This action might not be possible to undo. Are you sure you want to continue?
System Components and Experimental Verification
Master of Science Thesis
Abey Daniel & Samson Gebre
MSc in Electric Power Engineering 60p (Degree of Civilingenjörsexamen, eq. 180p)
Department of Electric Power Engineering
Division of Energy and Environment
CHALMERS UNIVERSITY OF TECHNOLOGY
Göteborg, Sweden, 2008
THESIS FOR THE DEGREE OF MASTERS OF SCIENCE
Analysis of Transients in Wind Parks: Modeling of
System Components and Experimental Verification
Abey Daniel & Samson Gebre
Performed at: ABB Corporate Research
Västerås,Sweden
Advisor: Ambra Sannino
ABB Corporate Research
Västerås,Sweden
Examinar: Torbjörn Thiringer
Division of Electric Power Engineering
Department of Energy & Environment
Chalmers University of Technology
Department of Electric Power Engineering
Division of Energy and Environment
CHALMERS UNIVERSITY OF TECHNOLOGY
Göteborg, Sweden, 2008
Analysis of Transients in Wind Parks: Modeling of System Components and
Experimental Verification
ABEY DANIEL & SAMSON GEBRE
© ABEY DANIEL & SAMSON GEBRE, 2008
Department of Energy and Environment
Chalmers University of Technology
SE412 96 Göteborg
Sweden
Telephone + 46 (0)31772 1000
Email: abiy1882@yahoo.com
gebrsa@yahoo.com
Chalmers Reproservice
Göteborg,Sweden 2008
i
ABSTRACT
This thesis deals with analysis of transients and experimental verifications in wind parks.
The focus is on accurate modeling of system components such as cables, vacuum circuit
breakers and transformers which make any branch in a wind park. Models are also
validated where possible. Although a lot has been done and fairly good models of cables
and VCBs have been developed, it seems that a unified wide band model of transformers
which works well for broad range of frequencies has been the most difficult to develop.
Critical as it is in transient studies, modeling of such a transformer has been extensively
treated in this thesis.
In the first part of the thesis, a frequency dependent model which represents high
frequency effects like wave propagation, damping and reflection is used to model the
cables using PSCAD/EMTDC. The model is also validated experimentally. Secondly, a
statistical model which emulates the occurrence of reignitions, prestrikes and current
chopping behaviors of the VCB during opening or closing operations is dealt with.
Finally, a terminal model of a transformer which represents its wide frequency response
is developed based on experimental measurement of admittance matrix over a wide range
of frequencies. The curve fitting algorithm is used to approximate the admittance matrix.
An RLC equivalent network is realized and implemented in the time simulation software
PSCAD/EMTDC.
ii
PREFACE
This thesis work was conducted at ABB corporate research. We would like to express our
deepest gratitude to our supervisor Ambra Sannino for providing us with the opportunity
to work in an excellent academic environment and for her brilliant ideas, guidance and
inspiration.
We are deeply grateful to our examiner Torbjörn Thiringer, for the valuable discussions,
excellent guidance and for the care with which he reviewed the thesis report.
Special thanks goes to Tarik Abdulahovic, who is responsible for helping us with the
modeling part and his constant guidance, suggestion and valuable input to this work. He
was always there to listen and to give advice.
We have learnt a great deal from those who have worked with us over the months and
gratefully acknowledge our debt to them, especially we are extremely grateful to Henrik
Breder, for the warm welcome ,assistance, generosity, and advice we received.
We would also like to acknowledge all of the other people who have helped and
encouraged us. Regrettably we will not be able to recall all of them, but here is a good
start. We want to thank Nilanga Abeywickrama,Yuriy Serdyuk,Dierk Bormann, Leif
Hederström, Josef Medvegy.
