an unpredictable inception time. These delays can be
represented as statistical distributions. The above-mentioned statistical de-
lays were included in the model to make a realistic calculation, as presented
in Section 3.1; however, the statistical distributions used in Section 3.1 cor-
responded to different lightning measurements.

With the aim of obtaining a proper representation of the tortuous path taken
by the discharge, several data from industrial tests of long spark gaps under
switching impulses made at ABB/PS-HVDC were used [44, 45]; Figure 11
presents a collection of paths obtained from a series of tests.

Figure 11. Tortuous path of the discharges for several laboratory tests of switching
impulses made by ABB-PS/HVDC Ludvika for a long spark gap. Taken from [44].

A Gaussian distribution was determined that represents the random behavior

of the leader channel in a suitable manner; it is presented in equation (3.19).
2
x b1
c1
f ( x) a1e (3.19)

45
Where a1, b1 and c1 are constants obtained from tests conducted during the
normal fitting of several long gap switching impulses. It is important to men-
tion that it was observed that the constants in the Gaussian distribution de-
pend on the electrode configuration; examples are presented in the following
table:

Table 2. Coefficients of the tortuous Gaussian distribution with confidence intervals

of 95%. distribution for different electrode configurations
Constant Square rod Hyperboloic rod Sphere
a1 3.67(3.13, 4.21) 4.03 2.33(0.69, 3.97)
b1 0.052(-0.07, 0.18) -0.5 0.89(-0.40, 2.18)
c1 0.29(0.17,0.41) 1.46 1.13(-0.42, 2.68)

3.2.2 Results of the simulations

A series of methodologies was implemented for comparison purposes: the

methodologies proposed by Becerra and Cooray, assuming a conical area for
the streamer region and using the charge simulation method [15], the simpli-
fied method of Goelian et al [13], the simplified method presented in section
3.1, and the new method presented here in section 3.2.

The laboratory geometry used for the calculations was the one used in Les
Renardieres’ Group’s configuration [2], which has been widely used in pre-
vious research to validate models [13, 15, 25]. In addition, to check the re-
liability of the methodologies for different gap configurations, the rod-rod
geometry used by Paris [40] was also studied. A brief summary of the results
is presented here and more details can be found in Paper VIII.

The first set-up was selected because all of the existing numerical methodol-
ogies have been validated using the same rod-plane electrode arrangement,
making it the ideal configuration to make comparisons with. The second set-
up was used because it is known that the electric field distribution of a rod-
rod arrangement is different from the rod-grounded plane one. In addition, as
the streamer zone depends on the electric field, it can be anticipated that the
shape of the streamer zone will be affected by the differences in the electric
field. Therefore, it is anticipated that the assumption of a constant streamer
zone will lead to results that are not in agreement with the measurements.

46
The aim was to show that the proposed methodology can be used for differ-
ent kinds of configurations, and to compare the results with those from other
methods proposed by different authors.

Overall, from the first set-up, one can conclude that the results obtained with
the different methodologies agreed with the work done by Les Renardieres’
Group [2] in with respect to the leader channel position and the breakdown
voltages (see Figure 12).

Figure 12. Streak sketch showing the development of a positive leader and the
streamer discharge in the laboratory in a long spark air gap with a 10 meter gap;
tfront = 500 µs and Vmax = 2.5 MV.

With the purpose of determining if the methodology can reproduce different

electrode arrangements, the rod-rod configuration was selected. Because the
rod electrode has a termination different from the one used by Les Renar-
dieres’ Group, the configuration square rod-plane was simulated; the simula-
tions can be seen in Figure 7 of Paper VIII. The results obtained using the
other authors’ methodologies indicate that, even though the results of the
breakdown can be reproduced for a gap comprised of a rod and plane, their
accuracy is not as good as that obtained with Les Renardieres’ group ar-

47
rangement as errors greater than 20% have been obtained with them. Instead,
using the methodology proposed in this research work, errors of less than
10% were obtained.

The simulations of the rod-rod arrangement are presented in Figure 13 and in

Papers VI and VIII. For this calculation, two different simulations were
made: one with the variable streamer zone without taking into account the
electrical charge that can accumulate on the grounded electrode (i.e., without
checking the condition that the electric field is high enough to generate
streamers on the grounded electrode, namely, that the electric field is
750 kV/m) and the other one taking into account the criterion for negative
streamer inception.

Figure 13. The 50% breakdown voltage for a rod-rod arrangement for different gap
distances. Comparison between the models of Becerra and Cooray CSM [15], and
Goelian and Lalande [13]; the model is introduced Section 3.1, and the method is
described in this section.

The simulations revealed that calculations that assumed a constant streamer

zone, such as those of section 3.1 [25], Goelian and Lalande [13], and Becer-
ra and Cooray [15] will lead to incorrect breakdown magnitudes for a rod-
rod gap configuration, with breakdown errors higher than 80% and values
that are insensitive to changes in the gap distance. In contrast, when a varia-
ble streamer zone is used for the calculation without checking the streamer
criterion on the grounded electrode, the errors in the breakdown voltages are
still close to 20% when compared with the experimental results, however
they follow the variation observed in the gap distance in the experimental
measurements.

The results assuming a variable streamer zone, which took into account the
streamer criteria in the grounded electrode, lead to an acceptable and con-

48
servative error of below 9%, for the distances evaluated and the variation
found with the gap distance in the experimental measurements made by Paris
[40, 41].

Thus, in the research presented in this part of the thesis, two changes have
particularly improved the calculation of the propagation of the leader strea-
mer in a long gap arrangement and the relevant research is discussed in
depth in Papers V to VIII:

- A successful three-dimensional calculation of a variable streamer

zone was proposed and implemented taking into account the mini-
mum electric field that characterizes the propagation of the streamer
region.
- The path of the discharge based on switching impulse measurements
was incorporated in the methodology and different distributions
were found to be required depending upon the electrode configura-
tion under consideration.
- The calculation method results in errors of less than 10% and can be
used in different gap configurations, such us rod-rod and rod-plane
ones.

49
4 Validation of the methodology through
ultra high voltage laboratory experiments

ABB Power Systems HVDC – Ludvika [46] performed several long gap
experiments under positive switching applied voltages to investigate the
insulating behavior of air at ultra high voltages. The series of experiments
conducted consisted of using different electrode configurations with gap
distances that were varied between 6 and 10 meters; and, to the gaps, switch-
ing impulse voltages were applied. The methodology developed and ex-
plained in Chapter 3 was utilized. The breakdown voltages and the time to
breakdown were calculated and compared with the experimental results and
very good accuracy was observed.

4.1 Experimental set-up

The experimental tests were performed at the Ultra High Voltage Laboratory
UHVEN at ABB – Ludvika. The applied voltage consisted of positive
switching impulses (250/2500 s). Two electrode arrangements were used:
rod-plane and sphere-plane. The gap distances were varied from 6 to 10 me-
ters; and the insulation media in all the tests was air under normal atmos-
pheric conditions. The humidity and temperature were registered. The up
and down method was used to test the different configurations following the
IEC standards [47]. For each voltage level, 30 shots were applied. The
breakdown voltage and the time to breakdown were measured.

Two cameras were installed at 90 degree angles to enable them to photo-

graph the path of the discharge in 3D-planes, and for every shot two photo-
graphs were recorded. A fast video camera was used as well to register the
speed of the leader propagation and the geometry of the path of the dis-
charge.

50
4.1.1 Rod-plane air gaps

The rod used was a 6 meter long metallic rod terminated at the top by a fine
electrode with a grounded metallic plane located on the floor. The distance
to the wall was chosen in such a way that there was no probability of break-
down occurring against the wall for any of the experiments performed. The
gap distance was varied between 6 and 10 meters. The high voltage source
was directly connected to the rod electrode and the plane was well grounded.

A total of 30 impulses was applied to every gap arrangement and the vol-
tage measurements were registered as a function of time until the breakdown
had occurred. The breakdown voltage and the time to breakdown were simu-
lated for each of these individual breakdowns and the values obtained were
compared with the experimental values. Figure 14, illustrates the path the
discharge took in a gap distance of 6 meters for a rod-plane configuration;
the sequence of figures was taken with the fast video camera, as the dis-
charge steps along the path.

(a) (b) (c) (d)

Figure 14. Sequence of the path of a discharge in a 10 meter gap obtained after
processing the video taken during the discharge. (a) the discharge at the initial time,
10 µs, (b) after 40 µs, (c) after 120 µs and (d) at 180 µs.

As explained in Section 3.2.1.2., the constants a1, b1 and c1 are required for
the calculation of the tortuous path of the discharge; the Gaussian distribu-
tion and its respective constants are obtained from fitting the photographic
evidence for each impulse applied to the 6 to 10 m long spark gaps. Table 3
presents the different magnitudes for the different kinds of rods used on the
high voltage electrode.

51
Table 3. Coefficients of the tortuous Gaussian distribution with 95% confidence
limits.
Constant Square rod
a1 3.65 (3.11, 4.0)
b1 0.051(-0.07, 0.17)
c1 0.28 (0.16,0.41)

Using the Gaussian distribution presented earlier and the characteristics of

the set-up, each applied switching impulse voltage was simulated and the
results are summarized in the following tables. The error between the simu-
lation and measurements is displayed as well.

Measurements Simulations

Applied Breakdown Time to Breakdown Time to % %

Voltage Voltage breakdown Voltage breakdown error error
[kV] [kV] [ s] [kV] [ s] in Vb in tb

1500 1464,8 2,39 287,81 18,43 1494,6 287,7 -2,03 0,04

1560 1523,5 3,53 339,13 26,09 1548,9 280,43 -1,67 17,31
1440 1409,83 1,72 290,23 14,52 1429 306,15 -1,36 -5,48
-
1380 1353 330,63 1342,2 391,6 0,80
18,44

Table 4. Simulations and measurements of the breakdown voltage and the time of
the breakdown with the percentage error for a rod-plane arrangement for a gap dis-
tance of 6 meters.

