Dynamical simulation of a Vacuum Switch with PSCAD

T. Wenzel T. Leibfried D. Retzmann

Inst. of Electr. Energy Syst. Inst. of Electr. Energy Syst.
and High-Voltage Technol. and High-Voltage Technol. Siemens AG
University of Karlsruhe University of Karlsruhe Erlangen, Germany
Karlsruhe, Germany Karlsruhe, Germany
wenzel@ieh.uni-karlsruhe.de leibfried@ieh.uni-karlsruhe.de dietmar.retzmann@siemens.com
Abstract - The switching behavior of vacuum
switches (VS) is different to that of other circuit break-
ers. There are several effects which have to be taken
into account, for example the current chopping, the di-
electric strength of vacuum gap, the arc voltage, the
quenching capability of high frequency currents, virtual
current chopping and prestrikes. In some earlier in-
vestigations the characteristics of vacuum circuit break-
ers had been simulated in different ways by taking into
account some of the effects mentioned above. There
are mainly two different approaches for modeling vac-
uum circuit breaker: The rst is to use an ideal circuit
breaker which is switched on/off after different criteria
have been checked. The second is to use a variable resis-
tor whose value is changed continuously. The VS model
discussed in this paper is modeled with PSCAD and
takes into account chopping current, dielectric strength
of vacuum gap, arc voltage and quenching capability.
It is represented by a variable resistance and is used to
simulate frequently switching of inductances and capac-
itances. The stress of VS is investigated and the energy
conversion in the vacuum tube: As the current through
a VS is owing from moment of contact separation un-
til next zero crossing, together with arc voltage, the en-
ergy conversion is very high, dependent on the moment
of contact separation.
Keywords - Vacuum Switch, Model, Simulation,
PSCAD
1 Introduction
Generally vacuum switches are limited in their
ratings, for example the maximum rated voltage is
about 24 kV and the maximum current is approxi-
mately 800 A [1]. These values differ slightly compar-
ing different manufacturers, but are in the same range.
VS offer the possibility to switch very often and much
faster compared to vacuum circuit breakers: As they
are opened and closed by a contactor, they only need
about three seconds to be switched again. A possible
application could be a frequently switched inductance
or capacitance. But in this case a problem occurs -
dependent on losses in VS, it can warm up very fast
and external cooling is possibly needed. In this paper,
the rst step is done to simulate the expected values of
power dissipation in VS while switching an inductive
load.
In the following section, the special characteristics of
VS during a switching process are presented. The
model of VS, which was created, is described after-
wards. Then the corresponding electric circuit is in-
troduced. Finally energy conversion in VS is assessed
and the next steps are discussed.
2 Switching characteristics of VS
There are mainly four different effects, which are
important to consider, if a model of VS should be cre-
ated. These are
Dielectric recovery of gap
Chopping current
High frequency quenching capability
Prestrikes
2.1 Chopping current and arc voltage
If VS is switched off, the contacts start to separate.
In the last point of contact, a very high current density
exists, which heats up contact material. Metal vapour
emerges and current is going on to ow through the
metal vapour arc. If it falls below an assigned value,
the metal vapour arc collapses [2]. This value is called
chopping current and strongly depends on contact ma-
terial. It can be calculated by [3]:
i
chopping
= (2f )
q
(1)
f = frequency in Hz
= current amplitude in A
= 6.2 10
16
s
= 14.3
q = (1 )
1
In the considered case, the used material is a
chrome copper alloy, and according to [4] or [5] the
chopping current is then lower than 5 A. This corre-
sponds very well to the calculated value by eq. (1).
While the contacts are open and current is owing,
there is a small voltage across the arc, which is about
10 30 V. This is because the arc is diffuse at smaller
currents [2][6]. In this investigation the current is ap-
proximately 800 A, so the arc voltage can be assumed
to be constant. It is not neglected, as voltage and cur-
rent form an electrical power dissipation, which heats
up the VS. If switching events occur frequently, tem-
perature can rise. Switching capability is degrading
with rising temperature.