Last but not least, we are deeply indebted to our friends and families for their support and
advice.
iii
CONTENTS
ABSTRACT........................................................................................................................ i
PREFACE.......................................................................................................................... ii
CONTENTS...................................................................................................................... iii
ABBREVIATIONS........................................................................................................... v
1 INTRODUCTION..................................................................................................... 1
1.1 Problem description ............................................................................................ 1
1.2 Overview of earlier work.................................................................................... 2
1.3 Aim of this thesis ................................................................................................ 2
1.4 Thesis outline...................................................................................................... 2
2 TRANSIENT OVERVOLTAGES .......................................................................... 4
2.1 Introduction......................................................................................................... 4
2.2 Classification....................................................................................................... 4
2.3 Shunt Capacitor Switching Transients................................................................ 5
2.4 Traveling Waves ................................................................................................. 7
3 CABLE MODELING............................................................................................. 10
3.1 Introduction....................................................................................................... 10
3.2 Cable Parameter Selection in PSCAD/EMTDC............................................... 11
3.2.1 Core Parameters ............................................................................................. 12
3.2.2 Insulation and Semiconducting screen.......................................................... 13
3.2.3 Sheath conductor........................................................................................... 14
3.3 Testing Model ................................................................................................... 14
3.3.1 Measurement SetUp .................................................................................... 14
3.4 Results and Analysis ......................................................................................... 17
3.5 Conclusions....................................................................................................... 21
4 VCB MODELLING................................................................................................ 22
4.1 Introduction....................................................................................................... 22
4.2 Phenomena causing over voltages in VCB....................................................... 22
4.2.1 Current chopping .......................................................................................... 22
4.2.2 Voltage Escalation ........................................................................................ 23
4.2.3 Virtual Current Chopping ............................................................................. 24
4.2.4 Prestrikes....................................................................................................... 24
4.3 Modeling of VCB ............................................................................................. 25
4.3.1 Current chopping .......................................................................................... 25
4.3.2 Dielectric strength......................................................................................... 26
4.3.3 High Frequency Quenching Capability ........................................................ 27
4.4 Simulation Results ............................................................................................ 27
4.4.1 Effect of Arcing Time.................................................................................... 28
4.4.2 Effect of rate of rise of dielectric strength ..................................................... 29
4.4.3 Effect of high frequency quenching capability.............................................. 30
4.4.4 Breaker unsuccessful and high frequency current successful operation........ 32
4.4.5 Prestrikes ...................................................................................................... 33
4.5 Conclusions....................................................................................................... 34
5 TRANSFORMER MODELING Background and Theory ............................... 35
5.1 Introduction....................................................................................................... 35
iv
5.2 Wide Band Modelling of Power Transformers................................................. 36
5.2.1 Measurements ................................................................................................ 36
5.2.2 Rational Approximation of Frequency Responses by Vector Fitting............ 40
5.2.3 Time Domain Implementation....................................................................... 44
6 TRANSFORMER MODELINGExperiments and Results................................ 47
6.1 Introduction....................................................................................................... 47
6.2 Measurement Setup........................................................................................... 47
6.3 Measurement Results ........................................................................................ 51
6.4 Rational Approximation of Admittance Matrices of The Transformers .......... 53
6.5 Model Validation .............................................................................................. 60
6.6 Accuracy of the Model...................................................................................... 65
6.7 Conclusion ........................................................................................................ 66
7 CONCLUSIONS AND FUTURE WORK.............................................................. 67
7.1 Conclusions....................................................................................................... 67
7.2 Future Work...................................................................................................... 68
REFERENCES................................................................................................................ 71
APPENDICES................................................................................................................. 74
Appendix A. Implementation of 3phase Transformer model in PSCAD/EMTDC..... 74
Appendix B. Network Realization................................................................................ 78
v
ABBREVIATIONS
VCB Vacuum Circuit Breaker
BIL Basic Insulation Level
MV Medium Voltage
WP Wind Park
EMTP Electro Magnetic Transient Program
PSCAD Power System CAD
EMTDC Electromagnetic Transient with DC
TOV Transient Overvoltage
EM Electromagnetic
RRDS Rate of Rise of Dielectric Strength
1
Chapter 1
INTRODUCTION
1.1 Problem description
Nowadays due to the development of modern circuit breakers; vacuum and SF6 for
example, the switching overvoltage has become a more sever problem, to some extent
caused by the fact that the current interruption capabilities of these breakers have
improved significantly [3]. These switching transient over voltages have the potential to
result in large financial losses. Transient overvoltages and other power quality
disturbances cause billions of dollars of losses each year due to damage to equipment and
loss of production [1]. Worldwide many transformer insulation failures have been
reported caused by switching transients [5], [6], while those transformers had previously
passed all the standard tests and complied to all quality requirements. These phenomena
can occur in both distribution and transmission networks. The problem is generally
associated with the fast transient over voltages produced by restrikes and prestrikes
during the opening or closing of a switching device. Especially the vacuum circuit
breaker (VCB), which shows a high ability of interrupting high frequency currents of
several hundreds of kHz, may cause switching overvoltages with fast rise time and thus
degrading the insulation and eventual failure due to the cumulative effect of voltage
transients of even less magnitude than the BIL.