Measurements Simulations
Time to
Applied Breakdown Breakdown Time to % %
break-
Voltage Voltage Voltage breakdown error error
down
[kV] [kV] [kV] [ s] in Vb in tb
[ s]
2160 2099,5 7,61 541,15 124,6 2158,15 479,17 -2,79 11,45
2070 2026,4 5,98 536,48 76,47 2026,3 467,31 0,01 12,89
1980 1937 2,64 524,1647 45,19 1948,7 425,35 -0,60 18,85

Table 5. Breakdown voltage and time to breakdown obtained in simulations and

measurements with the percentage error for a rod-plane arrangement with a gap
distance of 10 meters.

The results of the simulations have shown that the methodology can repro-
duce the experimental results obtained for the rod-plane configuration with
good accuracy. The maximum error obtained was 3% for the breakdown

52
voltage and the average error in the calculation of the breakdown voltage
was 1.13%. The maximum error determined in the calculation of the time to
breakdown for a single test was 19% and the average obtained error was
14.4%.

4.1.2 Sphere-plane configuration

The second set-up used for the comparisons was an arrangement comprised
of a sphere connected at the high voltage electrode and the grounded plane
electrode. Two different spheres of 1.3 and 1.6 m radii were tested, and the
gap distance between the sphere and the grounded electrode was varied be-
tween 4, 6 and 8 m. As in the previous test, 30 impulses were applied using
the up and down method. The test was implemented with the photographic
equipment described above. The data obtained were used to determine the
Gaussian distribution of tortuosity for the sphere-plane arrangement.

For the simulation, every case was analyzed individually and the coefficients
obtained for the Gaussian distribution are presented in the following table.
Table 6. Coefficients of the tortuous Gaussian distribution with 95% confidence
limits.
Constant Sphere
a1 2.30 (0.68, 3.97)
b1 0.88 (-0.40, 2.17)
c1 1.13 (-0.42, 2.65)

The data obtained from the measurements are summarized in tables 7, 8 and
9. The percentage error is compared for the breakdown voltages and the time
to breakdown.

Measurements Simulations
Break-
Applied Breakdown Time to Time to % %
down
Voltage Voltage breakdown breakdown error error
Voltage
[kV] [kV] [ s] [ s] in Vb in tb
[kV]
2150 2148 680,57 2063,7 439,32 3,924 35,44
2240 2175,79 5,3 352,38 20,1 2076,5 318,6 4,563 9,58

Table 7. The breakdown voltage and the time to breakdown obtained from simula-
tions and measurements, with the percentage error, for a sphere of 1.3 m diameter in
a sphere-plane arrangement with a gap distance of 6 m.

53
 Measurements Simulations
Break- Time to Time to
Applied Breakdown % %
down break- break-
Voltage Voltage error error
Voltage down down
[kV] [kV] Vb tb
[kV] [ s] [ s]
2400 2326,7 6,26 427,05 36,98 2146,4 404,32 7,75 5,32
2320 2300,5 4,14 521,8 37,5 2135 415,32 7,19 20,41
2240 2207 361,65 2135,5 334,39 3,24 7,54

Table 8. The breakdown voltage and the time to breakdown obtained from simula-
tions and measurements, with the percentage error, for a sphere of 1.3 m diameter in
a sphere-plane arrangement with a gap distance of 8 m.

Measurements Simulations
Break- Time to Time to
Applied Breakdown % %
down break- break-
Voltage Voltage error error
Voltage down down
[kV] [kV] Vb tb
[kV] [us] [us]
2440 2332 7,25 196,166 4,54 2202,6 177,61 5,55 9,46
16,6
2360 2337 8,8 264,3 28,17 2195,6 220,32 6,05
4

Table 9. The breakdown voltage and the time to breakdown obtained from simula-
tions and measurements, with the percentage error for a sphere of 1.3 m diameter in
a sphere-plane arrangement with a gap distance of 4 m.

From the results of the simulation is possible to conclude that the simula-
tions can reproduce the experiment with very good agreement. The maxi-
mum error for the breakdown voltage is 7.75% and the average error is
5.36%. The time to breakdown presented higher errors than the breakdown
voltage; the maximum error in the time was 20.41%, obtained for the gap
distance of 8 meters and the average error for all the tests was 15.6%. It is
important to note that the simulated magnitudes of the breakdown voltage
are lower than the measured values, which will give a conservative value
from the point of view of designing the insulation.

4.1.3 Variable distance to the wall

With the aim of evaluating the precision of the simulation when the back-
ground electric field changes because of the presence of a wall (another
grounded electrode) and when two gaps are competing (wall-high voltage
electrode, high voltage electrode-wall), simulations were made of a sphere-
plane gap with a variable distance to the wall. Simulations of breakdown to
the wall and breakdown to the grounded plane were evaluated independent-

54
ly. The experimental measurements were carried out at ABB Ludvika [48],
and the experimental set-up consisted of a 1.3 m diameter sphere and a
grounded metallic plane. Four different gap distances to the earthed elec-
trode were tested and four distances to the wall. As in the previous series of
experiments, for every test, 30 switching impulses were applied and every
single test was simulated. The magnitudes of the breakdown voltage and
current were calculated and the results, with their respective errors, are
summarized in the following table.

Dist. Breakdown Voltage 50% Time Breakdown 50%

Gap % %
to
m Simula- error Simula- error
Wall Measurement Measurement
tion tion
4 6 1825 1649.7 9.61 193.8 192 0.93
4 10 1869 1829 2.14 195.4 193.6 0.92
5 10 1962 1816 7.44 248.9 224.7 9.72
6 10 2025 1821 10.07 314.6 296.3 5.82
8 10 2125 2096 1.36 375.1 369.9 1.39
6 8 1965 1921 2.24 337.3 319 5.43
5 8 1918 1800.9 6.11 263.1 204 22.46
6 7 1937 1805.3 6.80 373 314.3 15.74
5 7 1918 1878.9 2.04 263.1 257.6 2.09

Table 10. The measured and simulated breakdown voltage and time to breakdown,
with the error calculation for a sphere-plane configuration. The diameter of the
sphere was 1.3 m for all the cases and the gap distance and the distance to the wall
were varied.

As is evident from table 10, the maximum error in the breakdown voltage
was 10% and the average error was 5.31%. For the time to breakdown, the
error was higher than the error in the breakdown voltage; the maximum val-
ue was 22.46% and the average magnitude was 7.16%.

These results show that the methodology can reproduce the breakdown vol-
tages and the times to breakdown more accurately by taking into account the
effect of changes in the background electric field. It is important to mention
that the magnitudes of the breakdown voltage for sphere-plane gaps are con-
servative from the point of view of determining the insulation requirements
(i.e., lower magnitudes are obtained for the breakdown voltages using the
proposed methodology than those found in the measurements).

From this chapter it is possible to conclude that:

55
 - The results showed that the methodology proposed in Chapter 3 can
reproduce the breakdown voltages and the times to breakdown for
the sphere-plane and rod-plane configurations with good accuracy.
The average error for the breakdown voltage is lower than 6% and
for the time to breakdown is lower than 22% in the worst case.
- The effects of nearby walls, which will change the background elec-
tric field, are successfully reproduced using the proposed methodol-
ogy.
- The methodology was tested with several experimental measure-
ments and it was observed that the magnitudes of the breakdown
voltages calculated are lower than the measurements. This gives a
conservative value that can be used to design protection distances.

56
5 Negative leader discharges under switching
impulses

The majority of the modeling work has been dedicated to the consideration
of positive discharges because of the complexity of the negative discharge
process. In nature, however, most of the lightning discharges are of negative
polarity. Therefore, the development of numerical models capable of repro-
ducing negative discharges will improve the design of lightning protection
systems and the design of equipment that can withstand overvoltages caused
by direct and/or indirect strikes.

A first step to attaining the goal of modeling negative lightning discharges is

to model laboratory gap discharges, as these are developed in controlled
environment; for example in the form of laboratory discharges under switch-
ing impulses, as presented in Papers IX and X.

5.1 Mechanism behind negative leader discharges

The physical description assumed here is based on the characteristics pre-

sented by Castellani 1998 [49], Les Renardieres 1978 [4], Lalande 2002 [50]
and Gallimberti et al 2002 [17]. This description corresponds to observations
of experiments conducted with a rod-plane arrangement to avoid interaction
between positive and negative discharges.

The first phenomenon that occurs during a negative discharge is the forma-
tion of a negative streamer (NC) at time ti. Then afterwards, at time t1, a dis-
charge process called a “pilot” appears (PL), in which simultaneous devel-
opment of a positive streamer propagating towards the cathode and a nega-
tive streamer propagating towards the anode occurs.

At time t2, a space leader (SL) develops from a space stem, and at t3, the
conditions are ripe for a negative leader (NL) to propagate towards the
anode. From the published results [11, 17], one can conclude that the ther-
modynamic conditions for the formation of positive leaders ought to be very
similar to those for negative leaders.

57
At time t4, a stem is formed in the space where the two pilots were located.
However, this stage is not observed for gaps of less than 4 m [4, 49]. From
this stem, a negative streamer and a leader crossing the space (SL) are
formed. The leader possesses two heads, one of which is a positive leader
and propagates towards the cathode, the other, being a negative leader, prop-
agates towards the anode.

