Switching Arcs
Switching Arcs
Switching Arcs
21548
SWITCHING ARCS IN
PASSIVE RESONANCE HVDC
CIRCUIT BREAKERS
A thesis submitted to attain the degree of
DOCTOR OF SCIENCE of ETH ZURICH
(Dr. sc. ETH Zurich)
presented by
Michael Max Walter
MSc ETH Zurich
born on 31.03.1984
citizen of Zurich (ZH)
accepted on the recommendation of
Prof. Dr. Christian M. Franck
Prof. Dr. Volker Hinrichsen
2013
Abstract
This work aims to systematically and accurately investigate switching arc characteristics in passive resonance high voltage direct current (HVDC) circuit breakers.
The replacement of classical band energy with uctuating wind
and solar power from peripheral locations in Europe will signicantly challenge the European transmission grid in the future. Energy must be transmitted via cables for distances > 50 km and via
overhead lines for distances > 1000 km with low losses. VoltageSourced-Converter (VSC) HVDC is considered to be superior to the
classical alternating current (AC) transmission for long distance energy transfer, because it has signicantly smaller losses and requires
no reactive power compensation.
Mechanical circuit breakers are standard fault protection devices
in AC networks but do not yet exist for HVDC with sucient interruption performance. The current zero crossing, essential for arc
extinction in mechanical breakers, is not inherently available in DC
systems. This makes DC interruption more challenging than AC
interruption. Passive resonance breakers excite an unstable current
oscillation and create articial current zero crossings by interaction
of the switching arc with an LC-commutation circuit. This principle has been successfully applied for Metal-Return-Transfer-Breakers
(MRTB) in operation. It is, however, limited in maximum interruptible current, takes too long for current zero creation and uses a large
capacitor, which signicantly contributes to the breaker costs.
Signicant improvement of the interruption performance is expected if the arc chamber and nozzle design is optimized for passive
resonant creation of current zero crossings in DC circuits. For this,
the improvement of the switching arc characteristic is shown to be
iii
iv
more eective and most probably also more economical compared to
the passive L and C components of the resonance path. The main
goal of this thesis is a systematic characterization of dierent arc
congurations for use in black-box simulations.
For that purpose, a novel arbitrary pulsed current source has
been developed. By creation of complex current waveforms, (eg.
staircase-like currents and spikes superimposed on a current slope,)
the transient and stationary arc characteristic can be measured independently of each other. Thereby, a more accurate parameter determination and a better validation of black-box models is achieved.
In principle, the source could also be used to characterize the arc
completely model-independent, by generation of step currents with
variable slope steepness. Furthermore, a novel improved method
for arc characterization has been developed. A exible model circuit breaker has been used to investigate the eect of blow pressure,
nozzle geometry, nozzle material and blow gas type.
The investigations conrmed that a falling stationary U I-characteristic with decreasing arc voltage at increasing current is a necessary condition for creation of passive resonance. A rising characteristic, the arc thermal inertia and high current gradients act as damping
terms and inhibit passive resonance. For current amplitudes < 2 kA
the following has been shown: a) an increase of blow pressure intensies the falling slope of the stationary U I-characteristic and improves
passive resonance, b) a narrow throat diameter and a large nozzle
throat length exhibit a rising U I-characteristic and should therefore
be avoided, c) nozzle material and blow gas type have shown only
minor inuence on the arc characteristics.
The results gained can be used to improve an existing HVDC
MRTB in two ways: Firstly, if the time for current zero creation (and
with it the total break time) is not important, the switching arc characteristic could be inuenced so that a) the size of the capacitance
is minimized or b) the interruption current amplitude is maximized.
Application for this are MRTB or HVDC load break switches. Secondly, the time for current zero creation can be minimized by suit-
v
able arc chamber design changes with consequent changes in the arc
characteristic. By this, together with an increase in maximum interruption current, existing MRTBs could be improved for use as
HVDC circuit breakers. Such a passive resonance HVDC circuit
breaker would be a low loss and low cost alternative to the recently
proposed hybrid breakers, which use expensive and inherently lossy
solid state components.
Zusammenfassung
Das Ziel dieser Arbeit ist die systematische und przise Untersuchung von Schaltlichtbogencharakteristiken in Hochspannungs-Gleichstrom (HG) Leistungsschaltern.
Der Ersatz klassischer Bandenergie durch uktuierende Wind- und
Sonnenenergie aus peripheren Standorten Europas stellt das Europische bertragungsnetz zuknftig vor grosse Herausforderungen.
Energie muss in O-shore-Kabeln ber Distanzen > 50 km und in
bertragungsleitungen ber Distanzen > 1000 km verlustarm zum
Verbraucher geleitet werden. Hochspannungs-Gleichstrombertragung
(HG) wird gegenber der klassischen Wechselstrom (AC)-bertragung bevorzugt, aufgrund der markant kleineren bertragungsverluste bei grossen bertragungsdistanzen und aufgrund der Blindleistungskompensation, welche bei ersterer entfllt.
Mechanische Leistungsschalter zur Unterbrechung von Fehlerstrmen sind Standardkomponenten in AC-Netzen, existieren fr HG
jedoch noch nicht mit ausreichender Kurzschlussunterbrechungsleistung. Die Schwierigkeit der HG Stromunterbrechung ist der fehlende Stromnulldurchgang, welcher in mechanischen Schaltern unerlsslich ist zum Lschen des Lichtbogens. Passiv resonante Schalter
erzeugen durch Wechselwirkung eines Schaltlichtbogens mit einem
LC-Kommutierkreis eine instabile Stromschwingung zur Erzeugung
eines Stromnullurchgangs. Das Prinzip wurde erfolgreich in MetalReturn-Transfer-Breakern (MRTB) umgesetzt, ist jedoch limitiert in
Bezug auf den maximal unterbrechbaren Strom, die Schalter bentigen zu lange zur Erzeugung eines Stromnulldurchgangs und verwenden einen grossen Kondensator, welcher wesentlich zu den Schalterkosten beitrgt.
Eine wesentliche Verbesserung des Unterbrechungsverhaltens sol-
vii
viii
cher Schalter ist zu erwarten, falls die Lichtbogenkammer und die
Dsengeometrie optimiert wird zur passiv resonanten Erzeugung von
Stromnulldurchgngen. Es wird aufgezeigt, dass die Verbesserung
der Lichtbogencharakteristik ezienter und hchst wahrscheinlich
auch konomischer ist verglichen mit der Optimierung der passiven
L und C Elemente des Resonanzkreises. Das Ziel dieser Arbeit ist
das systematische Charakterisieren von verschiedenen Lichtbogenkongurationen zur Verwendung in Black-Box Modellen.
Zu diesem Zweck wurde eine neuartige Pulsstromquelle entwickelt. Durch Erzeugung komplexer Stromformen (z.B. treppenfrmig
ansteigende Strme und Stromspitzen, die einer achen Stromanke
berlagert sind) kann das transiente und das stationre Lichtbogenverhalten unabhngig voneinander gemessen werden. Dies hat den
Vorteil einer genaueren Parameterbestimmung und einer besseren
Validierung von Black-Box-Modellannahmen. Prinzipiell wre auch
eine komplett modell-unabhngige Charakterisierung des Lichtbogens mglich durch die Verwendung von Schritten mit unterschiedlicher Flanken Steilheit. Es wurde zudem eine neue verbesserte Methode zur Bestimmung von Black-Box-Parametern entwickelt. Ein exibler Modell-Leistungsschalter wurde verwendet, um den Einuss von
Beblasungsdruck, Dsengeometrie, Dsenmaterial und Beblasungsgas zu untersuchen.
Die Untersuchungen haben besttigt, dass eine stationre U ILichtbogencharakteristik mit fallender Spannung bei steigendem Strom
notwendig ist zur Erzeugung von passiver Resonanz. Eine ansteigende Charakteristik, die Lichtbogentrgheit sowie steile Stromgradienten dmpfen hingegen das System und verhindern passive Resonanz.
Es konnte fr Strme 2.5 kA folgendes gezeigt werden: a) Ein Erhhen des Beblasungsdrucks verstrkt die passive Resonanz durch
Erhhen von Steilheit und Spannung der fallenden Charakteristik,
b) ein enger Durchmesser oder eine grosse Lnge des engen Bereichs
der Dse hingegen fhren zu einer ansteigenden Charakteristik und
sollten daher vermieden werden, c) verschiedene Dsenmaterialien
und Beblasungsgase zeigten nur kleinen Einuss auf die Charakte-
ix
ristik.
Die gewonnenen Erkenntnisse knnen auf zwei Arten zur Verbesserung eines bestehenden HG-MRTB Schalters verwendet werden: 1)
falls die Zeit zur Stromnulldurchgangserzeugung (bzw. die gesamte
Unterbrechungszeit) keine Rolle spielen, kann die Schaltlichtbogencharakteristik beeinusst werden, um a) die Grsse des Kondensators
zu minimieren oder b) den unterbrechbaren Strom zu maximieren.
Anwendungen dafr sind MRTB-Schalter oder HG-Lastschalter.
2) Die Zeit zur Erzeugung des Stromnulldurchgangs kann durch Anpassungen in der Lichtbogenkammer minimiert werden, welche sich
auf die Lichtbogencharakteristik auswirken. Dadurch, und durch Erhhen des Unterbrechungsstroms, knnen existierende MRTBs zur
Anwendung als HG-Leistungsschalter verbessert werden. Solch ein
passiv resonanter HG-Leistungsschalter wre eine verlustarme und
preislich attraktive Alternative zu den krzlich vorgeschlagenen Hybridschaltern, welche teure und verlustreiche leistungselektronische
Komponenten verwenden.
Acknowledgement
At the start of this Ph.D. project, Prof Christian Franck told the
following to me: The times, when a scientic breakthrough could
be achieved by a single genius person, working alone in his chamber,
are long gone. Today, successful science is performed in a team
of researches, working together and exchanging their ideas. This
sentence illustratively expresses that this work would not have been
possible without the support of many people.
I would like to thank Christian for the trust he set in me, to
conde me with one of his key projects. The interesting eld kept
me continuously fascinated and motivated for the last three and a
half years. In the daily work, Christian gave me the freedom to
develop my own approaches, supported me when I needed guidance
and always advised me with his expert knowledge. I believe that it
is a rare skill, to nd such a balance in leading his employees.
Warmly, I would like to thank Markus Bujotzek, Riccardo Bini
and Manolis Panousis from the ABB Corporate Research. Their
experience in circuit breakers research and valuable inputs in discussions, helped to set the right focus in the research work.
Carsten Leu from TU Illmenau was a very welcome discussion
partner during the work on a joint paper. At TU Illmenau, Carsten
applied for the rst time the Drebenstedt method on passive resonant
currents and is therefore very experienced in the eld of experimental arc parameter determination. Carsten contributed the literature
research on the classical parameter determination methods to our
common paper, which also served as a basis for the corresponding
theory section in this thesis.
Thanks to Dr. Timm Teich for careful proof reading of the manuscript
and for his valuable inputs.
xi
xii
Several students worked in the scope of their master and semester
projects on integral parts of the experimental setup. These were:
Michael Leibl, Daniel Rothmund and Andreas Ritter in their semester
project and Andreas Ritter, Moonjo Kang and David Gassner in
their master project. Michael Leibl supported the set up and execution of the rst measurements series on free burning arcs. Daniel
Rothmund built the setup for the second measurement series with
sinusoidal currents and performed experiments on free burning and
radially blown arcs. He also assembled and characterized a exible
coil for the arbitrary current source. Andreas Ritter assembled the
current source modules, programmed the code for the measurement
automation, based on a code structure from Dominik Dahl, and validated the IGBT-controller performance in an intensive measurement
study. Moonjo Kang and David Gassner assembled the model circuit
breaker and they performed a measurement series on blow-pressure
eects and contact movement eects in convection stabilized arcs.
The IGBT controller, a key component for the arbitrary current
source, has been developed by WEMEL GmbH. In particular, I
would like to thank Martin Weidmann from WEMEL for his excellent cooperation in the development phase. At the stage of controller
development it has been uncertain, how the arc would behave as a
load. With his foresight, Martin identied and avoided numerous
technical challenges already in a very early planning phase and was
always eager to maintain maximum possible exibility of the controller hard- and software. This exibility signicantly facilitated
later optimization of the controller for the investigated arc.
A major support in the whole three years was the technical sta
in our team: Thanks to Hans-Jrg Weber, Henry Kienast, David
Brhlmann, Claudia Stucki and Karin Sonderegger.
The mutual discussion with all colleagues in the High Voltage
Group was an important pillar in the daily work. Thanks in particular to my oce mates Ueli Straumann, Philipp Simka, Michael
Strobach, Andreas Ritter, Moonjo Kang and Pascal Higer for the
enriching time and mutual support. Especially, I would like to men-
xiii
tion Philipp Simka, who encouraged me to start as a Ph.D. student
in Christians group.
I would also like to express my gratefulness to my parents Erika
and Hansruedi Walter, who always supported me on my way and do
that still.