The use of a cable network and VCB in medium voltage (MV) networks could be a
source of insulation failure due to the fast voltage transients caused by multiple restrikes
of VCB. There is a similarity between wind parks (WP) and industrial MV networks in
this regard as both use VCB as switching device and consists of a cable network. Hence
there is a concern of transient overvoltages (TOV) in wind parks as well. One example
could be Denmark’s Horn Rev, one of the world’s biggest off shore wind farms built to
date, where all transformers and generators had to be moved to shore for maintenance.
Fast switching, high voltage transients are suspected to have contributed to the failure [7].
Due to the unpredictable threat of TOV in WPs and large MV systems it is worth
studying fast switching TOV in wind parks and MV systems experienced due to the
unavoidable switching action of VCB. To analyze these phenomena, an experimental set
up was designed in ABB Corporate Research with threephase cable representing a part
of a wind park, including one feeder with one turbine under test and the adjacent tower
and adjacent feeder to reproduce a unique wind farm in a lab. The experimental set up is
also useful to study large cable systems such as MV undergrounding and industrial
systems.
2
1.2 Overview of earlier work
A number of investigations have been done related to this work. It is not the intention of
this section to summarize these valuable works, but the interest in this section lies on
overview of the two works by Daniel Mireanu and Maialen Boyra as this work is the
continuation of these two. The first work provides an analysis of TOVs and their
propagation in cable systems [31]. This investigation is performed through the use of two
systems: an industrial steel mill electrical system and a wind farm. Cables and vacuum
circuit breakers, two essential components in any MV cable system, are discussed in
detail and modeled in PSCAD/EMTDC. The two systems are then simulated and, where
available, the results are compared to measured data for verification. In the second work,
TOVs in cable systems are characterized through both simulations in PSCAD/EMTDC
and laboratory experiments which characterize a section of wind power plant [30]. A
single phase laboratory set up was used which includes two single phase cables, a VCB
in between, and supply side and load side transformers. The laboratory system was
modeled. Analysis of measurement and simulation results were also made for different
loading conditions.
1.3 Aim of this thesis
The aim of this work is to study and characterize high frequency transients in wind parks.
A crucial part is accurate modeling of the cable, vacuum circuit breaker and transformer
which comprise a branch in any wind park. As the work is a continuation of [30] and
[31], a particular interest is to refine and improve models of the aforementioned
components. Particularly, a different approach is used to model the transformer than
those used in [30] and [31], one that enables representing the transformer by using a
single and unified terminal model for a very wide band of frequencies.
1.4 Thesis outline
Chapter 1 gives a general background of the thesis. An overview of previous literature
by is presented, [30] and [31]. The aim of this thesis work is defined as well.
Chapter 2 presents an overview of TOVs, how they are classified, their causes and how
they propagate through the system.
Chapter 3 presents modelling of a power cable which is to be used later in the simulation
of the whole system. The model derived in this chapter proposes a method to account for
the effects of stranding of the conductor and semi conducting layers for use in
PSCAD/EMTDC. Comparison of results obtained between the model and measurement
data during cable discharging is also included.
3
Chapter 4 Describes the VCB model treated in this paper. The model simulates most of
the effects (phenomena) of a VCB. Results obtained from model testing are also included
in this chapter.