When the space leader reaches the negative electrode, a negative streamer is
formed and the leader channel is illuminated brightly. This step is called a
step discharge. Briefly, the negative discharge propagates discontinuously,
hence the name, and with complex processes taking place between the nega-
tive and positive streamers. Figure 15 was taken from [17] to illustrate the
various stages of a negative discharge.

Figure 15. Negative leader development process. This figure has been reproduced
from [17] with the permission of Elsevier.

5.2 Negative leader models

The first lightning attachment model incorporating physical assessments was

proposed by Dellera and Garbagnati [51]. This model, called the leader pro-
gression model, includes the simulation of the propagation of both negative
and positive leaders using electrostatic considerations. Subsequently, from
Les Renardieres’ experimental work [1, 3], Gallimberti [10] developed the
physics pertinent to understanding the development of leaders. Subsequent-
ly, Gallimberti and Bondiou [11] built up a physical methodology that simu-
lates positive upward leaders under switching impulses. Taking the work of
Bondiou and Gallimberti [11] as their starting point, Becerra and Cooray
developed a model to study the attachment of lightning flashes to grounded
structures[15]. However, neither Dellera and Garbagnati [51] and Rizk [8,

58
7], nor Becerra and Cooray [15] modeled the negative steeped leader from
first principles.

In 1994, Bacchiega et al [52] developed the first theoretical model of the

propagation of a negative stepped leader across long air gaps. By 1998, Cas-
tellani et al [49] had conducted experimental measurements in such long
gaps to identify the bi-leader process, giving a great amount of detail of the
negative leader development.

In 2000, Mazur et al [53] presented the results of a physical model of nega-

tive leaders generated under the effect of a downward-coming positive lead-
er. In 2002, Gallimberti and Lalande et al [17, 50] presented a bipolar model
to reproduce triggered lightning in which they included positive and negative
leaders based on the measurements of Les Renardieres [3] and Castellani
[49] and the model of Bacchiega [52].

The other available models are stochastic, such us the electrical breakdown
models of Niemeyer et al. [54], Tsonis and Elsnet [55], and Sanudo et al.
[56], with some of them using the electrical characteristics of the leader
channel published by Petrova [57]. The latest available model by Beroual et
al [58, 59] takes electrical network parameters derived from electromagnet-
ism, wave propagation and gas discharge theory to calculate the evolution of
a lightning discharge.

All of the above-mentioned models are approximations that fail to reproduce

all the different stages of the negative discharge in one way or other. There-
fore, to reproduce processes such as the relaxation time or the positive leader
propagation or space leader, they tune parameters or assume circuital ar-
rangements.

5.3 A preliminary model to simulate negative leader

discharges under switching impulses

In Papers IX and X, a new physical model is proposed that takes into ac-
count the stages and characteristic measurements identified by Castellani
[49] and the subsequent phases of the physics of the negative discharge iden-
tified and used by Castellani, Mazur and their co-workers [49, 50, 53]. The
methodology considers processes such as the negative leader channel, com-
posed of a negative and positive streamer and the negative leader stem. For
the positive and negative streamer, we used the streamer criterion of Gallim-
berti [10].

59
Once the first streamer has been incepted, the charge in the streamer zone is
calculated, assuming that the streamer zone is characterized by a constant
electric field of 450 kV/m and 750 kV/m for positive and negative streamers,
respectively. The next stage is the so-called “pilot system”, and the calcula-
tion is made as described in Paper X.

The voltage drop in the leader channel was evaluated using the LTE equa-
tions of Gallimberti [10]. The step leader propagation is reproduced until the
final jump takes place. The condition for the final jump is satisfied when the
negative streamer reaches the grounded electrode. Figure 16 presents the
flow diagram for the calculation.

Figure 16. Flow chart for the simulation of the development of the negative leader.
Ei is the electric field at the start of the leader and E is the electric field at infinity
attributable to the leader channel

60
5.3.1 “Pilot streamer system”

Even though there is no clear explanation of the physics underlying the dis-
charge process because of its complexity and the lack of experimental data,
some experimental results have indicated that each pilot streamer discharge
starts at the lower tip of the preceding one and that this form of initiation is
associated with every current pulse [4, 49]. A consistent explanation has,
however, been proposed by Bacchiega and co-workers [52].

For the calculation of the potential profile and the reproduction of a pilot
system, the following considerations have to be taken into account:

1. It has been stated in the literature that the electric field necessary for
the propagation of a negative streamer is constant and has a value of
750 kV/m [10]. Therefore, the region in front of the tip that fulfills
this electric field criterion has been identified, and the maximum axi-
al length of the region has been calculated.
2. Once the negative streamer is formed, a positive leader discharge
will develop in the direction of the high voltage electrode. The simu-
lation of this positive leader follows the methodology presented in
Chapters 3 and 4. The inception and direction of the propagation of
this positive leader is assumed to be at the central axis of the ar-
rangement towards the high voltage electrode.
3. Laboratory measurements made by Les Renardieres’ Group have
shown that, before the formation of the first pilot system, there is an
enhancement of the electric field between the negative streamer and
the grounded electrode, leading the electric field in the local region
next to the tip of the negative streamer to have a magnitude of 1·106
V/m [4]. The existence of this electric field in the vicinity of the tip
of the negative streamer region and the requirement that a positive
discharge can propagate towards the high voltage electrode makes it
necessary to assume that, from the tip of the negative streamer re-
gion, a first stem leader develops towards the high voltage electrode
with an electric field magnitude of 1·106 V/m. In front of this stem
leader, a positive streamer is located.
4. For the simulation of the positive and negative leader channel Gal-
limberti’s LTE equation is used [10].
5. The potential distribution of every pilot streamer discharge will be
composed of the potential drop: over the positive streamer region, in
the positive leader channel and in the negative streamer region lo-
cated in front of the grounded electrode.

There are plenty of different phenomena going on that need to be

61
 considered, these effects are: the positive leader channel, the positive
and the negative streamers, and when they are combined, the relative
impact of the different phenomena gives the total charge, which
leads to the characteristic pulsing of the pilot system, defined by the
so-called “relaxation time”, which has been tuned to RLC electrical
circuits in other models.

The charge per unit length that is necessary to sustain the progression
of the pilot streamer discharge was assumed to be 13.8 C/m as this
was the value measured by Castellani and co-workers [49]. In Figure
17, one can observe the potential profile distribution before and after
the streamer development, and after the formation of the first pilot
streamer discharge.

Figure 17. Potential distribution before and after the negative streamer development
and after the formation of the first pilot streamer discharge. The insert shows an
enlargement of the potential of the incepted leader and the streamer channel.

5.3.2 “Negative leader phase”

The measured results extracted from the literature [4, 49] reveal that the
charge required to incept a leader channel is of the order of 5.4 C. This
value was used in the present calculation to initiate the simulation of the
leader stage. This charge includes all of the charge associated with the dif-
ferent streamers and the pilot systems developed up to the inception of the
leader.

We assumed that the inception mechanism for the negative leader is similar
to the one for the positive leader [60]. This is reasonable because the thre-
shold charge corresponds to the charge necessary to heat the “stem” to a
critical temperature, which will lead to the formation of the first section of
the leader channel. To calculate this negative leader, the thermo-equilibrium

62
equations of Gallimberti [10], described in Chapter 2, are used. The initial
condition of the potential gradient before a new leader segment is created is
assumed to be 7.5 x 105 V/m, taken from experimental evidence [10], which
is a different condition from the one assumed for positive leaders. The
charge per unit length necessary to sustain the negative leader propagation
used in the simulations was 108 C/m based on values published in the lite-
rature [49].

5.4 Application of the methodology

In order to check the validity of the model, two different rod-plane configu-
rations were tested, with two different switching impulses and gap distances.
The results are presented in Paper X. First, the methodology was applied to
the same geometrical set-up as that used by the Les Renardieres Group [3],
which consists of a gap between a conical rod and a plane. The conical elec-
trode had a tip radius of 10 mm. The simulation was made assuming a gap
distance of 2 m with an applied voltage of -1550 kV; a waveshape of
20/1600 s was used.

The results of the simulations presented in Figures 18 and 19 showed that the
trend exhibited by the simulated current agrees with the measured signatures
determined by the Les Renardieres’ Group. The maximum magnitude of the
current is 10 A for both the measurements and simulations. The simulation
of the current pulsates, which is because of the restarting process and be-
cause of the exchange of charge in the channel arising from the presence of
both a negative leader and a positive leader moving towards the high voltage
electrode.

Figure 18. Measured and simulated current pulses during the stem-space propaga-
tion in a 2 m wide gap between a conical rod and a plane with a switching impulse
of 20/1600 s, configuration used by Les Renardieres.

63
Figure 19. Results of simulations obtained using the measurements of Les Renar-
dieres for a 2 m conical rod-plane gap during a switching impulse of 20/1600 s.

Overall, the model results including the path of the discharge are in good
agreement with the experimental results of Les Renardieres [3]. However, it
is important to notice that, even though the shape of the pulses is in agree-
ment with the measurements, slight differences are evident in the timing. All
pulsation behavior corresponds to the pilot streamer system representation
and that the initiation of pilot streamers is dependent on the exchange of
positive and negative charges.
The final parameters available for comparison are the time to breakdown
and the breakdown voltage, which were calculated and are summarized in
Table 11.

Breakdown Time to Break-

Voltage [kV] down [ s]
Les Renardieres 1481 20.6
Simulations 1546 28
% error 4.5 35.9

Table 11. Comparison between the measurements and the results of the simulations
for a 20/1600 s impulse across a gap of 2 m between a conical rod and a plane.