Finally, I would like to thank my wife Ivana. She always encourages me, in what I am doing and supports me in so many dierent
ways. She is very understanding for the numerous evenings where I
was physically absent, trying to nish some work and for the even
more numerous times when I was physically present but absent in
my thoughts, trying to solve some problem. Thanks to her I found
enough sleep at night after the birth of our son Luka in April to
complete this work.
This work was nancially supported by ABB Switzerland, Corporate Research.
xv
xvi
[P6] M. Walter, C. Carstensen, J. Biela, C. M. Franck, A new
Method for Investigating the Arc Behavior in HVDC Circuit
Breakers based on a novel Current Source, Cigr Session 2012,
Paris, France, 26 - 31 August 2012.
[P7] M. K. Bucher, M. M. Walter, M. Pfeier, C. M. Franck, Options for ground fault clearance in HVDC oshore networks,
IEEE Conference: Energy Conversion Congress and Exposition (ECCE), Rayleigh NC, September 2012.
[P8] M. M. Walter, C. M. Franck, Inuence of arc chamber parameters on passive resonance circuit of HVDC circuit breakers,
Cigr International Symposium THE ELECTRIC POWER
SYSTEM OF THE FUTURE, Bologna, Italy, 13th - 15th
September 2011.
[P9] M. M. Walter, C. M. Franck, Flexible pulsed DC-source for
investigation of HVDC circuit breaker arc resistance, XVIII
International Conference on Gas Discharges and Their Applications (GD 2010), Greifswald, Germany, 5th - 10th September
2010.
xvii
The following master students have made contributions to this
project in the scope of their master and semester projects.
[S1] Author: A. Ritter, Supervisors: M. M. Walter, M. K. Bucher,
Disconnector and earthing switch operation in HVDC GIS,
master project, ETH Zurich, 2012.
[S2] Author: M. Kang, Supervisor: M. M. Walter, Inuence of
blow pressure and nozzle shape on HVDC circuit breakers,
master project, ETH Zurich, 2012.
[S3] Author: D. Gassner, Supervisor: M. M. Walter, Experimental
study of high current arcs during contact separation, master
project, ETH Zurich, 2012.
[S4] Author: A. Ritter, Supervisor: M. M. Walter, Aufbau, Inbetriebnahme und Automatisierung einer gepulsten DC Quelle,
semester project, ETH Zurich, 2011.
[S5] Author: D. Rothmund, Supervisor: M. M. Walter, Zeitgenaues Znden und Lschen einer Funkenstrecke fr einen Teslatransformator, semester project, ETH Zurich, 2011.
[S6] Author: M. Leibl, Supervisor: M. M. Walter, Bestimmung
der Kennlinie und Parameterfunktionen eines Lichtbogens zur
Beobachtung von Einussfaktoren, semester project, ETH Zurich,
2010.
Contents
Abstract
iii
Zusammenfassung
vii
Acknowledgements
xi
xv
Table of content
xix
List of symbols
xxv
1. Problem statement
3
3
5
10
13
13
14
20
22
23
4. Switching arcs
4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . .
4.1.1. Classes of arc models . . . . . . . . . . . . . .
25
25
25
xix
xx
Contents
4.1.2. Energy transfer mechanisms . . . . . . . . .
4.2. Nozzle constricted arcs under forced convection . .
4.2.1. Convection stabilized low current arcs . . .
4.2.2. Convection stabilized high current arcs . . .
4.2.3. Ablative arcs . . . . . . . . . . . . . . . . .
4.2.4. Characteristic arc cross-sections . . . . . . .
4.3. Black-box arc modeling . . . . . . . . . . . . . . .
4.3.1. Introduction . . . . . . . . . . . . . . . . . .
4.3.2. Generalized Mayrs equation . . . . . . . . .
4.3.3. Physical interpretation of black-box models
4.3.4. Transformation of stationary characteristic
4.4. Black-box parameter determination . . . . . . . . .
4.4.1. Intrinsic assumptions . . . . . . . . . . . . .
4.4.2. Stationary and transient arc characteristic .
4.4.3. Classical methods . . . . . . . . . . . . . . .
4.4.4. Comparison . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
31
32
34
35
36
39
39
40
41
42
43
43
44
46
49
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
mechanisms
. . . . . . .
. . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
53
53
57
58
61
63
65
68
6. Experimental methods
6.1. Arbitrary current source . . . . . . . . . . . .
6.1.1. Functionality . . . . . . . . . . . . . .
6.1.2. Creation of selected current waveforms
6.2. LC current source . . . . . . . . . . . . . . . .
6.3. Model circuit breaker . . . . . . . . . . . . . .
6.3.1. Functionality . . . . . . . . . . . . . .
6.3.2. Nozzle geometry . . . . . . . . . . . .
6.4. Free burning arc arrangement . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
69
69
69
73
77
78
78
79
79
Contents
6.5. Diagnostic tools . . . . . . . . . . . . . . . . .
6.5.1. Current measurement . . . . . . . . .
6.5.2. Voltage measurement . . . . . . . . . .
6.5.3. Pressure measurement . . . . . . . . .
6.5.4. Optical measurement . . . . . . . . . .
6.6. Arc cross-section determination . . . . . . . .
6.6.1. AV determination from arc images . .
6.6.2. AP determination from pressure signal
6.7. Typical experiment sequence . . . . . . . . .
xxi
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
81
81
81
81
82
82
82
83
85
.
.
.
.
.
.
.
.
.
105
105
106
110
115
118
119
119
128
131
xxii
Contents
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
133
133
133
136
139
139
141
143
145
148
148
152
10.Discussion
10.1. Blow pressure . . . . . . . . . . . . . . . . . . .
10.2. Nozzle geometry . . . . . . . . . . . . . . . . .
10.3. Blow gas parameters . . . . . . . . . . . . . . .
10.4. Nozzle material . . . . . . . . . . . . . . . . . .
10.5. Desirable and undesirable loss mechanisms . . .
10.6. Improvement by adapting the arc chamber . . .
10.6.1. Resonance capacitor reduction for lower
MRTBs . . . . . . . . . . . . . . . . . .
155
155
158
161
162
163
164
. . .
. . .
. . .
. . .
. . .
. . .
cost
. . . 164
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12.Outlook
12.1. Further experiments . . . . . . . . . . . . . . . . . .
12.1.1. Further phenomenological characterization . .
12.1.2. Arc characterization and understanding cooling mechanisms . . . . . . . . . . . . . . . . .
171
172
174
175
175
177
179
179
179
180
Contents
xxiii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
181
182
182
182
183
183
184
184
184
185
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
187
187
189
189
193
195
195
196
199
199
203
213
Curriculum Vitae
225
List of symbols
symbol
A
AE
unit
m2
m2
Ae
m2
AN
AP
m2
m2
AV
AT
apg
m2
m2
api
aui
bui
bpg
bpi
description
arc cross-section
arc cross-section (determined from electrical parameters)
eective cross-section of cold gas ow for a nozzle throat
with arc
nozzle throat cross-section
arc cross-section (determined from uid dynamic parameters)
arc cross-section (determined from optical parameters)
arc cross-section (theoretical prediction)
t coecient of power law for stationary arc power loss
P = 1W apg (g/(1 S))bpg
t coecient of power law for stationary arc power loss
P = 1W api (I/(1 A))bpi
t coecient of power law for stationary arc power loss
U = 1V aui (I/(1 A))bui
t coecient of power law for stationary arc power loss
U = 1V aui (I/(1 A))bui
t coecient of power law for stationary arc power loss
P = 1W apg (g/(1 S))bpg
t coecient of power law for stationary arc power loss
P = 1W api (I/(1 A))bpi
xxv
xxvi
symbol
C
Cp
c
Contents
unit
F
J/(m3 K)
s
d
d
dT
E
U/I
m
V/m
V/A
u/i
V/A
E
f
f
g
g
V/m
Hz
hp
b
h
J/kg
J/kg
I
IF
IT
A
A
A
i
iF
j
LT
m
S
m
sub
A
A
A/m2
m
kg/(m s)
kg/(m s)
S
S/s
description
capacitance
heat capacity at constant pressure
t coecient of power law for thermal arc inertia
= c(g/(1 S))d
t coecient of power law for thermal arc inertia
= c(g/(1 S))d
duty cycle
nozzle throat diameter (cylindrical section)
Electric eld
derivative of the stationary (non time dependent)
U I arc characteristic after the current
derivative of the transient (time dependent) ui arc
characteristic after the current
electric eld
frequency
fraction of of power loss by radiation
conductance
temporal derivative of a transient conductance signal
enthalpy of plasma
enthalpy for wall material evaporation and acceleration to sonic velocity
steady state current
steady state fault current
current amplitude with minimum steady state arc
voltage in the U (I) arc characteristic
time variant current
transient fault current
current density
nozzle throat length (cylindrical section)
supersonic mass ow rate
subsonic mass ow rate
Contents
symbol
P
unit
W
Pcool
Pheat
Pjoule
PH
p
pB
W
W
W
W
bar
bar
pdn
pH
bar
bar
pS
pup
pR
Q
R
R1
bar
bar
bar
J
J/(mol K)
R2
r
U
UT
UN
u
VB
VH
vs
vz
m
V
V
W/m2
V
m3
m3
m/s
m/s
xxvii
description
steady state cooling power of an arc (the parameter
is typically assumed to be a function of conductance)
instantaneous power removed from the arc column
instantaneous power added to the arc column
joule heating power
steady state cooling power of a hypothetical arc
absolute pressure
overpressure in pressure bottle (relative to pR =
1013 mbar atmospheric pressure)
absolute downstream pressure
overpressure in pressure chamber (relative to pR =
1013 mbar atmospheric pressure)
absolute stagnation point pressure
absolute upstream pressure
atmospheric reference pressure in pR = 1013 mbar
energy content stored in an arc
universal gas constant (R = 8.314 J/(mol K) )
stray resistance in the main path of a passive resonant breaker
stray resistance in the resonance path of a passive
resonant breaker
radius of arc
steady state voltage
arc voltage (determined by theoretical prediction)
net emission of radiation emitted from the plasma
time variant voltage
volume of pressurized bottles
volume of a pressure chamber
sonic velocity
axial plasma velocity (along z axis)
xxviii
Contents
symbol
T
TH
TR
unit
K
K
T
t
L
Xin
ms
H
m
Xout
x
z
W/(mK)
kg/m3
kg/m3
S/m
s
s
J
description
absolute temperature
absolute temperature in a pressure chamber
absolute temperature (T0 = 293 K laboratory conditions)
time
inductance
distance between upstream electrode and nozzle
throat inlet
distance between nozzle throat outlet and downstream electrode
normalized radius
axial distance measured from upstream electrode
ratio of specic heats (adiabatic index) (for air =
1.4)
thermal inertia of a single physical process in the arc
column
thermal conductivity
gas density
gas or plasma density
electric conductivity
thermal arc inertia (parameter is typically assumed
to be a function of conductance)
thermal inertia of a hypothetical arc
heat ux potential
1. Problem statement
The considerable growth of energy consumption and the replacement
of traditional band power (i.e. coal, gas and nuclear) with uctuating
renewable power (i.e. solar thermal, photovoltaic and wind) demands
major extension of todays transmission networks. In long distance
energy transmission and connection of o-shore power, HVDC transmission is superior to HVAC. This is so because DC transmission
losses are signicantly lower for cables with a length exceeding 50 km
and for overhead transmission lines with a length exceeding 700 km.
Limitations in HVAC transmission length exist due to the massive
consumption of reactive power in AC cables and transmission lines.
Meshed, VSC based HVDC networks, overlayed to todays HVAC
grid, are on a good way to become the backbone in a future European wide Supergrid [BF13a, FW13, CIG25, CCE11].
Among several unsolved problems to enable such a Supergrid, grid
protection in multi-terminal HVDC networks is one of the main obstacles [BF13a]. Arc based circuit breakers are standard measures
for fault current interruption in HVAC. HVDC circuit breakers have
been proposed and developed as passive resonance breakers, and hybrid breakers with limited interruptible currents (< 4 kA) [NNH+ 01,
PMR+ 88, HJ11], but yet no breaker technology became established
as circuit breaker, and performance increase would be desirable.
This thesis focuses on passive resonance HVDC circuit breakers
and HVDC load switches. Breakers without semiconductor switches
but only metal contacts have signicantly lower losses in the closed
state and tend to be signicantly cheaper. The motivation of this
thesis is to show ways to increase interruptible currents, decrease the
time required for current zero creation and reduce the component
1 Problem statement
Figure 2.1.: Cost break-down for AC and DC systems with line length
(copied from [HG06])
networks, b) long distance point-to-point connections and c) undersea connections [HG06]. Multi-terminal HVDC grids have been
proposed, but so far only three-terminal networks have been put
into operation [BTAM13, MMP+ 94]. Meshed multi-terminal HVDC
grids, in comparison to point-to-point connections, would bring major advantages:
Power transmission can continue even if one interconnecting
line is out of service [CJ97].