Chapter 5 and 6 propose the modelling of a transformer by measuring the admittance
matrix in a wide frequency domain. The admittance matrix experimentally determined is
subjected to a rational function approximation by vector fitting. A transformer model
compatible with electromagnetic transient programs (EMTP) can then be obtained by
network realization of the rational function.
Chapter 7 gives conclusions of the thesis work and suggests future work.
4
Chapter 2
TRANSIENT OVERVOLTAGES
2.1 Introduction
TOVs in electrical transmission and distribution networks result from the unavoidable
effects of lightning strikes and network switching operations. A TOV can be defined as
the response of an electrical network to a sudden change in network conditions, either
intended or accidental, (e.g. a switching operation or a fault) or network stimuli (e.g. a
lightning strike). A transient is a natural part of the process by which the power system
moves from one steady state condition to another [2]. Its duration is in the range of
microseconds to milliseconds. A brief explanation and analysis of transients can be
found in references [3, 4]. In this chapter, the fundamental concepts of TOVs with
especial emphasis on switching transients is presented to facilitate understanding of the
next chapters as the transients dealt with in this work are of this type caused by switching
operation of VCBs.
2.2 Classification
TOVs are broadly classified as impulsive and oscillatory.
(1) Impulsive transients
The most common cause of impulsive transients is lightning. Lightning is associated with
the discharge of a large current. This current can reach up to 200kA. The overvoltage
developed in this case is limited by the impedance of the system seen by the lightning
current. This overvoltage can reach several MV. This situation usually causes power
system faults due to insulation failure which in turn causes a supply interruption and
voltage sags throughout the distribution network.
Impulsive transients are characterized by their rise and decay time and peak values.
Lighting current and its corresponding voltage rises to its peak in a time which varies
from less than a microsecond to 10 or 20 microseconds and then decays in a few hundred
microseconds [3].
(2) Oscillatory transients
Switching operations within the network are a major cause of oscillatory overvoltage
transients. Such operations include switching of circuit breakers to clear faults, switching
5
of utility capacitor banks and switching of distribution feeders to rearrange the network
for maintenance or construction.
A switching overvoltage or as they are some times called switching surge is generated
due to the interaction of the inherent elements(inductance, capacitance and resistance)
associated with an electric network .When current flows through an inductance, it
produces magnetic flux. Any effort to change the magnetic flux (i.e. the current) will be
opposed by the inductance, which is manifested by the generation of a counter EMF in
the inductance in such a direction as to keep the magnetic flux (and the current) in the
inductance constant. Therefore, when a circuit breaker tries to interrupt a current, a
voltage is developed by the system inductance to oppose the change in current. The faster
a switch tries to interrupt a current, the higher the resulting switching surge voltage is.
Theoretically this voltage can be infinite if the current through the inductor is snapped off
instantaneously. However, as every electrical system can withstand a maximum level of
voltage, the system maximum voltage strength is reached causing a discharge and a
temporary or a permanent damage. In practice, all electrical systems have capacitances.
When the current is interrupted in an electrical system, the current through the inductance
flows through the capacitance, thus transferring the energy from the inductance to the
capacitance. The energy originally stored in the inductance is transferred back and forth
until it is finally damped by the system resistance.
In a system with large physical extension, such as a power transmission line where the
inductance and the capacitance are distributed, traveling voltage and current waves are
generated during a switching operation. Dangerously high voltages can be generated by
multiple reflections at the end of the line.