The time to breakdown and the breakdown voltages exhibited errors of

35.9% and 4.5%, respectively, for the particular case shown. The higher
error in the time to breakdown can be attributed to the path that the discharge
has taken. As was demonstrated in the references [60, 61], the path followed
by the discharge determines the time to the breakdown, but it does not have a
notable influence on the calculation of the breakdown voltage.

Owing to a lack of detailed experimental information in the literature, it is

not possible to extract more parameters for the same configuration, but data

64
for positive and negative streamer charges and for the velocity of the posi-
tive and negative leaders can be obtained from the calculation. A comparison
with a gap distance of 7 m, maximum voltage 2.8 MV and a waveform of
6/3000 µs using the configuration studied by Les Renardieres [4] was made
and the results can be found in Paper X.

A self-consistent numerical simulation based on the physical process behind

the negative leader discharge has been presented in Papers IX and X. It
includes a novel way to calculate the streamers for the pilot system and
enables the complete negative discharge to be predicted. The experimental
results are compared with simulations and show good agreement on the re-
presentation of the magnitudes of the current and path of the discharge.

65
6 The lightning attachment process and
applications to IEC standards

The theories and the design of the protection of structures against lightning
rely heavily on the data for long gap discharges. The methods accepted as
the standards were derived from empirical approximations or equations
taken from data from long gap experiments under switching impulses. Vari-
ous authors from the scientific community have proposed models that adopt
a more physical approach to the representation of the lightning attachment
process. Nevertheless, some assumptions concerning the initial values or
regarding the nature of the process have had to be made.

6.1 Lightning protection models and standards

Lightning protection standards laid down by the IEC [62, 63] specify three
procedures that can be used to implement a lightning protection system for
grounded structures. They are the rolling sphere, the mesh and the protection
angle methods.

The rolling sphere method comes from a simplified version of the electro-
geometric method [64-67]. According to the electro-geometric method, a
down-coming stepped leader will get attached to the first part of a grounded
structure that comes within a critical distance of the tip of the stepped leader.
This critical distance is called the striking distance, and was defined by
Golde [19].

In the rolling sphere method, the radius of the sphere is selected such that its
radius is equal to the striking distance. The radius of the sphere R is defined
as a function of the probable return stroke current according to the relation-
ship between the lightning striking distance and the peak return stroke cur-
rent derived by Whitehead for power transmission lines [67]. The design of
the lightning protection system uses the rolling sphere method, in which a
sphere with a specified radius is rolled over the surface of the ground and
over the structure to be protected, allowing the protected and the unprotected
objects or parts of the structure, covered by the shadow of the sphere or

66
touched by the sphere respectively to be visualised [62, 67]. Any part of the
structure that is in contact with the surface of the sphere is considered to be
vulnerable to a direct lightning strike; the untouched volume falling outside
the shadow of the sphere defines a lightning protected zone. The radius of
the sphere is defined as being equal to 15 m or 50 m, depending upon the
level of lightning protection required.

The second method proposed for the positioning of air terminals is the pro-
tective angle method. This procedure is recommended for simple structures.
The positioning of air terminals, masts and wires is determined by taking
into account the fact that all parts of the structure to be protected fall inside
the volume defined by the surface generated by projecting a line from the air
terminal to the ground plane, at an angle to the vertical.

The equation defining the protective angle was established by Wagner et al.
[68, 69]. The relationship between the connection distance for the upward
and downward leaders, called the final jump, and the point at which the re-
turn stroke current Ip starts was proposed to be:

S A I pb [m] (6.1)

S is the final jump distance in meters, Ip is the peak lightning current in kA,
and A and b are constants that take on different values depending upon the
author [65, 66, 70-75]. Fixed angles are still in use as a design tool today.

The third method, especially recommended for the protection of flat surfac-
es, is the mesh method. According to this method, a conducting mesh is used
with a cell size determined by the minimum return stroke current that is al-
lowed to strike the protected structure [62]. The mesh has to be located at a
critical distance above the flat surface to be protected to avoid a direct strike.

6.2 Lightning attachment model

A number of authors [12, 76, 77] have used lightning attachment models
with the aim of probing the lightning protection methods accepted by the
standards [12, 76, 77]. However, their results are not conclusive owing to the
parameters assumed and/or the complexity of the routines applied. Thus,
more results are needed to validate the methods accepted in the international
standards.

67
The methodology used for long gap simulations discussed in Chapter 3 was
extended and applied to the lightning attachment cases in an attempt to
create a more physical approximation to the international standards and to
improve the practical design of protection against lightning. A brief descrip-
tion of the methodology applied is presented here, a more detailed discussion
can be found in Papers XI-XIV.

6.2.1 The methodology

The procedure adopted in the calculation is divided into three different stag-
es. In the first stage, the structure is scanned to identify the locations at
which the streamer inception criterion of Gallimberti [11] is satisfied. The
second stage involves an analysis of stable leader inception. After the strea-
mer inception criterion is fulfilled, the dynamic leader inception procedure is
used to locate the points from which a stable leader can be incepted. The
final stage involves the analysis of the movement of the upward leader to-
wards the down-coming leader and the final attachment process.

6.2.1.1 Streamer Criteria

The methodology applied considers a straight downward-coming leader

approaching a structure, as was first proposed by the leader progression
model of Dellera and Garbagnati [27, 28] and later applied by Lalande [12]
and Becerra and Cooray [15, 37, 76-77]. It consists of representing the
downward leader as a line with a given charge density. The charge density is
a function of the probable return stroke peak current and, in this case, its
value is calculated using the equation proposed by Cooray et al [78]:

6
( z) 8 10 1 G ( z0 ) I p
H z0 (6.2)
a b
2
H ( z0 ) I p C/m
1 c d

with
z0
G ( z0 ) 1
H
z0 z0

H ( z0 ) 0.3 e 50
0.7 e 2500 (6.3)
z z0

68
where z0 is the height of the leader tip above the ground in meters, H is the
height of the cloud in meters (which is assumed to be equal to 4000 m), Ip is
the peak return stroke current, a = 7.2 10-5, b = 5.297 10-5, c = 1.316 and
d = 1.492 10-2.

The downward leader channel is divided into segments of increasing length

from the base of the cloud to the tip of the leader to compute the background
electric field. The charge density in each segment is assumed to be constant
and is equal to the charge in the lower part of the segment as given by equa-
tion (6.2).

Once the first corona has been incepted, the charge in the streamer zone is
calculated using the simplified electrostatic approach proposed by Lalande
[12] and N. Goellian et al [13], which assumes that the streamer zone is
characterized by a constant electric field. If the streamer inception criterion
is fulfilled in one of the corners of the building, the leader inception proce-
dure will be initiated.

6.2.1.2 Leader Inception

The stable leader inception criterion is divided in two parts: one involves the
calculation of the charge and the dimensions of the streamer zone, and the
other concerns the development of the leader channel. For the calculation of
the features of the streamer zone, the procedure used by Lalande and Goelian
et al [12, 13] is applied. This procedure assumes the streamer zone to have a
fixed geometry with the number of streamers determined by the experimen-
tal measurements of Les Renardieres’ Group [3]. However, the geometrical
constant, K, for the calculation of the streamer zone proposed by Becerra
and Cooray [37] was used in the analysis presented in Paper XII. The vol-
tage drop in the leader channel was evaluated using the LTE equations of
Gallimberti [10]. Once the radius of the leader channel was given as an input
parameter, these equations predict the development of the electric field in-
side the leader channel.

For each time step, the potential caused by the leader channel and the poten-
tial change arising from the streamer zone were analyzed. The charge inside
the streamer region was recalculated and was kept for the next time step, as
was done by Lalande [12] and Becerra and Cooray [15]. Nevertheless, it is
important to emphasize that the advance of the leader is not only calculated
from the charge in the streamer zone, but also from the charge attributable to
the leader channel. In addition, in the calculation presented here, no linear
approximation was made for the background potential as had to be assumed
in previous models [15, 76, 77].

69
6.2.1.3 Lightning attachment

Once an upward leader is incepted, it propagates in the direction of the

downward-coming stepped leader. The simulated lightning attachment pro-
cedure consists of analyzing the propagation of the stable upward leader
while the down-coming leader is approaching, as was done in the Becerra
and Cooray methodology [37]. The calculation was terminated when the
streamer zone of the upward leader reaches the downward-coming leader.

6.2.2 Applications to the lightning attachment process

The lightning attachment methodology was used to validate two well-known

engineering concepts, namely the protection angle cone and the mesh me-
thod, which correspond to the respective topics of Papers XI and XII.

6.2.2.1 The protection cone method

Shielding wires are required to protect power lines against external overvol-
tages, such as lightning discharges. The basic idea of a shielding wire is to
create a volume for the conductors of the power line that is protected, and
which will offer effective protection against lightning strikes, in accordance
with the standards. The angle of protection for the shielding wire can be
determined using “the effective electro-geometrical method” [74, 75] or the
“perfect shielding concept” [30, 66] if the magnitude of the current, and the
heights of the conductor hf and the shielding wire hg are known. Figure 20
illustrates how the calculation of the angle of protection is carried out.

M S

B E
2H
B S
S
2H
H
A C D
hg
hf

3H

Figure 20. Determining the protection angle. The basic electro-geometrical method
is illustrated on the left, where H is the height of the structure and corresponds to
the angle of protection. On the right, the perfect shielding concept is illustrated; S
corresponds to the striking distance, hg to the height at which the shielding wire is
located, and hf is the height of the conductor wire. is a factor function of the return
stroke current [79].