AC/DC converter stations are a major contribution to the
costs of HVDC connections. Figure 2.2 illustrates that in a
meshed network with 4 nodes, the number of required converter terminals can be reduced by a factor of 3 in comparison
to an identical point-to-point network. In the meshed network,
fault protection is carried out by HVDC circuit breakers.
Two dierent technologies for HVDC transmission with dierent
protection schemes exist. These are the thyristor based CSC (Current Sourced Converter) systems and IGBT based VSC (Voltage
10
11
12
structure was modeled with lines and with cables. One terminal is
connected directly to a common bus, two terminals are connected
via a 300 km line/cable to the same bus. A fault was assumed close
to the bus, on one of the connecting lines/cables. Figure 2.3 shows
the resulting fault current in the system, including a zoom of the
rst 10 ms. The total fault current is separated into its shares
that originate from dierent grid components. Critical components
that dominate the fault current are thereby identied and this serves
to improve the understanding of how fault currents evolve in VSCHVDC grids [BF13a, BWPF12]:
The steady state fault current is dominated by the AC infeed
over the freewheeling diodes of blocked VSC terminals. It increases with the number of terminals connected to a bus, and
decreases with the length of the line/cable over which the terminals are connected to the bus.
In the system with overhead lines, the initial slope was very
high and the rst peak of the fault current occured already at
a time < 2 ms. It is dominated by the discharge of the lter
capacitor. The current increases with the lter capacitor size
and with the number of terminals that are directly connected
to the bus [BF13a].
In the system with cable connections, the cable capacitance
discharge contributes signicantly to the fault current peak,
as well. However, its share reaches the circuit breaker location
typically with a small delay (eg. 1 2 ms), caused by the traveling time of the voltage wavefronts along the line. Thereby, a
second peak may result that exceeds the amplitude of the rst
one.
The simulation results have the following consequences for protection equipment in future HVDC grids: the time to interruption
or the time to current limitation will be a crucial parameter for the
choice of circuit breakers in overhead line and cable networks. In case
13
14
15
16
b)
surge arrester
surge arrester
LC- Resonance Path
mechanical breaker
d)
hybrid breaker
hybrid breaker
commutation to IGBT
absorber
unit
current
injection unit
absorber
unit
f)
commutation to IGBT
reactor
fast disconnector
IGBT
resonant
breaker reactor
fault current
reduction unit
surge arrester
17
hybrid mechanical
CB (2.4 c, 2.4 d)
expected
total
interruption
time
required
times for
commutation
/
energy
absorbtion
< 1 ms
< 2 ms
5 30 ms
commutation
< 0.1 ms
energy absorbtion
1 ms
commutation
< 0.2 ms
disconnector
opening < 1 ms
energy absorbtion
1 ms
contact
separation 20 ms
current
zero
creation
passive
resonance
30 ms
current
zero
creation
active
resonance 2 ms
current
state
of
development
contact
separation conventional
AC
breakers
20 ms
contact
separation
magnetically
driven
UFS
(UltraFast-Switch)
with
opening
speed > 20 m/s:
1.5 ms
- not yet avaliable
- slow AC breakers available
- UFS not yet
available
120 kV veried by
test (up to 320 kV
achievable)
AC
circuit
breakers > 500 kV
- UFS < 12 kV
< 5 kA expected
9 kA
experimentally proven
(up
to
16 kA
expected)
6 12 kA (estimated)
- passive: up to
4 kA in operation
active: up to 8 kA
possible
-able to survive
transient overcurrents
< 0.01 % (metal
contacts)
- development in
solid state device
technology to reduce on-state forward voltage or
number of modules in series
- eld experience
with prototype in
test grid
reduction
of
IGBT costs
-development
of
ultra-fast
mechanical drives to
reduce commutation time
maximum
rated
voltage
Un
max. DC
breaking
current
expected
power
loss
in
comparison
to
a
VSC
converter
station
further
development
steps
Mechanical passive
or
active
resonance
CB
(2.4 a, 2.4 b)
< 60 ms
- applied in CSC
HVDC
also
used
as
MRTB
(MetalReturn-TransferBreaker)
- optimizing DC
arc chamber for
passive resonance
to achieve higher
current
rating,
minimize
capacitor size and
minimize
time
for current zero
creation
18
19
solid state current injector unit in between. Each hybrid breaker can
interrupt only one current direction and consists of a vacuum interruption chamber with a parallel diode. First, the vacuum breakers
are activated. When they are in a state able to interrupt, the thyristor in the injection path is operated. At closing of the thyristor,
the current commutates into the current injection unit, via ground
and a diode path back to the conductor. It creates a current zero in
the vacuum breaker, at which the arc extinguishes. A capacitor in
parallel to the thyristor and an inductor in series to the latter form
an active resonance circuit. It interrupts the ground connection,
once the vacuum breaker has interrupted. The energy absorption is
performed in the diode path between ground and conductor. In addition to the diode, it includes a resistor in series plus an overvoltage
limiter.
The breaker has minor on-state losses, as no solid state elements
are placed in the main path. No details have been published on interruption time or on component size. However, the breaker is expected
to achieve interruption times in the order of active resonance HVDC
circuit breakers. In comparison to the latter, repetitive operations
are also expected to be possible, because no separate charging unit
is required, as the remaining fault current recharges the capacitor
automatically.
Mechanical passive or active resonance CB
Mechanical passive or active resonance breakers have been developed
for CSC HVDC systems and are based on AC gas circuit breakers. An additional LC-commutation circuit is placed in parallel to
the CB. This enables a current oscillation between the two parallel
paths and may create an articial current zero crossing in the main
path at which the CB can interrupt [PMR+ 88,TAYY85,Fra11]. The
oscillation can be achieved by an active current injection from a precharged capacitor or excited passively by the arc. Yet, no technical
solution has been found to increase the maximum interruptible current for passive resonance breakers to the desire level. This is a
20
21
22
line, as in AC, the breaker costs will play an important role when
choosing the technology. In particular for overhead line systems, fast
high performance breakers may be placed only selectively at strategic
locations. Such locations could be busses, where sub-grids or several lines are interconnected [BWPF12]. Arc based breakers could
be placed in a large number at less critical locations. In addition
to that, resonance based breakers can also serve as load switches in
such networks or they could be an integral part of hybrid breakers
(cf. gure 2.4f).
2.6. Conclusion
Long distance overhead line transmission and o-shore cable transmission for integration of uctuating renewable power is preferentially implemented in HVDC due to its low transmission losses.
HVDC circuit breakers have been identied as key enabling components for a future large-scale multi-terminal HVDC network. Among
the dierent candidate technologies for HVDC circuit breakers, so far
none meets all requirements for fault protection in such a network.
Passive and active resonance breakers are advantageous, as they are
a comparably low priced technology that can be installed in large
numbers in the grid. Breakers of this technology in operation consist
of a traditional AC circuit breaker with an LC-commutation path
in series and a surge arrester for overvoltage limitation. Presently,
they are limited in maximum interruptible current (< 4 kA), need
long times (> 30 ms) for current interruption and have comparably
large capacitors that contribute signicantly to the breaker costs.
23
24
4. Switching arcs
This chapter gives a theoretical introduction to the present knowledge in characterizing switching arcs. An introduction to cooling
mechanisms and chamber-arc interaction is given in section 4.1. Simple mathematical models for nozzle-constricted arcs under forced
convection are introduced in section 4.2. In section 4.3 phenomenological black-box models are discussed. Section 4.4 reviews classical
methods for experimental determination of black-box arc characteristics.
4.1. Introduction
4.1.1. Classes of arc models
Processes aecting an arc are very complex and can vary strongly in
their behavior as a consequence of the interaction with the environment. Electric, magnetic, chemic, thermodynamic and uiddynamic
processes simultaneously aect the plasma column. From the application perspective for HVAC and HVDC circuit breakers, mainly the
externally measurable parameters current i, voltage u and conductance g are of interest. The transient interaction of the arc with a
surrounding network during the high current phase and the extinction process as a consequence of intensive cooling at low currents
must be understood suciently to predict the breaker performance.
An arc can be completely described by seven coupled dierential
equations. These are [Rie67]: The coupling between voltage and
electric eld, the Poisson equation, the charge carrier balance, the
electron current density, the ion current density, the energy balance
25
26
4 Switching arcs
and the Saha equation. The coupled state variables voltage u, electric eld E, number of electrons Ne , number of ions Ni , electron
current density je , ion current density ji and temperature T vary in
time and space. For known start conditions, exactly dened geometry and known material parameters as functions of temperature, the
problem is theoretically solvable [Rie67]. However, under realistic
conditions a closed solution is almost never found. Limitations result of mathematical reasons from of complex boundary conditions,
imprecisely known material coecients and unknown start conditions. Therefore, in almost half a century of research on arcs, a large
number of strategies and models to predict the behavior of specic
arcs have evolved. The following three model classes are commonly
used:
Simple mathematical models limit the arc description to a certain type of arc under specic conditions [LL75,Low79,Low84,
SNSD06]. Often a single, or only a few processes can be identied that dominate the arc behavior. These processes are
precisely calculated, all others are neglected without signicant loss of accuracy. This approach requires only moderate
calculation eort.
Phenomenological models, also named black-box models describe
the arc by externally measurable parameters only. Typically,
a simple dierential equation is used whose parameters have
to be experimentally determined for each arc chamber [May43,
Cas39, Sch72, SK00]. By parameter tting, black-box models
achieve excellent transient accuracy and reect correctly the
dynamic interaction between an arc and its surrounding network by a two-port. In this case, the actual physical processes,
which are represented via the experimentally determined arc
parameter functions, are not of interest.
Computational Fluid Dynamics (CFD) simulations have emerged
strongly in the last decade. They solve the governing equations
4.1 Introduction
27
numerically for a given complex geometry and include the thermodynamic ows of the cold gas. Certain processes are also
neglected here, or the simulation is restricted to a thermodynamic equilibrium [SNSD06, IF08]. In such simulations the
eect of geometrical variations on dierent processes in the arc
can be visualized and thereby they can be studied easier.
(4.1)
28
4 Switching arcs
4.1 Introduction
29
30
4 Switching arcs
31
32
4 Switching arcs
a volume during the high current phase, which is used to blow the
arc during the low current phase towards current zero [FS06,CM97].
Three cooling mechanisms are considered to be relevant in the
energy balance equation to compensate the ohmic heating under
stationary conditions [LL75, TL75]:
1. the convective losses, caused by a plasma ow along an axial
temperature gradient T /z with sonic velocity vz , specic
heat Cp and plasma density p .
2. The net emission of the radiation power UN , not reabsorbed in
the plasma column. Part of this radiation is absorbed in the
cold gas vapor zone. The rest is absorbed by the nozzle wall
and causes mass ablation.
3. The radial conduction (including turbulent energy transfer)
due to a radial temperature gradient T /r with a thermal
conductivity k.
Even though all cooling mechanisms inuence the arc simultaneously, simple models have been developed that achieve acceptable
prediction accuracy by focusing only on one or two dominating processes. Depending on the current level, three dierent types of arcs
may result in a single experimental arrangement. These are nozzle constricted arcs stabilized by forced convection at low currents
and at high currents and ablative arcs at very high currents. The
exact current amplitudes at which a certain arc type appears depend strongly on nozzle geometry, blow gas parameters and nozzle
material.
33
currents, the arc plasma is not fully ionized and the electrical conductivity increases approximately exponentially with temperature
T . It has been shown by two dierent analytical approaches that
such arcs are characterized by a decreasing voltage with increasing
current (falling characteristic). Two authors [Sha06,Jen93] assumed
an axially uniform arc with radius x = r/rw , normalized to the
wall radius rw . They used the Elenbaas-Heller equation to predict a
radial temperature prole.
(
)
1
2 2
0 = E rw () +
x
.
(4.2)
x x
x
T
= 0 T is the heat-ux potential, () is the electric conductivity, is the thermal conductivity and E is the electric eld. The
equation results in an arc voltage U I 1 approximately inversely
proportional to the current [Sha06].
Tuma and Lowke [LL75, TL75] assumed a parabolic temperature
prole and can therefore represent the conduction losses by the following term: Pcond. = (4Tp )/A. The plasma temperature Tp is
uniform with radius. Tp and the arc-cross section A vary with the
axial distance z measured from the upstream electrode. The model
includes convective and radiative losses so that it is applicable also
for higher currents [LL75, TL75]:
0 = I 2 /(A2 ) p Cp vz (Tp /z) UN (4Tp )/A
|{z}
| {z } |
{z
}
| {z }
ohmic heating
axial convection
radiation
(4.3)
radial conduction
At low currents, the thermal conduction of this model causes the arc
voltage to decrease with increasing current. The phenomena would
be intensied by turbulence losses, which have been neglected in the
model [TL75].