2.2 Dielectric recovery mechanisms
It is important to distinguish between cold gap di-
electric strength and dielectric strength of vacuum gap
directly after current is interrupted. The rst one is
due to the fact, that opening of contacts of VS needs
some milliseconds. If contact distance of an open gap
is about 2 4 mm, a linear coherence between contact
distance and time can be assumed, which isnt possible
for greater contact distances (68 mm), that appear in
vacuum circuit breakers. As contact distance increases
nearly linear with time at small gaps, electric strength
does so too. It can be calculated by
U = a x (2)
a = Slope of dielectric strength in
kV
/mm
x = contact distance in mm
Typical rates of voltage rise are 40 60
kV
/ms. If
the voltage across VS exceeds dielectric strength, a
restrike will occur and the current continues to ow
through the opened VS. Worst case is switching off
VS shortly before a current zero. Then the current is
interrupted immediately and a transient recovery volt-
age of the electric circuit can exceed electric strength
of not completely opened contacts.
The second recovery mechanism is the dielectric
strength of gap immediately after current is inter-
rupted. In this case, the atoms of metal vapour have
to condense on the condensation shield [7]. During
this process, the dielectric strength of the gap increases
very fast. A typical value is 50
kV
/s [2]. After one mi-
crosecond dielectric strength rises with smaller slope.
If the transient recovery voltage across VS is higher
than this value, current interruption will be perma-
nently impossible, because this process reoccurs in ev-
ery attempt to interrupt the current.
2.3 High frequency quenching capability
A VS is able to quench high frequency currents.
They can arise if a restrike of VS occurred. After
the restrike, line current is superimposed by high fre-
quency currents, dependent on the connected electric
circuit. The resulting current can have many zero
crossings with different slopes. If a slope is smaller
than a particular value, current can again be interrupted
by VS shortly after restrike [3].
2.4 Prestrikes
When VS is switched on, contacts are moving to-
gether, which needs some milliseconds. As the di-
electric strength of gap decreases with smaller con-
tact distance, a breakdown can occur before contact is
completely closed. This effect is called prestrike and
causes high rates of voltage rise.
It should be mentioned, that there exists another
effect, which isnt part of this investigation, called vir-
tual current chopping. If the current in one line is
chopped after VS has been switched off, currents can
be induced into the other lines dependent on the elec-
tric circuit and coupling of the three phases. The su-
perimposition of line current and induced current can
result in a zero crossing in the respective phase if they
have opposite sign and are in the same range. VS can
interrupt this current according to its quenching capa-
bility of high frequency currents described above.
3 Simulation
3.1 PSCAD
The software used to create a model of VS is
PSCAD, which is an graphical user interface for
EMTDC (electro-magnetic transients simulation and
control simulation engine). The differential equations
of electric circuit and control circuit are solved by a
solver with xed stepsize. Size of computer memory
is limiting this value, because with decreasing step-
size, memory needed is increasing. An error message
will occur, if stepsize is chosen too small for a partic-
ular memory size. This causes a problem: as period-
ically switched events should be analyzed, a compro-
mise between accuracy and required memory has to
be found. The VS-Model, described in this paper, is
written in FORTRAN77 code. It is possible to create
own models and integrate them into component library
of PSCAD if they are not available. Parametrization
of VS is done by lling the input mask. Usually the
program runs without problems, and if there is one,
the error-description is adequate to get an idea of what
is wrong with the simulation case. This is only men-
tioned, because some of the other simulation software
differs considerably.
3.2 VS-Model
The VS-Model described in this paper is modeled
by a variable resistance. This is important for consid-
ering the arc voltage. Input parameters are VS-voltage,
VS-current and a switching command. The output pa-
rameter is the value of resistance. Furthermore there
are several parameters, which can be admitted by the
input mask.
Figure 1: Input mask of VS
Here electrical parameters are dened, such as re-
sistance in closed or opened position and maximum
voltage. Additionally, values of arc voltage, rise of di-
electric strength and rise of dielectric recovery can be
chosen. The value of resistance is changed dependent
on the different criteria discussed in section 2. The arc
voltage is modeled as a constant value (25 V) in order
to simplify the model. The corresponding value of re-
sistance is calculated by r = V
arc
/i
V S
. Both dielectric
recovery mechanisms are modeled by the equations
v
contact
(t) = 30
kV
/ms t (3)
effective only at moment of contact separation and
v
dielectric
(t) = 50
kV
/s t (4)
effective at every current interruption. These values
are limited by maximum voltage of VS (60 kV). Clos-
ing of contacts is modeled with 60 kV v
contact
. A
function checks
di
/dt of current at every zero cross-
ing to consider the current quenching capability of VS.