70
The calculation of the protection angle is based on the following equations:

hg
k1
S
hf (6.4)
k2
S
x 2 2
1 K1 1 K2
S
x
arctan
hg hf

In the analysis, presented in Paper XI, four AC power transmission line

structures were simulated, two for 230 kV and two for 500 kV. The position
of the downward leader was changed in the x-y plane. Figure 21 depicts the
variation of the position in the y coordinate for a transmission line of
230 kV. On average, sixteen different points were analyzed for each struc-
ture and each downward leader. The points considered correspond to the
vertices of the structure, where the conductors and the shielding wire are
supported, and, for each point on the structure an analysis was performed to
determine whether there would be streamer or/and leader inception and/or
leader propagation.

Figure 21. Position of the downward leaders and the distance between upward and
downward leaders. The red line corresponds to the downward leader at y = 10 [m],
the green color corresponds to the downward leader at y = 5 [m] and the black cor-
responds to the downward leader at y = 0 [m].

The most important results are presented here, however the detailed results
for every structure can be found in Paper XI. For a single circuit structure of
230 kV, the simulations showed that the shielding wire is always struck ra-
ther than the other wires. Even so, inception of streamers and leaders takes
place on the other conductors. Table 12 includes a summary of the position

71
of the downward and upward leaders when the final jump condition is
reached.

Downward
Downward leader conditions; z-coordinate [m] at the point analyzed
leader
Shielding wires Conductors
y-
x y y-coordinate y-coordinate y-coordinate y-coordinate
coordinate
[m] [m] = -3.96 m = 3.96 m = -7.92 m = 7.92 m
=0m
0 0 42.31 42.31 27.25 35.25 35.25
0 5 43.18 45.01 27.32 33.32 33.95
0 10 38.42 40.48 23.33 31.33 31.41

Table 12. Final Jump: The results for different y-coordinates for the downward lead-
er over a 230 kV single circuit structure. The z-coordinate at which the downward-
coming leader is located when the final jump condition is reached is displayed for
the shielding and the conductors.

For a double circuit structure, like the 500 kV one, the results showed that
there is leader inception on the points where the conductors are supported,
but there is no leader propagation from the points of inception.

Downward Leader
Downward leader z-position
Position
Shielding Phase Phase Characteristic of upward
x [m] y [m] wire A B leader for
40.36 [m] 52 [m] 37 [m] phases A and B
72.5
0 0 69.39 65.22 Streamer inception
Final jump
79.43 Leader inception, not pro-
0 5 69.29 70.22
Final jump pagation
79.41 Leader inception not pro-
0 10 69.28 66.43
Final jump pagation
83.43 Leader inception not pro-
0 15 69.28 64.54
Final jump pagation

Table 13. The calculated z-coordinate corresponding to different stages of the dis-
charge for the analyzed points on the structure.

The main conclusions of Paper XI are discussed below; the reader is re-
ferred to the publication for the other results.

If the downward leader goes inside the striking distance zone, connection
will take place on the shielding wire, as is claimed by the electro-geometrical
method [62, 63, 70, 71], and as was evident from the reported data of Brown
and Whitehead [66].

72
The 230 kV single circuit structures have a double shielding protection wire.
Nevertheless, from the points at which the conductors are supported, leader
inception and propagation are possible. This result explains why some shiel-
ding failures have been reported for this type of structure [57]. In the double
circuit configurations, upward leader inceptions at points other than the
shielding wire are obtained, but there is no connection to them.

The cone of protection concept ensures that the area inside the cone is pro-
tected against lightning strikes and that the lightning will strike on the shiel-
ding wire before the protected region. However, the cone of protection does
not prevent aborted connected leaders starting from the protected areas.
Therefore, one can conclude that the cone of protection concept was vali-
dated for the simulated conditions and cases.

6.2.2.2 The mesh method

The aim of Paper XII was to review the effectiveness of the mesh size pro-
posed in the standards by means of a procedure that takes into account the
physics of the lightning attachment process.

The ideal protection procedure, according to the principles of electricity, is

to enclose the protected object within a perfectly conducting shield of a
thickness adequate to prevent it from melting. In practice however, it is not
possible to perform such an ideal protection procedure, instead the structure
is surrounded by a conducting mesh. The size of the mesh that should be
used is specified in the standards [62, 63] and depends on the “Lightning
Protection Level” required, the definition of which is based on the maximum
amplitude of the return stroke current that the structure can accept without
causing any damage either to the structure or to the contents of the structure.
The mesh size is determined from the analysis conducted using the rolling
sphere method. The separation between the mesh and the grounded structure
should be larger than a critical value, and this critical value is also given in
the standards and presented in Table 14.

Class of Mesh Size Minimum Current Critical height of mesh

LPS [m] Standards [kA] above the structure [m]
I 5x5 3 0.15
II 10 x 10 5 0.42
III 15 x 15 10 0.63
IV 20 x 20 15 0.84

Table 14. Magnitudes and levels of protection proposed in the standards IEC 62305
and the height of the mesh with respect to the ground level.

73
In the analysis, the total size of the analyzed plane is 100 x 100 m. The mesh
was earthed and elevated above the ground level. The radius of the conduc-
tors of the mesh was assumed to be 2.5 mm as stipulated in the standards. In
the analysis, the vertical axis of the stepped leader is located directly at the
center of a cell, itself located at the center of the mesh. As the stepped leader
approaches the grounded structure, the electric field at the mesh continues to
increase and, when it reaches a critical value, a connecting leader is initiated
from the mesh. In order to check whether the down-coming stepped leader
will get attached to the connecting leader (i.e., the flash is intercepted by the
mesh) or to the ground plane (i.e., the stepped leader penetrates the mesh)
the following criteria are utilized: (a) If the streamers of the connecting lead-
er approach the stepped leader to such a distance where the background elec-
tric field generated by the stepped leader is 500 kV/m, the criterion for the
interception of the stepped leader by the connecting leader is assumed to
have been satisfied. (b) If the electric field in the ground level just below the
stepped leader reaches a value 106 V/m or more, after taking into account the
screening by the mesh, it is assumed to have met the conditions for the
stepped leader to terminate on the ground plane.

Condition (a) assumes that, once the positive streamers of the connecting
leader reach the zone where the electric field generated by the stepped leader
is larger than 500 kV/m, the streamers will propagate continuously until they
meet the streamer sheath of the leader channel, thereby satisfy the final jump
condition. The fulfillment of condition (b) assumes that the negative strea-
mers of the stepped leader will reach the ground plane, and, thereby, they
will meet the final jump condition between the stepped leader and the
ground plane. Depending on which condition materializes first, the stepped
leader will either get attached to the grid or to the ground plane.

The results from the simulations, displayed in Table 15, showed that the
minimum values for the current specified in the standards pertinent to a giv-
en mesh size are in reasonable agreement with the results obtained in this
study using a physically reasonable attachment model. The reason for this
agreement is probably the fact that, in the case of a mesh located close to a
ground plane, the leader has to come into the vicinity of the mesh before a
connecting leader is issued by the mesh. This is so because a mesh placed
rather close to the ground plane does not provide significant field enhance-
ment to promote connecting leaders. As the length of the connecting leader
diminishes, the attachment procedure becomes increasingly close to that
simulated by the rolling sphere method, thereby bringing the results from
both procedures closer to one another.

74
 Mesh Return stroke
Class Result
Size [m] Current [kA]
of LPS
5x5 2 Attachment to ground
I
5x5 3 Attachment to mesh
10x10 3 Attachment to ground
10x10 4 Attachment to ground
Conditions for attachment to the
II 10x10 5 ground and to the grid are fulfilled
almost at the same time
10x10 6 Probable attachment to mesh
10x10 7 Attachment to mesh
15x15 8 Attachment to ground
15 x 15 9 Probable attachment to mesh
III
15 x 15 10 Probable attachment to mesh
15 x 15 11 Attachment to mesh
20 x 20 13 Probable attachment to mesh
IV 20 x 20 14 Probable attachment to mesh
20 x 20 15 Attachment to mesh

Table 15. Results of the simulations. If the electric field at the ground plane is less
that 106 V/m, but greater than 7.5x105 V/m when the criterion for the attachment of
the leader to the mesh is satisfied, the outcome is denoted as probable attachment to
the mesh. In the simulations, the height of the mesh above the ground plane is 0.2 m,
0.5 m, 0.7 m and 0.9 m for LPS classes I, II, III and IV respectively.

6.3 Multiple connecting leaders from a grounded

structure

Simulations can be achieved by making assumptions, or by simplifying the

representation of the physical processes. In the simulations of lightning at-
tachment, one can approximate parameters such as the magnitude of the
minimum or maximum electric field, leader velocities, the electrical charge
required for leader inception, and the geometrical shape of the streamer
zone, among others. However, if one supposes that most of the parameters
assumed are correct, there is still another event that is potentially involved in
the phenomena of lightning attachment that has not been taken into account
in previous models [12, 37], which is the simultaneous inception of several
upward leaders.

The fact is that, when a downward moving leader approaches a given

structure, several connecting leaders could be initiated from several locations
on the same structure. This is the case since, when immersed in the

75
background electric field of the down-coming leader, the electric field at
several points on the structure may exceed the critical values necessary for
leader inception. Consequently, two or more upward leaders can be
generated and propagated from the structure. The initiation and propagation
of several upward leaders from the structure would completely change the
electric field configuration in the vicinity of the structure. This, in turn,
would affect the propagation characteristics of the upward leaders and hence
the lightning attachment process. None of the studies identified in the
literature has analysed the effect of these multiple leaders on the lightning
attachment process. Thus, in Paper XIII, a first approximation of the effect
that two upward leaders being emitted at the same time would have on the
lightning attachment process was presented, in which the simulated cases
assume symmetry for the downward-coming leader. With the aim of
introducing a more realistic condition, an assymetric position of the
downward- coming leader is analyzed in Paper XIV.