Decreasing voltages at increasing currents are typical for free burning arcs at currents < 100 A [SO91]. But if sucient blow pressure
is applied, they may prevail up to several kA, as shown in measurements of a puer circuit breaker [NNH+ 01] and an air blast
34
4 Switching arcs
A
I 2 (1 f )
=
.
z
A
(4.4)
For this to apply, several assumptions must be made. Firstly, the arc
cross-section is assumed large enough, so the convection losses overweigh the conduction losses. Still, A must be small relative to AN ,
to allow free radial expansion (eg. j < 10 A/mm2 ). This is given, as
long as the ablated mass does not aect the pressure prole in the
nozzle. Secondly, the arc is assumed to be fully ionized with saturated conduction 100 S/cm [LL75]. Further, a linearly pressure
dependent volume enthalpy p hp /p = const., constant fraction f of
power lost by radiation, and axial independence of f , p hp vz , and
is assumed. By integrating equation (4.4) for a cylindrical nozzle,
i
2(1 f )
AT = z
.
p
(p hp )vz
35
(4.5)
2p hp vz
U=
(4.6)
LT p .
(1 f )
A linearly increasing cross-section with current is in agreement
with several experimental studies from the 1970s [LL75] and has
been experimentally conrmed in section 9.2. Constant arc voltages
at increasing current have been observed at several kA in circuit
breakers with small puer diameter [NNH+ 01].
( )2
I
A
(4.7)
36
4 Switching arcs
37
downstream electrode
AE current conducting zone
(T20000K, 100mho/cm)
AV light emitting cross-section
upstream electrode
iLT
u
(4.8)
38
4 Switching arcs
section.
Fluid dynamic arc cross-section
In a nozzle with a CSA, nearly all mass ow is carried by the cold gas
[Low79]. In comparison, the hot plasma contributes only minimally
to the gas ow and acts as an obstacle, partly blocking the nozzle
outlet. This is due to the low density of the plasma in comparison to
the cold gas. The subsequent mass ow reduction can be rated by
the dierence between the known eective nozzle cross-section ANe
(without arc) and the eective cross-section Ae (with arc). A uid
dynamic arc cross-section AP is dened as:
AP = ANe Ae
(4.9)
For air, a sonic ow occurs if the upstream to downstream pressure ratio exceeds the critical value pup /pdn > 1.89 [Whi99]. Under
sonic conditions (cf. equation (4.10)), an adiabatic ow rate m
S
in a converging-diverging nozzle is dependent only on the upstream
pressure pup with the eective nozzle cross-section Ae , the specic
heat ratio for air = 1.4, the gas temperature T and the ideal gas
constant R [Whi99]. Under subsonic conditions, the downstream
pressure pdn aects also the ow rate m
sub (cf. equation (4.11)).
(
) +1
1
(4.10)
m
S = Ae pup
RT
+1
m
sub
= Ae k
RT
(4.11)
with
v
u(
u
u
k = pdn t
2
1
)(
pdn
pup
)(2/)
pdn
pup
(4.12)
39
The adiabatic mass ow can be determined from a pressure measurement pH in a circuit breaker pressure volume VH . A change in pH
results from an imbalance of inowing mass from a storage volume
m
in and outowing mass through a nozzle m
out (cf. equation (4.13)).
Hereby, the temperature of the heating chamber TV varies due to gas
expansion from start conditions T0 and p0 (cf. equation (4.14)).
RTin RTH m
out = VH
(
TH (t) = T0
pH (t)
p0
dpH
dt
(4.13)
(4.14)
40
4 Switching arcs
(4.15)
41
Under the assumption that the cooling power P (Q) and the conductance g(Q) are arbitrary functions of Q, the general form of the
dynamic arc equation [CW93, Kap11] can be formulated:
( 2
)
1
i
g = Q
P .
(4.16)
g
g
(4.17)
42
4 Switching arcs
( g )bpg
1S
(4.19)
( g )d
(4.20)
1S
of the stationary and dynamic arc characteristic. It correctly describes only certain types of arcs in a limited conductance range,
but is often used for characterization of switching arc chambers
[CW93, NNH+ 01]
The stationary characteristic is often plotted as U (I) or P (I) instead of P (g) [CW93]. All descriptions are equivalent under stationary conditions and can be transformed into each other using
P = U I and g = I/U . Many authors tted the stationary characteristic as potential function of either I or g. A conversion Table 4.1 has been included to relate the most common descriptions:
P (g) = 1 W apg (g/(1 S))bpg , U (I) = 1 V aui (I/(1 A))bui and
P (I) = 1 W api (I/(1 A))bpi . It has to be emphasized that dynamically the descriptions are not equivalent anymore. A stationary
description has to be converted to P (g) to represent the cooling
power in the twice modied Mayr equation (4.18).
(g) = c
43
to \ from
bui
bui
1 bui
bpi
bpi 1
bpi
bui + 1
1 bpi
bpg
bpg 1
bpg +1
2bgp
1+bgp
bpg
1+bui
1bui
bpi
2bpi
1 bgp
aui
api
agp
1 aui
1 api
to \ from
aui
api
apg
1 aui
(aui )
1
( 1+b
(agp )
1 api
2
)
( 1b
ui
(api )
gp
1
( 1+b
)
gp
(agp )
2
)
( 2b
pi
1 agp
44
4 Switching arcs
45
stationary
1
R = 5
voltage in kV
I = 0.1 kA
ms , = 10s
0.8
I = 1 kA
ms , = 10s
I = 1 kA
ms , = 20s
0.6
I = 10 kA
ms , = 10s
I = 10 kA
ms , = 20s
0.4
I = 100 kA
ms , = 10s
0.2
0
0
I = 100 kA
ms , = 20s
0.1
0.2
0.3
current in kA
0.4
0.5
0.6
At low current slopes 0.1 kA/ms, the transient voltage is dominated by the stationary characteristic P (g), because the current rise
time TR = 3 ms is much larger than . Under moderate current
slopes of 10 kA/ms, P and aect the transient voltage simultaneously, because TR = 30 s is in the same order of magnitude as
. The transient voltage response during a high current slope of
100 kA/ms is dominated by the stationary state prior to the current
increase. Because of the small rise time TR = 3 s, the arc does
not strongly change its conductance during the current slope, but
mainly afterwards. This leads to a linear increase of the arc voltage
with current during the slope.
A transient arc parameter can only be determined from those sections of the current-voltage oscillogram where the transient voltage
is signicantly aected. The comparison of transient voltage waveforms with = 10 s and = 20 s reveals that shows the greatest
eect around a 10 kA/ms slope and has no or only a minor eect at
slopes 0.1 kA/ms and 100 kA/ms (cf. gure 4.3). An exception are
46
4 Switching arcs
47
g(t2 )
g(t
2)
for i(t2 ) = 0.
(4.21)
(4.22)
i21 g 2 i22 g 1
g(g 2 g 1 )
(4.23)
g(i22 i21 )
i21 g 2 i22 g 1
(4.24)
Amsinck [Ams77] applied this method for the rst time to a single u,i measurement of an extinguished arc with thermal reignition
(cf. Table 4.2C1)). There, the assumption is made that the arc is
identical before and after current zero. Ruppe [Rup79] applied the
method to identical conductance values before and after the peak
48
4 Switching arcs
i
U
u*(t)
u(t)
g1
g2
- one experiment
- iterative recalculation of
voltage waveform
C1) Amsinck
g
g
g1
g2
g=g0
g=g0 g 1
g2
t
t=t1
t=t 2
i=i 1 i=0 i=i 2
g1
g2
t=t1
i=i 1
t=t 2
i=i 2
t
i=0
g1
g2
g=g0
t0=-1/
t=t1
i=i 1
t=t 2
i=i 2
P0
t
i=0
/g
49
of a current half cycle (cf. Table 4.2C2)). Ruppe also applied this
parameter determination method successfully to two or multiple similar experiments with varied current amplitudes or current slopes (cf.
Table 4.2C3)).
C4: Graphical Multiple Gradient Method: On the suggestion of Rother, Drebenstedt et. al. [DRWZ83] implemented the
calculative procedure of Amsinck-Ruppe in a graphical form and
applied it to pressurized gas and SF6 arc chambers. To do so, equation (4.18) can be written as
g
ui = (P ) + P .
g
(4.25)
4.4.4. Comparison
In the parameter separation method B and the multiple gradient
method C, a specic current waveform must be created. In the iterative method A, theoretically any waveform is suitable as long as
it contains sucient high current slopes, because only the four coecients a, b, c and d must be determined. Method C4 application
50
4 Switching arcs
51
53
54
- converted to U (I)
- converted to P (g)
circuit parameters
L0 = 130 H
R1 = 1 m
arc parameters
- specied in [NNH+ 01]
bpi = 0.5
d = 0.5
aui = 0.135 106
biu = 0.5
apg = 6.92 106
bpg = 0.33
C = 28 F
R2 = 1 m
uc
surge arrester
S2
R2
commutation path
S1
main path
a)
iF
L
R1
i2
g
u
i1
b)
i1
current in kA
8
6
4
2
0
2
4
a)
iabs.
iF
10
12
14
16
8
10
time in ms
12
14
16
voltage uC in kV
200
100
conductance g in S
b)
c)
i2
55
6
4
2
0
voltage in kV
15
d)
uc
10
5
0
5
1
3
4
time in ms
56
57
+ uc +
i2 = i2
(5.1)
dt
Lg R1 R2
L Lg
d
1
u c = i2
dt
L
dg
1
=
dt
(iF i2 )2
g
P
(5.2)
)
(5.3)
58
point
i
d 2
uc = A|I2 ,Uc ,G,IF
dt
g
i2
uc
g
(5.4)
with
R
1 R2 1
L
L
LG
1/C
A=
2(IF I2 )
a cGbpg +d
pg
L1
0
0
IF
LG
2 +
0
I2
LG2
1d
cGd
. (5.5)
(5.6)
59
60
12
excited
academic
current i1 in kA
10
rising IF
closing time of S2
8
6
4
2
a)
10
time in ms
15
20
5.5
academic
excited
arc voltage u in kV
4.5
4
3.5
3
2.5
2
1.5
1
0.5
b)
3
4
5
arc current i1 in kA
5.4 Discussion
61
5.4. Discussion
Asymptotically stable parameter combinations have only negative
eigenvalues of A (cf. equation (5.4)) and are indicated by black
circles in gure 5.4a)-b). They correspond, correctly, to the cases
where no current zero crossing occurred during Tsim .
Fault Current: The fault current inuences T in two ways.
Firstly, at high fault current amplitudes IFa , the stationary characteristic U (I) of a blown arc is typically at and the thermal inertia is large [BMRL85, BKM+ 70, NNH+ 01]. Hence, the system is
typically stable for high currents and instable for low ones. Secondly, a linearly rising fault current iF stands in competition with
the exponential resonance current growth. Consequently, a rising
fault current slope delays the current zero creation.
Stationary arc characteristic: Large stationary arc voltage
dU
U (I)|I=I1 and strongly falling characteristic dI
|I=I1 in the operation
1
point boost passive resonance.
Thermal inertia: Large thermal arc inertia , resulting from
large parameters c and d, delays the conductance change of the arc.
U
By that, a falling stationary characteristic I
|I=I1 results in a atter
1
u
or even rising transient characteristic i1 . For innite or innite
current gradient di1 /dt, the transient arc voltage u = i1 /G would
resistively follow the arc current. Consequently, the lowest possible
values of factors c and d are desirable.
Resonance frequency: Large current gradients add damping to
the system, because they increase the lagging of the transient arc
voltage. The high oscillation frequency in passive resonance creates
62
ref
normalized time T / T
10
C
L
R
10
R2
I0a
10
I0b
stable
10
a)
10
10
10
10
10
normalized parameter X/ X
10
ref
ref
normalized time T / T
10
c
d
U0
dU
10
stable
0
10
b)
10
10
10
10
normalized parameter X/ Xref
10
63
64
P =
Pji +
Pjg
j
(5.7)
(
(1 W) ajpi
I
1A
)bjpi
+
( g )bjpg
(1 W) ajpg
1S
(5.8)
The dierentiation of equation (5.9) with respect to changes in current results in:
U (I)
=
ajpi (bjpi 1)I (bjpi 2)
I
j
(
2)
ajpg I (bjpg
1 + bjpg
)
1)
bjpg
1+bjpg
(5.10)
65
( p ) ( g )bpg
1 bar
1S
(5.11)
with apg = 0.393 106 and bpg = 0.25. The tted curves (solid lines
in gure 5.5 a) and b)) describe the measurement data accurately
so that the eect of blow pressure on the arc characteristic is well
described by equation (5.11).
In the above 10 bar example the stationary characteristic is
P (10 bar) = 3.93 MW
( g )0.25
.
1S
(5.12)
66
10
70
Bonin 10bar
Bonin 30bar
Bonin 50bar
extrap. 100bar
60
stationary power loss P in MW
7
6
5
4
3
2
40
30
20
10
1
a)
50
10
20
30
current I in kA
40
b)
50
100
conductance G in S
( g )0.25
.