2.3 Shunt Capacitor Switching Transients
Shunt capacitors are used extensively in power transmission and distribution systems as a
means of supplying reactive power for voltage support. These capacitors are implemented
in the system in order to control the system voltage, to increase the power transfer
capability, to reduce equipment loading, and to reduce energy costs by improving the
power factor of the system. They are switched in and out depending on the level of
support needed. When a discharged capacitor is energized, the voltage of the bus bar to
which it is attached will momentarily collapse as the voltage across a capacitor can not
change instantaneously. This is followed by an oscillatory transient recovery voltage
which can be as high as 2p.u. An example of a capacitor switching waveform is shown in
Fig.2.2. The same effect can occur when a capacitor is switched off and a restrike occurs
(i.e. the circuit recovery voltage causes the dielectric between the switch contacts to break
down and capacitor current is reestablished). Capacitor switching transients can be
magnified to quite high values as they pass to a lower voltage level if certain network
conditions are met. This can occur when transients originating in the medium voltage
(MV) distribution network move into the low voltage (LV) network and there are power
factor correction (PFC) capacitors present at LV consumer’s installations. Such a
situation is given in Fig. 2.1 which shows an MV/LV network supplied by an MV feeder
6
R
1
L1
#1 #2
MV Bus
B
R
K
L2
LV Bus
B
R
K
C1 C2
P
+
j
Q
with inductance L1. A switched capacitor C1 is connected to the MV bus bar. The series
combination of the outgoing MV feeder and the MV/LV transformer has an inductance of
L2. The PFC capacitor C2 is connected to the LV bus bar. Magnification occurs when the
predominant frequency of the switching transient is approximately equal to the resonant
frequency of the LV system. This condition is mathematically represented by
Fig.2.1 Power system network used in PSCAD/EMTDC to illustrate
magnification of capacitor switching transients.
2 2 1 1
C L C L ≅ (2.1)
where L1 is the system source inductance seen from MV bus bar, C1 is the capacitance of
switched capacitor, L2 the inductance of stepdown transformer and associated MV feeder and
C2 is the capacitance of LV PFC capacitor.
The overvoltage associated with this phenomena can reach to a peak value up to 4 times
the corresponding power frequency system voltage.
7
Over voltage and current transients due to shunt capacitor switching
time[s]
0.00 0.10 0.20 0.30 0.40 0.50
...
...
...
3.0
2.0
1.0
0.0
1.0
2.0
3.0
V
[
p
.
u
]
Voltage
10.0
10.0
I
[
p
.
u
]
Curent
Fig.2.2 Voltage and current resulting from capacitor switching simulated.
2.4 Traveling Waves
The parameters of transmission lines, cables, as well as transformer and generator
windings are distributed. A characteristic of a circuit with distributed parameters is its
ability to support traveling transient waves of voltage and current. The influence of the
distributed parameters on the propagation of TOV depends on the frequency content of
the waves. Higher frequency transients will be more affected by the stray and distributed
inductance and capacitance of the system than lower frequency transients.
The current and voltage waves travel in both directions from the point of excitation or
disturbance. The ratio of the amplitudes of the voltage and the current waves on a
8
transmission line or cable is called the characteristic impedance Z
0
of the line and is for a
lossless line given by
C
L
Z =
0
, O (2.2)
where L and C are the distributed inductance and capacitance respectively of the line or
cable. Typical values of the characteristic impedance vary between 300 to 500 O for
overhead lines and for cables from 30 to 60O [2].
The velocity of propagation of the waves for a lossless line is given by
LC
v
1
= , m/s (2.3)
and depends on the medium of propagation. It is near the speed of light (3x10
8
m/s) for
overhead lines, and between one half and two thirds of this value for underground cables
[2].
Like all other waves, traveling waves initiated by disturbances in power systems also
have classic wave characteristics like reflection and refraction.
Reflection and refraction of traveling waves
When voltage and current waves propagate in transmission lines or cables, there is a strict
proportionality between the two. The proportionality constant is the characteristic
impedance of the line or cable. When a wave arrives at a point of discontinuity, which
could be an open end, an underground cable or transformer, where the characteristic
impedance changes, to keep this proportionality, two new wave pairs are generated, one
reflected back superimposed on the incident wave and another transmitted beyond the
discontinuity. The amplitudes of the reflected and refracted waves are such that the
voltages to current proportionalities are preserved for each.
The consequence of these discontinuities could be the production of complex sets of
traveling waves which add and subtract, possibly causing quite high voltages at some
points.
The magnitudes of the reflected and refracted voltage waves at a junction point with
characteristic impedances
a
Z and
b
Z on the incident side and refractive side, respectively,
can be quantified as
1 2
V
Z Z
Z Z
V
b a
a b
+
−
= (2.4)
9
1 3
2
V
Z Z
Z
V
b a
b
+
= (2.5)
where
1
V is the incident wave,
2
V is the reflected wave and
3
V is the refracted
(transmitted) wave.
b a
a b
Z Z
Z Z
+
−
is called the reflection coefficient and
b a
b
Z Z
Z
+
2
is the
refractive coefficient.