For both papers, the applied methodology considers a leader coming straight
down over a structure, as described in Section 6.2.1.1. The structures envi-
saged were rectangular buildings, and details of their dimensions can be
obtained in the respective papers.

The results pertinent to both papers showed that the propagation characteris-
tics of the upward connecting leader are influenced by the presence of other
leaders. They very clearly demonstrate that the moment one connecting
leader starts accelerating towards the downward-coming stepped leader, the
growth of other connecting leaders is arrested. The reason for this is that the
charge deposited in space by the accelerating connecting leader reduces the
electric field in the vicinity of other connecting leaders, thereby impeding
their motion. It is also observed that a slight advantage from the background
electric field is enough for one connecting leader to take over, forcing the
others to abort the attachment process.

The calculations performed in Paper XIV showed the current and velocity
of two upward leaders generated from grounded rods; the location of the
downward-coming leader is asymmetric to the rods, being closer to the rod
denoted by the letter A than to the one called B. Figure 22 displays the cur-
rent associated with the connecting leaders from rods A and B and depicts
the speed of the connecting leaders. Note that the current in the connecting
leader from A grows monotonically until the final jump condition is satis-
fied, whereas the current from B increases initially, but then drops off to zero
as the leader from rod A takes over. A similar tendency can also be seen in
the case of the leaders’ speeds. As one can appreciate that, after several un-
successful attempts (unstable leader inception), the two leaders would start

76
off with similar speeds. However, as the stepped leader comes closer to the
structure, the leader from A starts to grow very rapidly. In coincidence with
this, the growth of the connecting leader from B is arrested, and it is aborted
almost immediately.

Figure 22. Current in the connecting leaders issued from rods A (red) and B (black).
Taken from Paper XIV. Notice that the downward-coming leader is located closer
to rod A than rod B (for further information, see Paper XIV).

77
7 Conclusions

The research work presented in this thesis has been divided into two main
topics. One is numerical simulations of long gap discharges, and the other is
the modelling of the lightning attachment process.

First topic “Numerical simulations of long spark gaps”:

A procedure for the calculation of long gap discharges based on

physical principles was proposed. The numerical methodology was
initiated with simulations of the engineering approximation of Rizk’s
equation for the leader channel calculation and a simplified method
for the computation of the streamer zone. The methodology was
improved to include a variable calculation of the streamer zone and to
use the physical equations of Gallimberti to describe the potential of
the leader channel. Statistical delays for the inception of the first
corona and a random deviation of the path of the discharge based on
experimental measurements for long spark gaps under switching
impulses that were not taken into account in previous methodologies
have been introduced.

All the simulations used to validate the methodology have demon-

strated that, to reproduce the discharges across long gaps under
switching impulses appropriately with different electrode configura-
tions, it is necessary to consider the streamer zone to be a variable re-
gion. Clear evidence of this can be seen in Papers V to VIII. The
model has been validated and compared for different electrode confi-
gurations with successful results, and acceptable errors of less than
10% on average were obtained.

The simulations and studies conducted have shown that, to obtain re-
liable values for the time to breakdown for discharges in long gaps
under switching impulses, it is necessary to include the statistical de-
lays representing the first corona inception and the random path
taken by the leader channel, which depends on the electrode con-
figuration.

78
 It is important to emphasize that inclusion of the statistical represen-
tation of the leader channel mainly affects the magnitude of the time
to breakdown; it does not have a significant effect on the magnitude
of the breakdown voltage. Preliminary simulations using the path of
the discharge obtained from lightning measurements were presented
in Papers II to IV and good results were obtained for the break-
down voltage. However Papers V to VIII use a random normal dis-
tribution derived from experiments conducted under switching im-
pulses in long spark gaps in which a very good accuracy was
achieved with respect to the time to breakdown. In Section 3.2.1.2
different normal distributions were calculated for the path of the dis-
charge for a number of electrode configurations.

The validation of the methodology is presented in Chapter 4, where

recent ultra high voltage experimental tests from different electrode
configurations are presented and very good accuracy in the calcula-
tion of the time to breakdown and the breakdown voltage was ob-
served. The maximum errors obtained for the time to breakdown and
the breakdown voltage were 22 and 10%, respectively.

With the methodology developed to calculate positive discharges

and the new procedure for the calculation of the streamer region, a
consistent numerical simulation taking the physical process underly-
ing the negative leader discharge into account is presented here. It
includes a novel way to calculate the streamer for the pilot streamer
system and enables the complete negative discharge to be predicted.
The experimental results were compared with simulations and
showed good agreement for the representation of the magnitudes of
the current and the path taken by the discharge.

Second topic: “The lightning attachment process”

The method applied to simulate long gap discharges described in

Papers I to IV was extended to the lightning attachment simula-
tions, and the improvements to the previous models are:
o The space charge considered in the advance of each segment
is the new charge that is generated in the region between the
leader channel and the streamer zone, not only the charge
that is produced in the streamer zone.
o The background electric field takes into account the total
effect of the structure. This effect was avoided in previous
methodologies where linear approximations were applied.

79
 o A calculation including the attachment of multiple upward
connecting leaders from grounded structures to a downward
stepped leader was introduced. It was demonstrated that the
effect of one upward leader channel can reduce the electric
field over the other upward leader channel incepted, thereby
changing the attachment process. The results showed that a
small advantage for the growth of a connecting leader from
one point may drastically reduce the ability of other success-
ful connecting leaders to be launched.

The methodology developed was applied to test two lightning pro-

tection methods accepted in the standards: the mesh method and the
cone of protection. The results have shown that the accepted meth-
odologies are in good agreement with the physical methodology,
despite the fact that they are based on engineering approximations.

80
8 Future Work

Two methodologies for the calculation of positive and negative long spark
gap discharges under switching impulses were presented here. One of the
important contributions of the methodologies was to incorporate a variable
streamer region in the calculation, which had not been implemented previ-
ously in any other numerical model. The proposed methodology uses physi-
cal equations for the calculation of the development of the leader channel,
and ensures that the 3D region for the streamer zone has an electric field
equal to or higher than the defined streamer criteria [10]. The methodology
was tested by conducting several experiments and through the use of differ-
ent configurations, such as rod-rod, rod-plane, and sphere-plane configura-
tions, and good results were obtained. Despite this good agreement, the au-
thor is of the opinion that the next step should be to apply the method to
more complicated configurations, such as physical configurations presented
in typical power systems, to determine its accuracy. Additionally, it would
be useful for the calculation of insulation requirements in complicated con-
figurations to analyze several spark gaps simultaneously and compare the
results with experiments.

In Papers VI and VIII the methodology was applied to a rod-rod configura-

tion. The author believes that it would be valuable to combine a self-
consistent model to simulate negative leader discharges with the positive
discharge, thereby representing the case when the two rods are separated a
distance greater than 4m. This situation can frequently arise in ultra high
voltage substations.

A preliminary model was proposed to reproduce the propagation of negative

leaders, however it is the author’s opinion that the negative leader process
should be analyzed in greater detail, especially taking the physics behind the
first pilot leader formation into account, as this has not been clearly under-
stood by the scientific community.

The author considers that the methodology for the calculation of positive
leader discharges should be extrapolated to the lightning attachment process
and implemented in conjunction with the correct calculation of the streamer
zone. The methodology may require a considerable amount of time to run
simulations for a lightning attachment case because of the distances between

81
the downward-coming leader and the upward leader, however, it is still nec-
essary to validate all the assumptions that have been made in the various
models proposed.

Finally, it is the author’s belief that obtaining a complete lightning

attachment methodology is necessary for the downward coming leader to be
modelled properly using the physics of the negative discharge. In the future,
work should be dedicated to combining both negative leader discharge and
positive leader discharge models.

82
Svensk sammanfattning

Den här avhandlingen behandlar två olika fenomen – beräkning av urladd-

ningar vid lång stöt och långt avstånd samt uppåtgående blixtberäkning.
Uppåtgående, positiva stegurladdningar och utvecklingen av urladdningar
över långt avstånd består av två sammanflätade fenomen: leader - kanalen
och koronaytan. Långa avståndstest från laboratorier ligger till grund för de
fysiska beskrivningarna och de föreslagna beräkningarna.

Den här föreslagna metoden använder en geometrisk approximation för att

representera koronaytan. Vidare används och jämförs två olika modeller för
att beskriva leader-kanalen. Beräkningsmetodiken av leader-kanalen bygger
på teknisk approximation kombinerat med en fysikalisk ekvation som tar
med termodynamisk jämvikt. Statistik fördröjning av urladdninginges början
samt en böjning av leader-kanalen infördes för att modellera ett mer realis-
tiskt beteende hos urladdningen. En rak ledarkanal och en krokig jämfördes.
Den fysiska modellen och laboratorietesterna visade god överensstämmelse.

En modell är föreslagen för att simulera negativa urladdningar vid lång stöt,
med hjälp av den metod som utvecklas för att simulera positiva urladdningar
och fysiken baserade på de negativa ledar fenomenen. Utvärderingen av
metoden visade att fenomen som ”pilot leader” och ”negative leader current”
är framgångsrikt bevisat.

En ny metod, baserad på tidigare forskning om blixtar och uppåtgående blix-

tar, har utvecklats och testats. Metoden förfinar tidigare beräkningar; det
elektriska bakgrundsfältet och det joniserade området som beaktas för lea-
der-segmentets progression beräknas med en alternativ metod. Den föreslag-
na metoden användes för att utvärdera två tekniska designmetoder som god-
känts av normerna; ”the mesh method” och ”the electro geometrical
method”. Resultaten visade god överensstämmelse mellan den tekniska upp-
skattningen och den fysiska metoden.