1S
(5.13)
67
10
ref
normalited time ( T / T)
10
10
10
10
10
normalized blow pressure ( p / p
H
10
)
Href
68
5.7. Conclusion
Simulations of a passive resonance MRTB revealed that for its satisfactory operation with a given arc chamber resonance circuit parameters are suitable in a limited range of values only and can add only
relatively small enhancement of performance to the breaker. In comparison, small changes of the arc characteristics P (g) and (g) have
a signicant eect on the interruption time. Hence, an optimally
designed arc chamber allows the interruption of higher currents or
the choice of a smaller capacitor.
In order to create a falling characteristic even at high currents,
it must be investigated, how strongly dierent loss components (ie.
conduction, convection, turbulence and radiation) scale with the current or with the conductance. In a second step, technical measures
must be found to modify the arc chamber in a way that undesired
loss components are reduced or that desired loss components are
enhanced. Irrelevant loss components are, in principle, desirable as
well. They may not inuence if passive resonance occurs, but they
may speed up the passive resonance current increase for a system
which is already instable. Further, they may support arc extinction
once a current zero crossing is reached. The increase of blow pressure
was identied as an example measure to increase desirable losses.
A systematic search for arc congurations with strongly falling
voltage-current characteristic and lowest possible arc inertia is advisable to improve the performance of HVDC circuit breakers. This
research requires an accurate and ecient determination of arc characteristics. An improved method for accurate characterization will
be discussed in chapter 7.
6. Experimental methods
In this chapter, the experiment setup and experimental methods
are discussed. Section 6.1 describes the arbitrary current source.
Its development is an integral part of this thesis. An LC current
source used for sinusoidal current measurements is briey discussed
in section 6.2. Section 6.3 characterizes the model circuit breaker
used for the systematic study with blown and wall stabilized arcs.
Section 6.4 briey species a spark gap used for free burning arc
experiments. An overview on diagnostic measurement equipment is
given in section 6.5. Three methods for the determination of arc
cross-sections are introduced in section 6.6.
6.1.1. Functionality
The novel arbitrary current source is used to create complex current
waveforms with repetitive constant current sections interspersed by
fast current slopes (up to 150 kA/ms) in a single experiment [WF10].
The source consists of three parallel interleaved modules with a capability to provide a voltage of 3 kV and a current of 1 kA each and
is equipped with a common IGBT controller. The load current iload
is the superposition of all three module currents in . The functional
principle of the modules is similar to that of a buck converter. An
69
70
6 Experimental methods
uc
C3
C2
F3
S3
F2
S2
F1
S1
R3 L3
i3
R2 L2
iload
i2
R1 L1
i1
C1
Sc
Rc
Uc0
D1
D2
D3
uload
Rload
equivalent circuit of the current source is shown in gure 6.1, a picture of the three modules is provided in gure 6.2.
Each module n consists of a pre-charged capacitor Cn , a fuse Fn ,
an IGBT Sn with antiparallel diode, an inductor Ln and a freewheeling diode Dn . For closed IGBT Sn , a positive voltage dierence
uC uload is applied across the inductor Ln and if uC > uload a
positive current gradient results:
uC uload
din
|on =
.
dt
Ln
(6.1)
(6.2)
71
Figure 6.2.: Arbitrary current source: The picture was taken in the
source room. The three parallel modules are shown.
72
6 Experimental methods
uC d uload
.
Ln f
(6.3)
(uc u
eload )
.
In
(6.4)
Each inductor Ln is equipped with 10 mechanical tabs, which are distributed logarithmically along the windings. This allows to vary the
number of windings and with it the inductance with small steps in a
wide range (40 H-4.5 mH). Complex current waveforms are achieved
by combined or opposed operation of several modules with dierent
current slopes. An ideally constant current period is achieved by
two modules with an inductance ratio
L2
u
eload
=
,
L1
uc u
eload
(6.5)
73
74
6 Experimental methods
current Iload in kA
0.8
a)
2
0.6
1.5
0.4
0.2
0.5
0
0.2
0.4
0.6
0.8
b)
0.5
1.5
2.5
current Iload in kA
2.5
c)
2
1.5
0.5
1
0.5
d)
10
0.5
20
30
current Iload in kA
1.5
current Iload in kA
e)
g)
1
1
0.5
0.5
0
0.2
0.4
0.6
0.8
f)
1.5
1.5
0.5
0.5
2
time in ms]
h)
10
15
20
time in ms
1.5
25
75
Current slopes
A rising slope with constant gradient
Uc Uload
I =
Ln + LB
(6.6)
76
6 Experimental methods
77
78
6 Experimental methods
tion of the arc and the resistance of the test circuit. Inductances
and capacitances in the range of 4 H - 10 mH and 3.5 14 F are
used. The resulting oscillating frequency of the test current is in the
range of 0.5 17 kHz.
6.3.1. Functionality
In the model circuit breaker an arc is burning vertically in a nozzle
arrangement with top-down current direction, the forced axial blowing is bottom-up along the arc. The breaker is equipped with two
vertical copper-tungsten electrodes: a xed 18 mm diameter electrode at the bottom and a pneumatically movable 5 mm diameter
upper electrode. The lower electrode is mounted inside a closed pressure volume of 1 liter, into which gas from eight pressurized bottles
of 1.5 liter each can be released instantaneously by parallel opening of 8 valves. A vertical nozzle is mounted on top of the pressure
chamber, so that an arc between the two electrodes is axially blown
in a single ow arrangement. The geometry of valves, pipes and the
pressure volume is arranged in a way that the minimum cross-section
79
is inside the nozzle throat. This ensures that sonic conditions if once
they are reached, always occur in the nozzle throat. If not specied
otherwise, dry air is used as blow gas.
In AC circuit breakers, the contact separation is typically achieved
by pre-tensioned springs. Such mechanism is very complex and has
the disadvantage that contact velocity and distance are xed. Instead, a pneumatically movable upper electrode has been used. It
achieves a variable opening speed up to 3.4 m/s and a variable electrode gap distance up to 200 mm. The arc is ignited by applying
a 100 A pre-current during the contact separation time of typically
80 ms.
80
6 Experimental methods
Iarc
pA=1bar
movable electrode
gas flow
Xout
Lout
LT
dT
nozzle
Xin
Lin
fixed electrode
pB
pB
Iarc
pH
current-gradient-excited transient voltage overshoot and the stochastic uctuations. With respect to that, a free-burning horizontal arc
is optimal because it features a relatively low voltage but strong
uctuations and thus has a low signal to noise ratio [SO91]. In comparison to free burning arcs, wall stabilized arcs and blown arcs have
higher arc voltages and exhibit lower uctuations.
Spherical copper electrodes of 50 mm radius and a 4 mm spacing
gap are used. The arc is ignited by means of a spark plug for the
experiments in the LC-resonance test circuit. For the measurements
in the new arbitrary current source, a copper igniting wire with
50 m diameter is used. To minimize the eect of the igniting wire
on the arc characteristic, a constant current of 75 A is applied for at
least 100 ms prior to the evaluated current interval.
81
82
6 Experimental methods
nozzle inlet. The sonic cold gas traveling time reduces the dynamic
resolution to t = 0.3 ms.
Pressures pH and pB are not measured as absolute values but as
overpressures relative to atmospheric pressure and temperature pR =
1013 mbar, TR = 293 K.
83
1
(
y)) .
n n m
(6.7)
84
6 Experimental methods
85
6 Experimental methods
pressure pH in bar
pressure pB in bar
86
time in ms
time in ms
87
Safety circuit
1
0
1
0
3kV
0
Upload Data
Capacitor voltage VC
1
0
1kA
0.1k
A
1
0
1
0
Bottle pressure pB
Electrode positon d
Blow pressure pH
2s
10s
240s
Charge capacitors
5s
Fill bottles
1ms
Create pre-current
80ms
Separate contacts
5ms
15ms
1ms
80ms
Close contacts
5s
discharge
89
0.1 kHz
i
u
10* g
500
a)
b)
c)
0
500
0
10
20
30
40
1 kHz
i
u
10* g
500
0
500
0
0
0.1
0.2
0.3
0.4
80 kHz
500
500
0
0.01
0.02
0.03
0.04
time in ms
0.5
i
10* u
10* g
0
d)
5
i
u
10* g
10 kHz
500
500
0
50
0.05
0.06
91
700
0.1kHz
1kHz
10kHz
80kHz
Stationary
600
arc voltage u in V
500
400
300
200
100
0
100
100
200
300
arc current i in A
400
500
7.1.1. Validation
Hypothetical arcs with 0.1 and 10 kHz sinusoidal currents have been
simulated. The simulations are identical to the ones in gure 7.1a)
and c), but with weaker damping in the current waveform, so that
1020 subsequent oscillations resulted. All methods were used to reextract the arc parameters PH (g) and H (g) from an ideal hypothetical arc without any noise in the voltage signal uH . A correct extraction would result in constant parameters for all conductance values.
The extracted arc parameters by methods B, C1 and C4 are compared in gure 7.3a) and b). Black dashed lines indicate the correct
parameters, red triangles (B), blue circles (C1) and green squares
(C4) indicate the re-extracted points. The extracted parameters by
methods B, C1 and C4 agree within a 1 % band with the assumed parameters in the hypothetical arc. The largest deviations resulted at
low conductance values. (g) values could be extracted much better
from the high frequency signal. At 0.1 kHz the determined values are
correct, but could be evaluated for very low conductance values only.
Further simulations and recalculations have also been performed for
non-constant arc parameters (e.g. PH = 1W apg (g/(1 S))bpg and
H = c (g/(1 S))d ). In their case as well an error < 1 % resulted.
The simulations conrm that all methods have been implemented
correctly. All methods can be used to recalculate the parameters
93
10
P in kW
a) 9.9
10.1
Undesturbed
10.2
10
9.8
9.6
9.4
10.2
10
0
11
in s
3 V gaussian noise
Filtered
10
Filtered
10
0
0
20
40
conductance g in S
in s
P in kW
20
10
20
10kHz
10
20
0.1kHz
10
0
d)
10kHz
0
in s
P in kW
P in kW
c)
20
0.1kHz
10.1
10
9.9
b)
e)
Undesturbed
in s
10.1
in s
P in kW
20
10kHz
10
0
0
20
40
conductance g in S
95
7.1.3. Consequences
Simulations with hypothetical arcs revealed the sensitivity of methods B, C1 and C4 to arc uctuations. In particular, values cannot
be extracted with sucient accuracy to compare dierent arc chambers by comparing their arc parameters. Method A is much less
sensitive to noise, but is not considered an alternative, because it
is restricted to arcs, whose parameter shape function can be specied before the measurement (e.g. P = 1 W apg (g/(1 S))bpg and
= c (g/(1 S))d ).
The performance of the classical methods is not sucient for the
planned schematic comparison of dierent arc chambers. It was,
therefore, necessary to develop an improved parameter determination method, with the same robustness to noise as method A but
without the requirement to pre-specify an analytical function, which
the arc parameters are assumed to follow. The improved method
presented in the next section tries to separately measure the stationary and transient arc characteristic and to simplify the parameter
extraction by a choice of a more complex current waveform.
38
37
0.52
0.5
36
0.48
35
0.46
34
0.44
i in kA
0.42
u in V
0.4
75.55
75.6
75.65
75.7
33
32
75.75
time in ms
97
constant current intervals, but uctuations are superimposed. Furthermore, a transient arc voltage spike occurs during the current step
followed by a subsequent exponential decrease in the constant current phase towards the new stationary voltage. From the observed
transients one can conclude that the free-burning arc under investigation is quite well described by a rst order dierential equation,
i.e. the current steps are used to identify a suitable arc model.
It is now either necessary to characterize the arc model-independently, by using a large number of such experiments with dierent
current gradients, or to choose a black-box modeling equation and
to determine its parameters as a free function of an arc state. A
model-independent arc characterization is discussed in section 7.2.2,
an improved method for determination of arc parameters is proposed
in section 7.2.3. In agreement with other authors equation (4.18) is
chosen [Sch72] and the parameters P (g) and (g) are determined as
free functions of conductance g.
Pj Pj1
Gj Gj1
(7.1)
and bj = Pj aj Gj .
4. The conductance gj (t) = ij/uj and its gradient g j (t) are calculated for each section. The chosen model equation (here
equation (4.18)) is also solved for the current gradient using
equation (7.1)
(
)
1 i2
gj .