The value is set to a maximum of 80
A
/s. If this value
is exceeded, current will go on owing until next zero-
crossing.
3.3 Electric Circuit
Figure 2 shows the electric circuit, in which VS
is simulated. A voltage source provides a line to line
voltage of V = 18 kV. Its resistivity is 1 m. The
load is an inductance in delta connection. Its value is
L = 105 mH. The resulting current is I
rms
= 786 A.
Figure 2: Electric circuit with inductive load
As can be seen in gure 2, VS-Model returns a
value, which is directly assigned to the variable resis-
tance component. There are stray capacitances across
VS (12 pF) and from line to ground (9 nF). The resis-
tance (0.1 ) and inductance (50 H) represent proper-
ties of the supply line to the load. The resistors (0.1 )
in delta connection represent ohmic losses of the in-
ductances.
As mentioned in section 2.2, worst case for VS is be-
ing switched off directly before a zero crossing of the
current. This is shown in gure 3:
Figure 3: Restrikes in VS
In the upper graph, the dielectric strength (upper
curve with slope), and the transient recovery voltage
(TRV) is plotted. The lower graph presents the cur-
rent. As current reaches the chopping value (5 Ain this
case), it will be immediately interrupted. TRV now in-
creases until the dielectric strength of gap is exceeded.
Then value of resistance is set to r = V
arc
/i
V S1
, cur-
rent starts to ow but is superimposed by a high fre-
quency part. At the zero crossing of the high frequency
current, VS quenches the current and TRV arises again
with opposite sign. Dielectric recovery again is ex-
ceeded, current starts owing but this time without
zero crossing. In a 60 Hz grid, the current will con-
tinue to ow for approximately 8 ms until the next zero
crossing occurs.
3.4 Power dissipation and energy conversion
As VS is modeled by a resistor with variable value,
it will produce power loss which is converted into heat.
It can be calculated by
P
loss
(t) = i
2
(t) R
V S
= |v
V S
(t) i
V S
(t)| (5)
If P
loss
(t) is integrated over time, energy, which is in-
serted into VS can be obtained:
E(t) =

P
loss
(t) dt (6)
The next gure shows the power dissipation and en-
ergy in a VS when a restrike occurs after the VS was
switched off shortly before a zero crossing of current.
Figure 4: Voltage and current of VS while being switched off
The highest frequency, which occurs in the con-
sidered circuit, is approximately 25.6 kHz and a re-
sult of the interaction of load inductance and stray
capacitance to ground. The stepsize of the solver is
set to 1 s. The moment of contact separation can
be identied in the top graph, when dielectric strength
(v
recovery
) starts to increase. The small negative peak
in the same moment shows the unsuccessful current
interruption in line 1. Meanwhile, current of line 3
(I
V S3
) reaches zero and is successfully interrupted.
This is because the contacts are opened completely and
the rise of TRV (V
V S3
) doesnt exceed the dielectric
recovery of the gap. Simultaneously, currents of line
1 and 2 are inuenced due to the missing possibility
of current ow in line 3. As their values fall under
the chopping current, they are successfully interrupted,
too.
Power dissipation can be seen in the third graph. The
peak in the beginning is a miscalculation, but has no
big effect on energy in VS, which can be seen in the
last graph as a function of time. Power dissipation
reaches a peak value of

P
loss
= V
arc

i
V S1
24 kW.
The value of the curve in the last graph at t = 0.07 s is
the energy which is converted in VS after one unsuc-
cessful current interruption. It is E 98 Ws.
Figure 5 shows repeated switching operations every
three seconds. This is what special vacuum switches
are able to handle. The simulation stepsize in this case
was set to 10 s. The worst case is regarded here, too.
Every switching-off operation is taking place shortly
before current zero in the respective phase. It can be
seen, that prestrikes do not have a big impact on the
energy increase when the load is inductive. The cur-
rent is increasing slowly by reason of inductance, and,
together with the arc voltage, power loss is compar-
atively small. Like before, the increase of energy is
approximately 98 Ws for every switching-off process.