The reflected and transmitted currents are given by
a
Z
V
I
2
2
− = (2.6)
b
Z
V
I
3
3
− = (2.7)
10
Chapter 3
CABLE MODELING
3.1 Introduction
This chapter treats modeling of a real power cable in PSCAD/EMTDC. The objective is
to find a close representation (model) of a real cable that best depicts its response. The
model can then be used to study the effect, at a wider range of frequencies, of different
network operations (both normal and abnormal) in a real medium voltage power network
of which the cable is a part. A model is developed in PSCAD/EMTDC which suitably
works not only at power frequency but also at higher frequencies. The focus is on high
frequency effects like EM transient propagation, reflections, skin effect, etc. Both single
phase and three phase cables have been modeled. These models are later to be used to
simulate and study effects of TOVs in real power networks during switching operations,
earth faults, etc.
It can be sufficient to model a power cable by using simple RLC Hnetwork to study its
response at a single frequency such as steady state. However, for higher frequencies,
frequencydependent models where the frequency dependence of the distributed cable
parameters such as capacitance, inductance and resistance per meter are well accounted
for should be used to include high frequency consequences such as skin effect, etc.
PSCAD/EMTDC provide three options for modeling of a power cable [11]:
i) Bergeron model
This model treats the cable as an infinite number of H sections of L, R, and C. It
represents accurately only one frequency.
ii) Frequency dependent (mode) model
This model represents well the frequency dependence of the distributed
parameters over a wide range of frequencies, but uses a constant transformation of
internal matrices. Thus, it may not be good if the conductors are not ideally
transposed.
iii) Frequency dependent(phase) model
This model also represents well the frequency dependence of the distributed
parameters over a wide range of frequencies. The frequency dependence of
11
r1
Cable # 1
r2
r3
r4
d [m]
0[m]
Conductor
Insulator 1
Sheath
Insulator 2
internal transformation matrices is well taken care of. This model is, therefore, the
best of all three. This model has been used in our work.
3.2 Cable Parameter Selection in PSCAD/EMTDC
The basic parameters used to represent cables and transmission lines are series impedance
matrix Z and shunt admittance matrix Y [9], [10].
L R Z ω ω ω j ) ( ) ( + = (3.1)
C G Y ω ω ω j ) ( ) ( + = (3.2)
where R, L, G, C are series resistance, series inductance, shunt conductance and shunt
capacitance respectively. Both Z and Y are functions of frequency . All the commonly
used programs for simulation of transients, EMTPtype programs including
PSCAD/EMTDC have dedicated support routines for calculating an electrical
representation of cable system in terms of the series impedance matrix Z and admittance
Y [9], [10].
In PSCAD/EMTDC the cable parameters are calculated based on the geometry and
material property of the cable specified by the user. There is a major difference, however,
on the geometric representation of the actual cable from that used in PSCAD/EMTDC as
the later does not account for the presence of a semiconducting screen that is present in
the actual cable and it assumes that the core conductor is a homogeneous solid conductor
which is different from the core conductor made up of many strand conductors in the
actual cable. The situation is further complicated by the fact that some of the geometrical
dimensions of the cable are not given by the manufacturers. Fig.3.1 below shows the
difference between the actual cable and its representation in PSCAD/EMTDC.
Fig.3.1 Actual cable versus PSCAD/EMTDC representation.
As it is clearly shown in Fig.3.1 PSCAD/EMTDC requires the geometric parameters for
the core, semiconductor and sheath. An actual cable however includes a conductor, an
12
inner semiconducting screen insulator, an outer semiconducting screen sheath among
others. Accordingly the user needs to decide how to represent the following cable
features which PSCAD/EMTDC fails to represent.
1) Core stranding
2) Inner and outer semiconductor layers
3) The wire screen
The following sections are dedicated to obtaining the cable parameters for simulation in
PSCAD/EMTDC. An attempt will be made to represent the above features.