(7.2)
gj (t) =
j p j
(g j g j )2
99
(7.3)
101
current i in kA
stair
c)
e)
1.5
1.5
0.5
0.5
a)
spikes
0.2
100
0.4
0.6
0.8
b)
(u )=5 V
100
80
(u )=1 V
80
60
undesturbed
60
10
15
10
15
5
10
time in ms
15
N
N
40
40
20
20
0.2
0.4
0.6
0.8
d)
14
14
12
(u )=500 V
12
10
(uN)=100 V
10
undesturbed
0.2
0.4
0.6
time in ms
0.8
f)
stairs
a)
c)
spikes
15
15
10
10
(uN)= 5V
103
(uN)=1 V
undisturbed
original
100
200
300
conductance g in S
b)
7000
7000
6000
6000
5000
5000
4000
4000
3000
(uN)=500 V
(uN)=100 V
2000
undisturbed
1000
0
original
0
0.2
0.4
conductance g in S
100
200
300
conductance g in S
0.2
0.4
conductance g in S
3000
2000
1000
d)
stairs
15
10
10
(u )=5 V
N
(uN)=1 V
undisturbed
original
a)
spikes
15
100
200
conductance g in S
300
b)
30
c)
100
200
conductance g in S
0.2
0.4
conductance g in S
300
30
(u )=500 V
N
(uN)=100 V
25
25
undisturbed
original
20
20
15
15
10
10
0.2
0.4
conductance g in S
d)
8. Experimental comparison of
arc parameter determination
methods
In this chapter, a free burning arc and a wall stabilized arc are used to
experimentally validate and compare the dierent classical with the
new arc parameter determination methods. In section 8.1 the novel
improved method D1 proposed in section 7.2 is compared with the
classical methods of section 4.4.3. In section 8.2, the current waveform is optimized for the improved determination methods D1and
D2.
105
106
107
current i in A, voltage u in V
0.84kHz
a)
150
i
100
50
0
50
100
150
200
6
7
time in ms
10
11
cooling power P in kA
10
6
4
2
0
b)
Method A
Method B
Method C
10
15
conductance g in S
20
25
10
15
conductance g in S
20
25
thermal inertia in s
20
c)
15
10
5
0
108
current i in A, voltage u in V
3.6kHz
a)
200
i
150
100
50
0
50
100
150
3.5
4.5
time in ms
5.5
cooling power P in kW
40
b)
30
20
Method A
Method B
10
0
Method C
0
10
15
conductance g in S
20
25
10
15
conductance g in S
20
25
thermal inertia in s
20
c)
15
10
5
0
109
current i in A, voltage u in V
8.3kHz
a)
400
200
0
200
400
1.2
1.4
1.6
1.8
time in ms
cooling power P in kA
50
b)
Method A
40
Method B
30
Method C
20
10
0
10
15
conductance g in S
20
25
10
15
conductance g in S
20
25
thermal inertia in s
20
c)
15
10
5
0
110
Methods A-C all extracted arc thermal inertias of 0 20 s. However, the resulting strongly varies with method and current oscillation frequency: Method A results in a curve (g) = c(g/(1 S))d with
c = 4.3 s and d = 0.3. A similar result was found for the 8.4 kHz
measurement with method B and for the 3.5 kHz measurement with
method C.
For a reliable evaluation also at high conductance values, method
B requires high frequency oscillations as it uses current zero crossings for the evaluation. At low conductances, method B yielded
unexpectedly high values.
Method C extracts P and simultaneously and, consequently,
works reliably only in a narrow frequency range of the arc current.
At 840 Hz the voltage waveform is mainly stationary and cannot be evaluated at all. At 8.4 kHz the uctuations in P cause, as
explained above, strongly scattered and rather low values. The
3.6 kHz test current provided reasonable results for in the range
3 S < g < 13 S. At very low conductances, unexpectedly low values resulted, and at very high values a strong scatter is observable.
The evaluation method C is very sensitive to the chosen test current
frequency. Evaluation of from experiments with 1 kHz and 5.8 kHz
show comparably poor results like those from 840 Hz and 8.4 kHz,
respectively.
111
current step, a voltage peak with an approximately exponential decay of duration tfast = 10 15 s is observable. A quasi-stationary
arc voltage section follows. This gives a rst rough estimation of
the thermal inertia . During the constant current phase, the arc
voltage is not fully constant but uctuations with time constants
of tslow > 500 s occur. This behavior is due to uctuations of the
arc geometry or in its temperature distribution and is not described
by the black-box equation (4.18). Thus, even a perfect parameter
evaluation method will show a certain error in predicting the quasistationary behavior. However, the large dierence in the dynamics
tslow tfast justies a extraction with a minimum error.
Parameter evaluation results
The resulting arc cooling power during one experiment with staircaselike current is shown in gure 8.4b). The transient heating power
pm is plotted with a solid blue line and the quasi-stationary values Pj from the constant current phases are indicated by red circles. These values were extracted as the mean value of the arc
voltage during the constant current sections starting 50 s after the
current step up to the next step. This assured that the fast dynamics have already vanished and the slow dynamics are averaged
out. The stationary arc characteristic is approximately described by
Pt = 1 W apg (g/(1 S))bpg with apg = 659, bpg = 1.24. From the
shape of the Pj (gj ) curve (cf. gure 8.4b)) it is possible to judge if a
linearization between the points is reasonable or if another measurement with smaller step size is required. Moreover, a large scatter in
Pj (gj ) would indicate that the external physical conditions are not
maintained suciently constant.
The extracted (g) values are displayed in gure 8.4c). A general
increase with increasing conductance can be seen, and (g) can be
approximately described by t = c (g/(1 S))d with c = 6.7 s and
d = 0.23. The standard deviation is 4 s.
Recalculation of the step current experiment with evaluated arc parameters: As a consistency check, the arc voltage can
112
voltage u in V
current i in A
40
0.5
Ii
a)
75.6
cooling power P in kW
40
75.8
76
76.2
76.4
76.6
time in ms
u
76.8
77
30
77.2
ppm
P
P
30
P
P
=a*g
fit
20
10
b)
10
15
20
conductance g in S
25
30
thermal inertia in s
20
15
10
t j
t =c*gd
fit
c)
10
15
20
conductance g in S
25
113
voltage u in V
40
38
36
34
U
um(t)
32
a)
75.6
75.8
76
76.2
76.4
time in ms
76.6
U
usim(t)
76.8
77
77.2
cooling power P in kA
60
40
30
20
10
0
b)
measurement 1
measurement 2
measurement 3
measurement 4
50
10
15
20
conductance g in S
25
30
10
15
20
conductance g in S
25
30
thermal inertia in s
20
c)
15
10
5
0
114
115
result is very satisfying, deviations only occur directly after currentzero as the re-ignition is not covered in this black-box model.
116
100
voltage u in V
840Hz
a)
uum
uusim
50
0
-50
-100
100
voltage u in V
3.6kHz
50
0
-50
b)
-100
voltage u in V
60
4.2
4.4
4.6
4.8
8.3Hz
40
20
0
-20
-40
-60
c)
1.75
1.8
1.85
1.9
1.95
time in ms
2.05
2.1
2.15
117
instances with g = 0 that are part of the arc dynamics but are easily
misinterpreted as stationary points. Secondly, at current zero the
values g and g are not dened, due to division by zero, and interpolation methods must be used. Thirdly, any time shift between
voltage and current zero crossing leads to unreasonable values.
Such a shift occurs quite regularly and can be the consequence of a
measurement oset or a delay from the measurement equipment and
long cables. The cooling power is successfully evaluated only for low
frequencies as the arc is truly stationary at the current peak only
for these current shapes. For increasing test current frequency, the
arc is no longer in the stationary state which leads to an increased
scatter in the evaluation. The opposite is true for evaluation of .
From stationary arcs (low frequency oscillations), the thermal time
constant cannot be evaluated. To obtain reliable data from method
B, at least two measurements with signicantly dierent oscillation
frequencies need to be performed.
Method C relies strongly on the validity of equation (4.18) because
it determines P and simultaneously from the dynamic waveform.
Consequently, only a narrow band of current frequencies allows an
accurate parameter extraction. In our example, only the measurement at 3.6 kHz was successful. Experiments with slightly dierent
test current frequencies have been performed (2.5 kHz and 5.5 kHz;
not shown here). Their evaluation showed considerably less accurate
results, similar to the ones shown for 840 Hz and 8.3 kHz. The graphical implementation C4 of method C produces a certain intrinsic averaging by the straight line tting but has strong error propagation
for and is thus very sensitive to arc uctuations. Consequently, it
requires considerable eorts in implementation to deal with all the
numerical problems and a manual identication of poorly extracted
points.
Reliable evaluation of P and with low scatter can be done using the new method D1. From a single experiment, it is possible
to determine the arc characteristic over of the entire conductance
range. From gure 8.4a) it can be seen that the arc voltage is not
118
8.1.4. Conclusion
The most important challenge in black-box modeling is the choice
of a parameter function to describe the arc and the determination
of the parameter values for all states of the arc. The limitations
of conventional parameter determination methods from experiments
with sinusoidal test currents have been demonstrated by the example
of a free-burning arc. These are mainly three problems: a) the
119
120
121
a)
u
2
0.5
10
15
20
time in ms
25
voltage u in kV
current i in kA
0
35
30
voltage U in kV
0.8
rising
falling
mean
min
max
0.7
0.6
0.5
0.4
0.3
0.2
b)
0.5
time in ms
1.5
0.6
0.3
voltage u in kV
current i in kA
0.4
c)
12.2
12.4
12.6
12.8
13
time in ms
13.2
13.4
13.6
0.2
122
voltage u in kV
0.45
123
stationary
0.4
I =1kA
0
0.35
I =0.75kA
0.3
I =0.5kA
0.25
I =0.2kA
0
0
0
0.2
a)
0.15
0.5
1
current i in kA
1.5
voltage u in kV
0.45
stationary
kA
= 36 ms
0.4
kA
= 10 ms
0.35
0.3
= 5:5
kA
ms
0.25
= 2:2
kA
ms
0.2
voltage u in kV
b)
0.15
0.2
0.3
0.4
0.5
current i in kA
0.6
0.7
0.45
stationary
0.4
I=0.7kA
0.35
I=0.6kA
I=0.4kA
I=0.22kA
0.3
0.25
0.2
c)
0.15
0.4
0.6
0.8
1
current i in kA
1.2
124
125
steps i
current i in kA
1
1
0.5
voltage u in kV
0.5
a)
0.1
0.2
0.3
0.4
time in ms
1.4
0.5
0.6
spikes i
u
0.7
1.4
1.2
1.2
1
0.8
0.6
0.8
0.4
0.6
voltage u in kV
current i in kA
0.2
0.4
b)
6
time in ms
10
cooling power P in MW
126
0.6
0.5
0.4
0.3
0.1
0
a)
stairs
spikes
stairs
spikes
0.2
0.5
1
1.5
conductance g in S
2.5
voltage u in kV
1.4
1.2
1
0.8
0.6
0.4
b)
0.2
0.4
0.6
0.8
current i in kA
1.2
thermal inertia in s
40
c)
30
20
10
0
0.5
1
1.5
conductance g in S
2.5
thermal inertia in s
40
I at I0 =0.4kA
I at I0 =0.15kA
30
20
10
0
a)
100
200
300
400
500
spike size I
thermal inertia in s
40
b)
127
600
700
800
at I0 =0.4kA
at I0 =0.15kA
30
20
10
0
10
15
20
25
spike size I in A
600
20
400
10
200
c)
5
6
error in %
30
I(I0 =0.4kA)
(I 0 =0.4kA)
I(I0 =0.15kA)
(I 0 =0.15kA)
800
128
8.2.2. Discussion
Black-box modeling accuracy: For correctly determined arc parameters, the agreement between measured voltage waveform and recalculation by integration of equation (4.18) was extremely good.
Generally errors < 1 % were achieved. This conrms that the en-
129
ergy balance equation used is suitable to predict transient arc voltages. Larger errors resulted if stochastic uctuations occurred either
in the transient phase or between the instant where the stationary
point was determined and the subsequent current slope. Because of
the inability to identify very small uctuations, an average error of
2 4 % for the transient voltage resulted.
Arc uctuations: In nearly all constant current intervals and low
gradient slope intervals with duration > 100 s, stochastic arc uctuations were present. Up to 20 % of all current steps and spikes
showed voltage uctuations also during the transient phase. Intervals of the voltage curve that contain such uctuations are not
described correctly by equation (4.18) and can lead to wrong P and
values. Typically, a too high time constant results. Therefore, the
arc must be treated as a stochastic process and the quality of a parameter determination method depends on how well it is possible to
exclude or average out such uctuations. A large number of repetitive experiments usually results in a stray band of P (g) and (g).
This can be combined with an algorithm to identify large uctuations and exclude such sections from parameter determination. In
the present implementation this was rstly achieved by re-simulation
of the voltage waveform and secondly by identication of unexpected
maxima and minima in the arc voltage during the constant gradient
current slope.