Det finns dock effekter av den uppåtgående blixtprocessen som inte har ta-
gits med eller som har undvikits för att förenkla beräkningen. Faktum är att
jordade konstruktioner i närvaro av ett högelektriskt fält kan producera flera
uppåtgående blixter. Ändå har denna effekt inte tagits med i de olika nume-
riska modeller som finns tillgängliga förrän nu. De publicerade modellerna

83
beaktar varje uppåtgående blixt individuellt. En approximation till processen
att generera flera uppåtgående blixtar presenteras här. Resultaten visar att det
är möjligt att märka ett inflytande på det elektriska bakgrundsfältet när en
uppåtgående blixt utvecklas samtidigt som andra uppåtgående blixtar.

84
Acknowledgements

First, I want to express my deepest thanks to Professor Vernon Cooray for

his guidance during the years my doctoral studies have taken. I am particu-
larly grateful to have had the opportunity and privilege to share scientific
moments with a brilliant scientist. He has shown me that there is always a
very easy way to look at difficult equations and concepts, and he has also
shed light on my work, helping me to find solutions and clarify theories even
when I thought everything was already clear! Additionally, I am very grate-
ful for the advice that has helped me to follow a new path and look for new
challenges in my life as a researcher.

I am sincerely thankful to ABB/HVDC/U Ludvika, and especially to Björn

Jacobson, Dong Wu and Urban Åström. Thanks for each and every opportu-
nity, for giving me the chance to complete my doctoral work in cooperation
with industry and for being able to experience “the real work and world on
the high voltage side”, for all the knowledge imparted, for sharing your ex-
perience with a student, for listening to my presentations and for giving me
your expert opinion and for your input to my work. Thanks to their excellent
guidance, good results have been obtained in this research, which makes me
feel confident to face the upcoming challenges as a researcher in the future.

No words can express my gratitude to Dr. Raul Montaño, not only for his
advice, new ideas and challenges in the scientific area, but also for his help
and valuable friendship.

I would like to thank all my friends and colleagues from the Lightning Re-
search Group, Dr Mahendra Fernando, Azhlinda, Zikri, Prasan, Oscar,
Mona, Dr. Prasana Liyanage, Dr. Mahbuhbur Rahman and all the others who
joined our group during the time I was part of it, for their help, for the eve-
ryday learning experience about lightning or life itself, obtained during
meetings or coffee breaks. I want to thank my long-time friend and trustwor-
thy colleague Oscar Diaz especially: thanks for the time you have dedicated
to understanding my work, to reviewing my writing, to debating about dis-
charges and numerical simulations. Also, I want to express my gratitude to
all my colleagues at the Division of Electricity: to the people from the “Fly-
wheel Group” and the “Diamond Group” for the good dinners, the nice skat-
ing trips and all of the pleasant conversations in the corridors! Special thanks

85
to Thomas Götschl for technical help and to Gunnel Ivarsson and Ingrid
Ringård for the administrative collaboration. The English structural and
grammatical suggestions provided by Suzanne Lidström are deeply appreci-
ated.

I have had the opportunity to meet many wonderful people here in Sweden,
who have made me feel at home and have supported me during the good and
the difficult times, in particular Marianne, Victoria, Sarah and Sra Marianne.
I do not have enough words to say how grateful I am for listening, and for
giving me advice and support.

There are three special people who have helped me to learn a new culture,
language and way of life and I want to thank them for being so open and
nice Tinus, Truls and Troy “Tack för att vara så underbara”. Last, but not
least, I want to express my gratitude for the support, help, patience and for
all the special moments to Magnus Lindström “Tack för att göra allt så
perfekt och fantastiskt till mig”.

I have to express my deep gratitude to the National University of Colombia

for the excellent teaching in the electrical engineering field and on high volt-
age technology. Furthermore, I want to thank my friends from Colombia,
who always wonder how I am doing far away from home; particularly
Francisco Santamaria for always been there, helping, wondering and sup-
porting my work “Gracias pachito!”

Finally, I want to thank my family, the persons who have always stood by
me, helped me and given me everything to make my dreams come true:
Mujita, Mamita, Clarita, Teo, Myriamcito, Guillo, Marthica and Vivis “Alis”
thanks for teaching me and giving me the most important things in my life:
for teaching me to pursue my dreams, I would have never been able to
achieve any of them without you. Every small step I make in my life I give
thanks for having you as my family and as my support. I have no words to
express how much I love all of you, thanks a lot: you are the best of my life.
I want also to express my special thanks to Myriamcito, for reading my
thesis, for the comments and all the effort you have put into my work.

86
References

1. Group, Les Renardiéres. Research on long air gap discharges at Les

Renardieres. Electra 23. 1972, pp. 53–157.
2. —. Research on long air gap discharges—1973 results. Electra 35. 1974,
pp. 47–155.
3. —. Positive discharges in long air gaps—1975 results and conclusions. Electra
53. 1977, pp. 31–132.
4. —. Negative discharges in long air gaps. Electra 74. 1981, pp. 67-216.
5. C. Carrara, L Thione. “Switching surge strength of large air gaps: A physical
approach. IEEE Transaction on Power Systems. March / April 1976, Vols.
Vol PAS – 95, N 2, pp. pp 512 – 524.
6. Waters, N. I. Petrov and R. T. Determination of the striking distance of
lightning to earthed structures. Proc. R. Soc. A 450. 1995, pp. 589–601.
7. Rizk, F.A.M. Switching impulse strength of air insulation: Leader inception
criterion. IEEE Transaction on Power Delivery. October 1989, Vol. 4, 4,
pp. 2187 - 2195.
8. —. A model for switching impulse leader inception and breakdown of long air
gaps. IEEE Transaction on Power Delivery. January 1989, Vol. 4, 1,
pp. 596 - 606.
9. —. Modelling of transmission line exposure to direct lightning strokes. IEEE
Transaction on Power Delivery. November 1990, Vol. 5, 4, pp. 1983 - 1997.
10. Gallimberti, I. The mechanism of long spark formation. J. Physique Coll. 1972,
Vol. Suppl. 7 40, C7, pp. 193-250.
11. A. Bondiou, I. Gallimberti. Theoretical modeling of the development of the
positive spark in long gaps. Journal of Physics. D: Applied Physics. 1994,
Vol. 27, pp. 1252-1266.
12. Lalande, P. “Study of the lightning stroke conditions on a grounded structure,”
Doctoral Thesis". s.l. : Office National d’Etudes et de Recherches
A’erospatiales (ONERA), 1996.
13. N. Goelian, P. Lalande, A. Bondiou-Clergerie, G. L. Bacchiega, A. Gazzani,
I. Gallimberti. A simplified model for the simulation of positive-spark
development in long air gaps. Journal of Physics. D: Applied Physics. 1997,
30, pp. 2441-2452.
14. A. Bondio, P. Lalande, P. Laroche P, J. C. Willet, D. Davis, I. Gallimberti.
The inception phase of positive leaders in triggered light-ning: comparison
of modeling with experimental data. ICAE 99: 11th Int. Conf. on
Atmospheric Electricity. June 1999.
15. Becerra, M and Cooray, V. A self-consistent upward leader propagation
model. Journal of Physics. D: Applied Physics. 2006, Vol. 39, pp. 3708–
3715.
16. Baldo G., Gallimberti I. Breakdown phenomena of long gaps under switching
impulses conditions influence of distance and voltage level. IEEE

87
 Transactions on Power Apparatus and Systems. July - August 1975, Vols.
PAS - 94, 4.
17. Gallimberti I, Bacchiega G, Bondiou-Clergerie A, Lalande P. Fundamental
processes in long air gap discharges. C.R. Physique appliquée/ Applied
physics. 2002, 3, pp. 1335 - 1359.
18. Golde, R. Lightning protection. Edward Arnold. 1973.
19. —. The lightning conductor. Journal of the Franklin Institute. 1967, Vol. 283, 6,
pp. 451-477.
20. Lemke, E. “Beitrag zur abschatzung der durchschlagspannung langer
luftfunkenstrecken”. Z. Elecktr. Inform u. Engietechnik. 1973, Vol. 3, 4,
pp. 186 – 192.
21. Aleksandrov, G. N. Special features of spark discharge development in long air
clearences of insulating structures. Elektrichestvo. 1975, N 8, pp. pp. 15 –
18.
22. B, Jones. “Switching surges and air insulation”. Phil Trans R Soc. London A.
1973, Vol. Vol 275.
23. Hutzler. “Leader propagation for determination of switching surge flashover
voltage of large air gaps”. IEEE Transaction on Power Systems Delivery.
July/Aug 1978, Vols. Vol. PAS-97, N4, pp. pp. 1087 – 1096.
24. Bazelyan, E. “The leader of a long positive spark”. Electric Technology USSR.
1987, N2, pp. pp. 47 - 60.
25. Arevalo, Liliana. Numerical simulations of long spark gaps-Lightning
attachment and its applications. Uppsala : Uppsala Universitet,
2010. 0349-8352.
26. Rizk, F.A.M. Critical switching impulse strength of long air gaps: Modelling of
air effects. IEEE Transactions Paper Summer Power Meeting. July/August
1991.
27. L. Dellera, E. Garbagnati. Lightning strike simulation by means of the leader
progression model, Part I: Description of the model and evaluation of free-
standing structures. IEEE Transaction on Power Delivery. 1990, Vols.
PWRD-5, pp. 2009-2023.
28. —. Lightning strike simulation by means of the leader progression model: II.
Exposure and shielding failure evaluation of overhead lines with assessment
of application graphs. IEEE Transactions on Power Delivery. 1990, Vols.
PWRD-5, pp. 2023–2029.
29. Eriksson, A. The lightning ground flash –an engineering study. Pretoria : Ph.D
thesis, Faculty of Engineering, University of Natal, 1979.
30. —. An improved electrogeometric model for transmission line shielding
analysis. IEEE Transaction on Power Delivery. July 1987, Vols. PWRD-2,
N 3, pp. 871 – 886.
31. —. The incidence of lightning strikes to power lines. IEEE Transaction on
Power Delivery. 1987, Vols. PWDR-2, pp. 859-870.
32. D’Alessandro, F. The use of “Field intensification factors” in calculations for
lightning protection of structures. Journal of Electrostatics. 2003, Vol. 58,
pp. 17-43.
33. Akyuz Mose, Cooray Vernon. The franklin lightning conductor: conditions
necessary for the initiation of a connecting leader. Journal of Electrostatics.
2001, 51-52, pp. 319-325.
34. D’Alessandro, N. I. Petrov and F. Theoretical analysis of the processes
involved in lightning attachment to earthed Structures. Journal of Physics.
D: Applied Physics. 2002, Vol. 35, pp. 1788–95.