Figure 5: Periodically switching of VS
3.5 First considerations about heat balance
The power loss, which occurs during the switching
operations heats up the contacts of VS so that switch-
ing capability is reduced. P
loss
is an equivalent to the
heat ux

Q. The value of E, calculated above, is an
equivalent to heat Q. In the example mentioned above,
an inductive load is periodically switched and in worst
case, E 98 Ws has to be transferred out of VS
at every switching-off operation. There are different
heat transport mechanisms, for example thermal con-
duction, heat emission and convection. Referring to
[10], heat transport consists mainly of thermal transfer
into contact material. As vacuum is the surrounding
medium, there will be no heat transfer by convection.
The Temperature of copper contact material in the root
point of an arc is approximately 1353 K [2], which
is slightly higher than the melting temperature. Now
it has to be investigated, if therefore heat dissipation
occurs by heat emission. If yes, the only face which
could absorb emission is the condensation shield. It
surrounds the complete contact gap and is connected
to one of the two contacts. So even if some heat was
absorbed by this shield, it would have to be transferred
towards the respective contact by thermal transfer. The
major part of heat dissipation will be thermal conduc-
tion through both contacts of the VS. Therefore a high
thermal conductivity is necessary, which is the case
when copper or a chrome copper alloy is used [2]. In
order to estimate, how fast heat can be transferred out
of the VS, the next step will be a nite element cal-
culation of this problem. It can be reduced to a two-
dimensional problem, because the VS-tube is rotation-
ally symmetric. The calculated values of P
loss
can be
taken as the input parameters for the heat source.
4 Conclusion
A model of VS was created which takes into ac-
count special characteristics of vacuum switches like
chopping current, high frequency quenching capabil-
ity, dielectric recovery mechanisms and prestrikes. Arc
voltage is considered, too, in order to estimate the
value of power loss in VS connected to a electric cir-
cuit, where an inductive load is switched on and off
periodically. This VS-Model can now be used, to sim-
ulate overvoltages at VS connected to different electric
circuits as well as power dissipation. By calculating
this value, which is proportional to the heat ux, it is
possible to get input parameters for a nite element
calculation of the geometry of VS. This assessment is
necessary in order to estimate the inuence of the ther-
mal transport mechanisms, thermal ux and heat emis-
sion. The results of the simulation will then be used
to analyze, if VS can be switched more often than ev-
ery three seconds, and if additional cooling devices are
necessary and generally able to assure a sufcient heat
discharge.
REFERENCES
[1] Siemens Mittelspannungsgeraete Katalog HG
11.21 - Vakuumschuetze 3TL, 2002
[2] H. J. Lippmann, Schalten im Vakuum - Physik
und Technik der Vakuumschalter, Berlin, Offen-
bach, VDE-Verlag GmbH, 2003
[3] J. Helmer, M. Lindmayer, Mathematical Mod-
eling of the High Frequency Behavior of Vac-
uum Interrupters and Comparison with Measured
Transients in Power Systems, XVIIth Interna-
tional Symposium on Discharges and Electrical
Insulation in Vacuum, Berkeley, 1996
[4] Siemens Vakuum-Schaltroehren fuer die Mit-
telspannung, Produktbeschreibung
[5] R. D. Garzon, High Voltage Circuit Breakers -
Design and Application, Marcel Dekker, Inc.,
New York, Basel, 2002
[6] E. F. J. Huber, Wiederzuendmechanismus im
Vakuumschalter bei der Unterbrechung Netz-
frequenter Kurzschlussstroeme, Dissertation,
Zuerich, 1999
[7] S. Temborius, M. Lindmayer, Stromnullverhal-
ten unter Vakuum-Lastschalter-Bedingungen,
15. Fachtagung Kontaktverhalten und Schalten,
Karlsruhe, 22.09.-24.09.1999
[8] EMTDC - The Electromagnetics Transients
& Controls Simulation Engine, Users Guide,
Manitoba HVDC Research Centre Inc., 2005
[9] PSCAD - Electromagnetic Transients, Users
Guide, Manitoba HVDC Research Centre Inc.,
2005
[10] R. Orama, Breakdown Phenomena of a Vac-
uum Interrupter after Current Zero, Interna-
tional Conference on Power System Transients
2001, Rio de Janeiro, Brazil, June 2001