Momentary quasi-stationary point: Transient voltages created by
current slopes are superimposed to the instantaneous quasi-stationary
state and not to any averaged stationary curve. This stands in agreement with the observations of [MSS80]. Unless a uctuation occurs
during a current spike, a higher quasi-stationary start voltage leads
to a constant voltage oset during the whole transient phase. For
this reason, must be determined with respect to the instantaneous
stationary characteristic. This is the case with higher probability,
the closer the quasi-stationary point is determined to the subsequent current slope, from which is extracted. For this reason, Rijantos [Rij75] approach to merge dierent gradient slopes from two
130
131
8.2.3. Conclusion
The novel arbitrary current source has made it possible to create
complex current waveforms including staircase-like and spike currents. By that, the arc parameters can be determined with a smaller
stray band of a lower number of experiments required. The arc power
loss P (g) is ideally determined from constant current sections. The
arc thermal inertia (g) can be determined either from moderate
current slopes or at constant currents subsequent to a very large
slope. Therefore, an optimum current waveform combines repetitive
constant current sections and high current gradients very closely together. This clearly favors a series of small current steps with as large
as possible current gradient and constant current sections > 5 for
arc parameter determination. In the experiments, such waveforms
were technically limited to a gradient of 15 kA/ms. At this gradient, current spikes superimposed on a current slope with gradient
< 1 kA/ms achieved much smaller scatter in P and . (g) could be
determined at 20 dierent conductance values with a stray band of
132
< 5 s from a single experiment. Best results were achieved for spike
amplitudes of the magnitude of the previous quasi-stationary current. The accurate transient voltage recalculation with error 2 4 %
for correctly determined parameters conrmed the validity of the
energy balance equation for the wall-stabilized arc investigated.
The improved determination accuracy now allows a better relative
comparison of similar arc chamber congurations using their blackbox parameters. By that, nozzle shape optimization of AC and DC
circuit breakers for dynamic processes, such as passive resonance or
self blast pressure build-up, is possible more easily.
9. Arc chamber
characterization
In this chapter, arcs with dierent chamber parameters are systematically characterized via their black-box arc parameters. In section 9.1 the comparison is performed for varying blow pressure, nozzle diameter, throat length, blow gas and nozzle material. The arc
cross-section variation at varied current amplitude and varied blow
pressure is investigated in section 9.2.
133
134
results. The improved parameter determination method D2 of section 7.2.3 has been used for arc parameter extraction.
Blow pressure: An experiment series with a wide throat nozzle
dT = 25 26 mm diameter was performed for various blow pressures
of pH = 0.915 bar. The large nozzle cross-section results in current
densities < 3 A/mm2 , where ablation is small [STCA06].
Nozzle Diameter: An experiment series was performed with cylindrical nozzles with a throat length LT = 50 mm for dierent nozzle
throat diameters of dT = 625 mm at a low and a high blow pressure
of pH = 1.5 bar and pH = 7.7 bar, respectively. The lling pressure
of the bottles was adjusted for each experiment, so that the desired
chamber pressure resulted at the end of the low current phase right
before the rising slope with spikes. The measured current range
0.1 2 kA in combination with dierent throat diameters resulted
in current densities of j = 1 70 A/mm2 .
Throat length: An experiment series with throat length varied
in the range of LT = 10 75 mm was performed for a) a narrow
nozzle dT = 7 8 mm without forced convection, b) a narrow nozzle
dT = 6 7 mm with moderate blow pressure pH = 3 bar and c) a
wide nozzle dT = 25 mm with strong blow pressure pH = 9 bar. In
the experiments, the electrode gap width was adapted to maintain
the throat outlet to downstream electrode distance at xout = 20
30 mm. Unlike case c) with j < 4 A/mm2 , in cases a) and b) the
current densities exceeded 40 A/mm2 at high currents and major
nozzle diameter change from wall ablation was observed.
Blow gas: An experiment series with cylindrical nozzles of diameter dT = 12 13 mm and throat length LT = 50 mm was performed
for three dierent blow gases: dry air (air), carbon dioxide (CO2 )
and helium (He). The absence of an encapsulation restricted the
candidate gases to those that could be freely released into the open
air. He was chosen because it has three times larger speed of sound
cS and ten times lower density at a given pressure than CO2 and
air. CO2 was chosen because it has a specic heat CP , 20% lower
than air. The parameters of air, CO2 and He in Table 9.1 for normal
135
air
26.2
1.2
331
1.01
100%
100%
CO2
16.8
1.8
259
0.84
64%
99%
He
156.7
0.16
965
5.19
600%
200%
136
PTFE is the standard nozzle material in HVAC arc chambers. Macor is a very temperature resistant glass-ceramic, which has already
been gased out in the process of manufacturing and has nine times
higher thermal conductivity than PMMA. It does not release gases
below 800 . Quartz glass has a softening temperature of 1683 C and
is therefore expected not to ablate.
137
2
i
u
a)
i in kA, u in kV
b)
c)
1.5
pB
1
0.5
time in ms
10
15
i
u
1.5
0.5
10
15
time in ms
20
i
0.9
0.8
0.7
0.6
0.5
11
11.1
11.2
11.3
time in ms
11.4
11.5
138
1.8
15.5bar
12.4bar
8.6bar
1.9bar
0.9bar
1.6
voltage U in kV
1.4
1.2
1
0.8
0.6
a)
0.4
0.2
0.4
0.6
current I in kA
0.8
1.2
thermal inertia in s
15
10
5
15.5bar
1.9bar
0.9bar
b)
0
0.5
1.5
2
conductance g in S
2.5
139
140
1.2
voltage U in kV
0.8
0.6
0.4
0.2
a)
0.5
1
current I in kA
1.5
16
thermal inertia in s
14
12
10
8
7.7bar, dT=6mm
7.7bar, dT=30mm
1.5bar, dT=6mm
1.5bar, d =8mm
T
1.5bar, d =11mm
T
b)
3
4
5
conductance g in S
141
142
1.5
25
a)
p =0bar
H
thermal inertia in s
voltage U in kV
LT=75mm
LT=50mm
LT=25mm
LT=10mm
0.5
0.5
1.5
thermal inertia in s
voltage U in kV
0.5
0.5
1.5
thermal inertia in s
voltage U in kV
10
20
15
10
5
25
pH=8.5bar
0.5
0
d) 0
1.5
e)
10
25
pH=3.1bar
15
0
b) 0
1.5
c)
20
0.5
1
1.5
current I in kA
f)
20
15
10
5
0
2
3
4
conductance g in S
143
144
1.5
HE
CO
voltage u in kV
air
a)
0.5
0.5
1.5
current i in kA
20
18
thermal inertia in s
16
14
12
10
8
6
4
2
b)
0.5
1.5
2
conductance g in S
2.5
145
and gas density, the gas type only shows a small inuence on U (I)
and . At I > 1 kA, where at U I-characteristics are present, the
arc voltages of air, CO2 and He are identical. At low currents, CO2
has the highest voltage and lies 300 V above that of He, which is
lowest.
pressure p in bar
146
pressure p in bar
a)
voltage u in kV
b)
current i in kA
c)
3
Macor
PMMA
Teflon
Quartz
2.5
2
1.5
10
15
20
10
15
20
10
15
20
10
time in ms
15
20
3
2.5
2
1.5
0.5
2
1.5
1
0.5
0
147
Macor
PMMA
PTFE
Quartz
1.2
voltage U in kV
1
0.8
0.6
0.4
0.2
a)
0.2
0.4
0.6
0.8
1
current I in kA
1.2
1.4
1.6
25
thermal inertia in s
20
15
10
b)
0.5
1.5
2
2.5
conductance g in S
3.5
148
149
arc diameter in mm
pressure p in bar
time in ms
150
arc
in mm2
120
A
100
AP
AE
80
AT
60
40
20
0
200
400
current I in A
600
800
151
arc
in mm2
60
50
40
30
AV
20
AP
AE
AT
10
0
4
6
8
pressure pH in bar
10
12
152
153
154
10. Discussion
In this chapter, the results of the previous chapters are discussed.
In the experiments, the theoretically expected falling, at and rising stationary U I-characteristics were conrmed. All experiments
have in common that the arc voltage decreases with increasing current at low I, followed by a constant voltage section and a rising
region were the voltage rises with current I. The at and rising
regions were not measured in all experiments, because in some cases
the transition current IT (current amplitude with minimum voltage)
was above the maximum current amplitudes which are possible in
the present setup.
Amongst all experiments, no case was found where the thermal
inertia could be inuenced independently from the stationary characteristic. The correlation of with U (I) is reasonable, considering
its denition as the ratio = E/P between stored energy and stationary cooling power. Typically, a rising (g) corresponds to a
section of falling U (I) characteristic and a falling (g) to a rising
U (I) characteristic. Another choice of arc parameters (e.g. P (g)
and E(g)) may allow better separation of physical eects.
155
156
10 Discussion
157
for the optical, the uid dynamical and the electrical cross-section
(cf. gure 9.10) and is in agreement with the theoretical predictions
of Lowke [LL75]. A higher current is required for the arc to reach a
specied cross-section. As a consequence, the current amplitude is
increased, at which wall-constriction becomes dominant and strong
ablation sets in. This could explain, why the falling characteristic
can be maintained up to higher currents, if higher blow pressures
are applied.
The observations in section 9.2.1 showed also an increase of arc
brightness at increasing blow pressure. This is an indication that
also the arc temperature increases with increasing blow pressure.
The observations showed a reduced cross-section at increased pressure. Assuming that a thermodynamic equilibrium was present before a blow pressure increase, the cross-section reduction at increased
pressure has the following consequence: the current cannot be carried anymore, so that additional ohmic heating is created. This
causes an increased plasma temperature and subsequently increases
conductivity, which partly compensates for the additional losses. A
new thermal equilibrium is found at smaller cross-section but higher
plasma temperature. It is probable that at the new thermal equilibrium the plasma pressure will be increased and the plasma density
reduced due to the increased plasma temperature.
The increased power loss at increased blow pressure (cf. gure 9.2a)) is dominated by an increased convective enthalpy ow. In
general, the enthalpy ow scales with hvs [LL75] and is increasing
if a higher temperature is reached due to the higher blow pressure.
However, to predict an absolute amount of increase, the change in
axial cross-section prole must also be considered, as it is a dominant
parameter for convective losses.
A reduced nozzle widening from test to test was observed for experiments with a higher blow pressure. According to the LambertBeer law, the radiation absorption in the cold gas vapor zone increases exponentially with the cold gas density and with the plasmawall distance. As a consequence, at higher cold gas pressure, a
158
10 Discussion
smaller fraction of radiation reaches the nozzle wall and can create
ablation. It is therefore expected that ablation becomes dominant
at higher current amplitudes.
Taking the eect on P and into account only for wide nozzles,
an increase of blow pressure increases the desired losses and thereby
improves the passive resonant current creation.
159
160
10 Discussion
161
162
10 Discussion
163
a similar slope as in the experiments with the strongly ablating materials, which is surprising. This is a clear indication that a rising
characteristic is not simply a consequence of ablation. Probably it is
caused predominantly by the wall constriction and only intensied
or shifted to lower current values by additional constriction caused
by the ablative pressure build-up.
Interestingly, arcs burning in Macor nozzles have the highest thermal inertia of all investigated nozzle materials (cf. gure 9.7b)). As
it is neither the most strongly ablating nor the most weakly ablating material, this must be the eect of other parameters than the
amount of ablation. Possible explanations are in the radiation absorbtion performance of Macor vapor or in its thermal performance.
Further experiments are required to clarify this point.
The observation that small values occurred predominantly at
high currents and that PMMA as the most strongly ablating material has the smallest thermal inertia (cf. gure 9.7) leads to the
speculation that the reduction of the thermal inertia at high current amplitude is caused by ablation. This would in principle be
desirable for passive resonance. However, a small is of no use if
the stationary U (I) characteristic is rising. Experiments with low
mostly showed also a rising characteristic already from low current
amplitudes onwards.
164
10 Discussion
165
166
10 Discussion
12
+
cooling power P in MW
10
a)
0.5
1.5
2
2.5
conductance g in S
3.5
0.5
1.5
2
2.5
conductance g in S
3.5
60
thermal inertia in s
50
40
30
20
10
b)
167
168
10 Discussion
8
C=30 F, L=200 H
C=8 F, L=300 H
C=5 F, L=1000 H
current iarc in kA
6
5
4
3
2
1
0
a)
10
15
20
time in ms
25
30
35
10
12
14
8
C=30 F, L=200 H
C=8 F, L=200 H
C=3 F, L=200 H
current iarc in kA
6
5
4
3
2
1
0
b)
6
8
time in ms
169
10m
10m
a)
10m
b)
4m
171
172
173
Figure 11.1.: Project overview: Identication of the state of the art gap
and specication of project work packages
174
175
176
177
178
12. Outlook
In section 12.1, further experiments are proposed, to clarify questions which are important for passive resonant breakers, but have
not been investigated within the scope of this work. Some part of
the experiments would be possible with the available experimental
tools but others would require extensions of the measurement setup.
The required modications of the model circuit breaker and of the
arbitrary current source are discussed in sections 12.2 and 12.3.
179
180
12 Outlook
181
182
12 Outlook
183
184
12 Outlook
185
A. Specication of the
arbitrary current source
A.1. Requirements
Precise arc characterization in its stationary and dynamic behavior
is the main application of the source designed. As shown in section 4,
an arc plasma is a very dynamic load. It changes its resistance by
orders of magnitude during a single experiment. A free burning arc
reaches low resistance < 0.1 , a strongly blown arc has resistance
values in the order of 1 10 . After extinction, the resistance increases to very high values. The dynamic arc behavior also varies
signicantly with the current slope applied. At low current slopes,
the arc follows its highly nonlinear stationary characteristic, at moderate slopes its resistance changes with time delays in the order of
1 30 s, at high current slopes the arc increases its voltage proportional to the current.