88
35. A, Castellani. “Calcul du champ électrique par la méthode des charges
equivalentes pour la simulation d’une décharge bi-leader” Doctoral thesis.
Orsay : Université Paris XI, 1995.
36. Becerra M, Cooray V. “A simplified physical model to determine the lightning
upward connecting leader inception”. IEEE Transactions on Power Delivery.
April 2006, Vol. 21, 2, pp. 897 – 908.
37. Becerra, M. On the attachment of lightning flashes to grounded structures.
Doctoral Thesis. 2008.
38. Amarasinghe D, Sonnadara U, Berg M and Cooray V. Channel tortuosity of
long laboratory sparks. Journal of Electrostatics. 2007, Vol. 65, pp. 521-526.
39. E, Idone V and Orville. Channel tortuosity variation in Florida triggered
lightning. Geophysical Research Letters. July 1988, Vol. 15, 7, pp. 645 -
648.
40. L, Paris. “Influence of air gap characteristics on line-to-ground switching surge
strength”. IEEE Transaction on Power Apparatus and Systems. August
1967, Vol. PAS 86, No. 8.
41. Paris L, Cortina R. “Switching and lightning impulse discharge characteristics
of large air gaps and long insulator strings”. IEEE Transaction on Power
Apparatus and Systems. April 1968, Vols. PAS – 87, 4.
42. Veldhuizen, van. Electrical discharges for environmental purposes:
fundamentals and applications. s.l. : New York: Nova Science, 2000. ISBN
1-56072-743-8.
43. Yu, Dyakonov M I and Kachorovskii V. Zh. Eksp. Teor. Fiz. Soviet Physics
JETP. 1988, Vol. 67, 1049.
44. ABB PS/HVDC Ludvika. Switching impulse tests for different electrode gap
configurations. Ludvika : s.n., 2006.
45. Ming Li, Wu Dong, Åström Urban and Asplund Gunnar. Intriguing
observation on the breakdown trajectory of large air-gaps under switching
impulse voltages. 16th International symposium on high voltage
engineering. 2009.
46. ABB PS/HVDC R&D. Experimental investigations of air insulation. Ludvika :
ABB, 2010.
47. Commission, IEC International Electrotechnical. IEC 60060-1 High voltage
test techniques - Part 1 General definitions and test requirements. s.l. : IEC,
2010. ISBN 978-2-88912-185-4.
48. ABB PS/HVDC Ludvika Wu Dong, Ming Li, Åström U, Ericsson A.
Experimental investigation of air insulation under positive switching impulse
voltage for HVDC System (Part 1. Sphere/plane air gaps). ABB Internal
confidential report. Ludvika : s.n., 2006.
49. Castellani A., Bondiou A, Lalande P, Bonamy A and Gallimberti I.
Laboratory study of the bi-leader process from an electrically floating
conductor Part 2: Bi-leader properties. IEE Proc. Sci Meas Technol.
September 1998, Vol. 145, 5.
50. Lalande P., Bondiou-Clergerie A, Bacchiega G., Gallimberti I. Observations
and modeling of lightning leaders. C.R. Physique. 2002, Vol. 3,
pp. 1375-1392.
51. Dellera L, Garbagnati E. Lightning stroke simulation by means of the leader
progression model Part I: Description of the model and evaluation of free-
standing structures. IEEE Transaction on Power Delivery. 1990, Vols.
PWRD-5, Issue 4, pp. 2009-22.
52. Bacchiega G, Gazzani A, Bernardi M, Gallimberti I, Bondiou A. Theoretical
modelling of the laboratory negative stepped leader. Proceedings of the 1994

89
 international Aerospace and Ground conference on Lightning and Static
Electricity. 1994.
53. Mazur V, Ruhnke L, Boundiou-Clergerie and Lalande P. Computer
simulation of a downward negative stepped leader and its interaction with
ground structures. Journal of Geophysical Research. September 2000, Vol.
Vol 105, No. D17, pp. page 361 – 369.
54. Niemeyer L, Pictronero L, Wiesmann H. Fractal dimension of dielectric
breakdown. Phys. Rev. Lett. 1984, 52, pp. 1033-1036.
55. J, Tsonis A and Eisner. Fractal characterization and simulation of lightning.
Beitr. Phys Atmos. 1987, Vol. 60, (2), pp. 187-192.
56. Sanudo J, Gomez J, Castaño F and Pacheco A. Fractal dimensions of
lightning discharge. Non-linear Process in Geophysics. 1995, Vol. 2,
pp. 101 - 106.
57. G, Petrova. Lightning stroke simulation by means of the fractal approach:
Attractive and protective zones for structures. 24th International Conference
on Lightning Protection. 1999.
58. Beroual A, Rakotonandrasana J, Fofana I. Predictive dynamic model of the
negative lightning discharge based on similarity with long laboratory sparks
– part 1: physical process and modeling. IEEE Transactions on Dielectrics
and Electrical Insulation. October 2010, Vol. 17, 5, pp. 1551 – 1561.
59. —. Predictive dynamic model of the negative lightning discharge based on
similarity with long laboratory sparks – part 2: Validation. IEEE
Transactions on Dielectrics and Electrical Insulation Vol 17, N 5, 1562 –
1568. 2010.
60. Arevalo L, Cooray V and Montano R. Numerical simulation of long
laboratory sparks generated by positive switching Impulses. Journal of
Electrostatics. May 2009, Vol. Vol 67, Issue 2 – 3.
61. Arevalo L, Cooray V and Wu D. Laboratory long gaps simulation considering
a variable corona region. International Conference on Lightning Protection.
ICLP – 2010. 2010.
62. IEC. International standards on protection against lightning Part 1. General
principles. IEC 62305 -1. Geneva, Switzerland : s.n., 2006.
63. —. International standards on protection against lightning Part 2. Physical
damage to structures and life hazards. IEC 62305 -3. Geneva, Switzerland :
s.n., 2006.
64. H. R. Armstrong, E. R. Whitehead. A lightning stroke pathfinder. IEEE Trans.
Power Apparatus and systems. 1964, Vol. 83, pp. 1223-1227.
65. —. Field and analytical studies of transmission line shielding. IEEE Trans.
Power Apparatus and systems. 1968, Vol. 87, pp. 270-279.
66. G. W. Brown, E. R. Whitehead. Field and analytical studies of transmission
line shielding: Part II. IEEE. Trans. Power Apparatus and Systems. 1969,
Vol. 88, 5, pp. 617-625.
67. D. W. Gilman, E. R. Whitehead. The mechanism of lightning flashover on
high voltage and extra-high voltage transmission lines. Electra. 1973, Vol.
27, pp. 65-96.
68. C. F. Wagner, G. D. McCann, G. D. and G. L. MacLane. Shielding of
transmission lines. AIEE Transactions. 1941, Vol. 60, pp. 313-328.
69. C. F. Wagner, G. D. McCann, C. M. Lear. Shielding of substations. AIEE
Transactions. 1942, Vol. 61, pp. 96-100.
70. Golde, R. The frequency of occurrence and the distribution of lightning flashes
to transmission lines. Transactions AIEE. 1945, Vol. 64, pp. 902 - 910.

90
71. Wagner C, Hileman R. “The lightning strike” II. Trans. AIEE. 1961, Vol. 80,
pp. 622 – 642.
72. Young, Clayton, Hileman. Shielding of transmission lines. IEEE Transaction
on Power and Apparatus and Systems (supplement). 1963, Vol. 83, pp. 132 -
154.
73. Flisowski, Z. Trends in the lightning protection of structures. PWN. 1986.
74. Horvarth, T. Computation of lightning protection electronic and electrical
engineering research studies. New York : Series Wiley, 1991.
75. IEEE, The Institute of Electrical and Electronics Engineers. Std 998-1996,
IEEE Guide for direct lightning stroke shielding of substations. 1996. ISBN
1-55937-768-2.
76. Becerra, M, Cooray, V and Hartono, A. Identification of the lightning
vulnerable points on complex grounded structures. Journal of Electrostatics.
2007.
77. Becerra M, Cooray V, Roman F. Lightning striking distance of complex
structures. IET Generation Transmission and Distribution. 2008, Vol. 2,
pp. 131 - 138.
78. Cooray V., Rakov V., Theethayi N. The lightning striking distance - Revisited.
Journal of Electrostatics. 2007, Vol. 65, pp. 296 – 306.
79. EPRI. Transmission line reference book 345kV and above. s.l. : EPRI,
1982. 2nd edition.

91