Traditionally, sinusoidal current waveforms have been used for arc
characterization. The disadvantage is that current gradient and current amplitude are linked to each other. The chosen approach for arc
characterization was to measure the transient characteristic and the
stationary characteristic independently from each other. In a single experiment, each conductance value must be created as steady
state by applying dierent current amplitudes. From each steady
state, a transient reaction to a fast changing current must be excited, from which the transient parameter is determined. Stationary
characterization requires constant current intervals. Transient characterization can only be performed during current slopes or at a
constant current subsequent to a slope. Therefore, the source must
187
188
189
wall
charge
Current Source
Source room
discharge
optical transmission
bushing
controler
model breaker
blow gas
190
13
14
14
15
11
16
17
10
11
12
Supporting equipment
11 IGBT gate driver
12 Isolation transformer ATT supply
Measurement equipment
7 current measurement shunt
8 voltage probe
9 ATT1 optical transmission In
10 ATT2 optical transmission Vc
Main equipment
1 capacitor
2 fuse
3 IGBT with antiparallel diode
4 freewheeling diode
5 coil
6 busbar to high voltage laboratory
Figure A.2.: Zoom view and component list of a single current source
module: front view (left top), back view (right top), side view (left
bottom), top view (right bottom)
192
RD
UC0
C3
RG3
C3
RG3
C3
RG3
SD
SG3
SG3
SG3
193
(A.1)
194
14
x 10
module inductance L in H
12
10
8
6
4
2
0
0
10 15 20 25 30 35 40 45 50 55 60 65 70
number of windings between taps n
195
A.2.4. Fuses
At overcurrent or thermal overstress, the IGBT could be destroyed
thermally and the possibility of explosion of the housing cannot be
excluded. An U RB303250 Ferraz Shawmut fuse with melting point
of 34 103 A2 s interrupts the current in case of an IGBT short circuit
fault and protects the other source components from excess current.
196
197
70
1kHz
2kHz
3kHz
4kHz
5kHz
6kHz
7kHz
8kHz
9kHz
10kHz
60
50
40
30
20
10
200
400
600
800
module set current In in A
1000
1200
198
of 2800 V only. It is assumed that the losses scale linearly with the
voltage. To account for the uncertainties, the allowed temperature
increase is limited to 80 C, which is 20 C below the thermal limit of
IGBT and diode.
In the IGBT, three types of losses occur. These are on-state conduction losses PIGBTcond. , turn-on losses PIGBTon and turn-o losses
PIGBTo . Conduction losses are evenly distributed in time, during
which the corresponding device is conducting. Switching losses arise
as instantaneous amounts of thermal energy and are converted to
mean power by multiplication with the switching frequency f :
( u
)
C
PIGBTon (c1 (Ion ) + c2 )
f
(A.3)
2800 V
2
PIGBTo (c3 Io
+ c4 Io + c5 )
( u
)
C
f
2800 V
2
PIGBTcond. d(VIGBT IN + RIGBT IN
)
(A.4)
(A.5)
(A.7)
199
optical conv.
optical conv.
charger
optical conv.
Gate Driver
experiment
A/D
A/D
ATT
ATT
optical transmission
optical conv.
optical conv.
A/D
A/D
IO card
optical conv.
ATR
optical conv.
ATR
UC
current source
IN
UC
D/A
D/A
3x
modifications
for upper band
controller
Original PID
controller
IN
Experiment computer
3x
ISN
OR
blow trigger
open trigger
Uarc
contact separation
pb
pH
Iarc
OFF
ON
oscilloscope trigger
<
AND
BNC connector
blow valves
3x
timer
PID
model breaker
Imax
Ramp
generator
DSP
IN
ISN
optical connector
controler
P
config.
optical
conv.
Oscilloscope
serial connector
200
A Specication of the arbitrary current source
201
202
203
204
205
207
13
11
10
20
20
18
19
15
15
12
16
17
Blow equipment
16 storage bottles
17 blow valve
18 filling valve
19 filling tubes
20 gas storage bottles
14
Measurement equipment
10 voltage probe +
11 voltage probe 12 hall current sensor
13 pressure sensor heating chamber
14 pressure sensor bottle
15 position sensor
Main equipment
1 nozzle
2 busbar high voltage
3 movable high voltage contact
4 high voltage electrode
5 low voltage electrode
(not visible, placed inside chamber)
6 bus bar low voltage
208
B Specication of the model circuit breaker
Figure B.1.: Model circuit breaker components: (right top) full view,
(left) zoom of arc chamber, (center bottom ) gas bottle bank, (center
right bottom) wall bushing and current sensor, (right bottom) zoom of
valves and pressurized bottles
Moving contact
209
Nozzle
Arc
Gas flow
Fixed contact
Valves
Pressure chamber
Pressure bottles
Figure B.2.: Schematic drawing of the model circuit breaker (from [S2])
210
Sonic speed in the nozzle throat can only be achieved if the crosssection of the pipe system from the storage volume to the nozzle
throat has everywhere larger cross-section than the nozzle throat
itself. For a maximum throat diameter of 30 mm the piping system
has to exceed a cross-section of 848 mm2 at its narrowest section.
To control the blowing, fast-operating valves are required, which are
not commercially available for suciently large cross-sections. It was
therefore necessary to use several valves in parallel. Eight VCH400
valves with 120 mm2 each and 2 ms opening time have been used.
To minimize the pressure loss from the storage volume to the
nozzle throat, a radially distributed placing of eight valves and eight
separate storage bottles was chosen. This has the advantage that
for smaller throat diameters the storage volume size and by that the
gradient of pressure decrease during the experiment can be varied
by selective operation of only some of the valves and bottles.
A pressure chamber was manufactured from a high-tensile aluminium block to merge the ows from the eight valves. This reduces
possible turbulences in the nozzle, caused by the gas ow splitting,
and achieves a more stable gas pressure. The wall thickness is overdimensioned, so that it can bear three times higher shear stress than
resulting at 50 bar. In this way, possible overpressure created by ablation in the nozzle will not exceed the mechanical limits. In addition, the circuit breaker is equipped with a housing of non breakable
PVC.
The uid-dynamic equivalent circuit of the model circuit breaker
is shown in gure B.3. In total, the eight bottles represent a total
storage volume VB = 12000 cm3 connected via an eective crosssection AV = 9.2 cm2 to the chamber. A piezo-resistive pressure
sensor that measures pB is placed at the neck of one bottle, and a
second one that measures pH , is placed inside the pressure chamber
at 10 cm distance from the nozzle throat.
Pressure
bottles
Pressure
chamber
211
Atmosphere
VR = inf
VB
TB
Sensor pH
Sensor P
B
VH
TH
Valves
effective
area AV
.
.
mass flow m V
Nozzle
effective
.
area
. AN
mass flow mN
pR
TR
C. Bibliography
[Ams77]
[And97]
P. Andr. The inuence of graphite on the composition and thermodynamic properties of plasma formed
in ablated vapour of PMMA, PA6-6, PETP, POM and
PE used in circuit-breakers. J. Phys. D: Appl. Phys.,
30:475493, 1997.
[AXW02]
[BF13a]
[BF13b]
[Bis54]
213
214
C Bibliography
breakers with adapted arc characteristics and energy absorbers. CIGRE Conf. Paris, 1970.
[Bla43]
[BM07]
[CCE11]
[CGR11]
215
[CIG25]
WG B4.54 CIGRE.
Guidelines for life extension of existing HVDC systems.
Avaliable from:
http://b4.cigre.org/WG-Area/B4-54-Guidelines-ForLife-Extension-of-Existing-HVDC-Systems, accessed on
2013.07.25.
[CJ97]
[CM97]
M. Classens and K. Moller. A computational uid dynamics simulation of high- and low-current arcs in selfblast circuit breakers. J. Phys. D: Appl. Phys, 30:1899
1907, 1997.
[CS91]
[CS05]
[CSC11]
[CW93]
[Des13]
Foundation Desertec.
Das DESERTEC-konzept.
http://www.desertec.org/de/konzept/, 2013.
216
C Bibliography
[Duf09]
[Fra11]
C. M. Franck. HVDC circuit breakers: A review identifying future research needs. IEEE Transactions on
Power Delivery, 26:9981007, 2011.
[FS06]
C. M. Franck and M. Seeger. Application of high current and current zero simulations of high-voltage circuit
breakers. Contrib. Plasma Phys., 46:787797, 2006.
[FW13]
[HG06]
[HJ11]
[IF08]
217
[Jen93]
J. Jenista. Static characteristics of a conduction and radiation dominated arc in nitrogen. Czechoslovak Journal
of Physics, 43:10891105, 1993.
[JHL+ 11]
[Kap11]
[Kc05]
[KCP09]
[KXLS02]
[Leu01]
C. Leu. Exitation and course of current chopping oscillation as an intermittent oscillation in SF6. SAP, Lodz,
9:1826, 2001.
[LL75]
[LO03]
218
C Bibliography
[Low79]
[Low84]
J.J. Lowke.
Theoretical predictions of ablationstabilised arcs conned in cylindrical tubes. J. Phys.
D: Appl. Phys., 17:11971212, 1984.
[Mar11]
[May43]
219
[MSDD04] C. Meyer, S. Schroder, and R. W. De Doncker. Solidstate circuit breakers and current limiters for mediumvoltage systems having distributed power systems. IEEE
Transactions on Power Electronics, 19:13331340, 2004.
[MSS80]
[Mul93]
[Nie87]
220
C Bibliography
[PRT75]
[Rie67]
[Rie90]
[Rij75]
H. Rijanto. Ein experimentelles Verfahren zur Bestimmung von Lichtbogenkenngrssen - Stromberlagerungsverfahren. Dissertation, TU Hannover, 1975.
[Rot80]
[RRU07]
[Rup79]
[Sch72]
[SFHK03]
M. Steurer, K. Frhlich, W. Holaus, and K. Kaltenegger. A novel hybrid current-limiting circuit breaker for
medium voltage: Principle and test results. IEEE Trans.
Power Del., 18:460467, Apr. 2003.
[Sha06]
221
[SK00]
[SNSD06]
M. Seeger, L. Niemeyer, M. Schwinne, and R. Dommerque. An integral arc model for ablation controlled
arcs based on cfd simulations. J. Phys. D: Appl. Phys.,
39, 2006.
[SO91]
[STCA06]
[SVKS03]
[TO02]
L. Tang and B. T. Ooi. Protection of VSC-multiterminal HVDC against DC faults. IEEE 33rd Annual Power Electronics Specialists Conference, 2:719
724, 2002.
222
[Too]
C Bibliography
The
Engineering
http://www.engineeringtoolbox.com.
2013.06.13.
Toolbox.
accessed on
M. M. Walter and C. M. Franck. Flexible pulsed DCsource for investigation of HVDC circuit breaker arc resistance. Int. Conf. on Gas Discharges and their Applications, page 170, 2010.
[WF11]
M. M. Walter and C. M. Franck. Inuence of arc chamber parameters on passive resonance circuit of HVDC
circuit breakers. CIGRE Session, Bologna, 2011.
[WG]
[Whi99]
[WKF13]
M. M. Walter, M. Kang, and C. M. Franck. Arc crosssection determination of convection stabilized arcs. ISH
Conference, Seol, South-Korea, September 2013.
[WLB11]
[YFO12]
223
[YMQB08] L. Yangping, Z. Manjiang, Z. Qing, and G. Bo. Research on arc ablation resistance of PTFE improved by
introducing inorganic ller. International Symposium on
Electrical Insulating, Sep. 2008.
[ZAM+ 02] J. L. Zhang, J. D. Anthony, B. Murphy, W. Hall, and
M. T. C. Fang. Computational investigation of arc behavior in an auto-expansion circuit breaker contaminated by ablated nozzle vapor. IEEE Trans. Plasma
Sci., 30(2):706719, 2002.
[ZoC26]
Curriculum Vitae
Name
Date of birth
Citizenship
Civil status
E-Mail
2007-2009
2004-2007
1999-2003
2010-2013
2009
2009
2008
2008
Personal data
Michael Max Walter
1984.03.31
Swiss, Zurich
married
[email protected]
Education
Master Studies, Electrical Engineering and Information
Technology, ETH Zrich
Bachelor Studies, Electrical Engineering and Information Technology
Swiss Maturity, Physics and Applied Mathematics,
MNG Rmibhl Zurich
Work Experience
ETH Power Systems and High Voltage Laboratories,
Scientic Assistant, Zurich Switzerland
MITSUBISHI Electric, Master thesis, Inuence of an
opening resistor and surge arrestor on UHV transient
recovery voltages, Osaka Japan
ABB Switzerland, Internship, Assembling of a UNITROL6080 prototype test cabinet, Turgi Switzerland
Mihajlo Pupin Institute, Internship, Telecommunications, Belgrade Serbia
ABB Switzerland, Semester thesis, Impedance Tomography Level Detection Systems, Dttwil Switzerland
225