3.2.1 Core Parameters
The main parameters of the core to be given to PSCAD/EMTDC are the resistivity and
radius. The actual cross section area of the core given by the manufacturer is different
from the area of circle obtained using the radius of the conductor given by a manufacturer
for stranded core design. It is obvious that the area of the core needed for resistance
calculation in PSCAD/EMTDC is obtained based on the radius specified, hence there is a
need to make some form of correction to account for the fill factor as to give the correct
resistance of the conductor. Stranded conductors can be modelled in PSCAD/EMTDC in
two different ways. The first way is to model it as a solid conductor by increasing the
resistivity by a factor equal to the inverse of the fill factor as given by the formula
below
c
A
r
2
1
'
π
ρ ρ = (3.3)
where ´ is the corrected resistivity to account for the space between the strands, is the
resistivity of the core material, r
1
is the radius of the conductor given by manufacturer,
and
c
A is the efficient cross section area of the core.
The second way is to model it as a hollow cylinder whose cross section is the difference
of areas between two circles with area equal to
c
A as shown by Fig.3.2.
The inner radius in this case is given by
r
in
=
π
c
A
r −
2
1
. (3.4)
If the manufacturer provides the DC resistance for the core R
DC
, the corrected resistivity
can alternatively be calculated as
l
r
R
DC
2
1
'
π
ρ = (3.5)
where l is cable length. Resistivity of commonly used core materials is given in Table
3.1.
13
r1
Cable # 1
r2
r3
r4
rin
d [m]
0[m]
Conductor
Insulator 1
Sheath
Insulator 2
Fig.3.2 Alternative approach of dealing with stranding of core.
Table 3.1 Resistivity of commonly used core material.
MATERIAL COPPER ALUMINIUM
p[O.m]
1.72E
8 −
2.82 E
8 −
3.2.2 Insulation and Semiconducting screen
The main insulation of highvoltage cables is always sandwiched between two semi
conductive layers. Unfortunately, none of the EMTPtype programs for instance
PSCAD/EMTDC in our case allows the user to directly specify the semiconducting layers
[9], [10]. These must therefore be introduced by a modification of the input data. As
explained in [9] and [10], this is done by allowing the insulation to extend between the
core conductor and sheath conductor, and increasing the permittivity proportionally to
leave the capacitance unaltered. This data conversion procedure is summarized as
follows.
1) Calculate r
2
, r
1
plus the sum of the thickness of the semi conducting screens and
the main insulation. This radius is given as outer radius of the insulation to
PSCAD/EMTDC.
2) Calculate the corrected permittivity c
1 r
as
0
1 2
1
c 2
) / ln(
π
ε
r r C
r
= (3.6)
where C is the cable capacitance and c
0
=8.854E
12 −
.
If the capacitance is unknown, the permittivity is corrected as
) / ln(
) / ln(
1 2
1
a b
r r
rins r
ε ε = (3.7)
14
where a and b are the inner and outer insulation radii respectively and
rins
ε is the
permittivity of the insulation material.
3.2.3 Sheath conductor
PSCAD/EMTDC allows the user to specify the sheath conductor. When the sheath
conductor is made of wire screens, as it is the case with the cable we are dealing with, the
best way to model it is with a tubular conductor having a cross section area equal to the
total wire area A
s
, inner sheath radius of r
2
and outer radius r
3
given by
2
2
3
r
A
r
s
+ =
π
(3.8)
The other input data that needs to be specified to PSCAD/EMTDC that is worth
mentioning is the resistivity of the sheath material. Materials commonly used as sheath
conductor along with their resistivity is given in Table 3.1.
3.3 Testing Model
The cable models developed in the foregoing sections are tested to characterize different
frequency dependent effects like EM propagation, damping (due to skin effect and
proximity effect) and reflections. The cables at the wind cable lab were charged to 800V
and discharged through a COMgap to excite a very steep fronted voltage at one end and
measure the transients at the other end for different lengths and arrangements of the cable
system. The measured waveforms are compared with results obtained by simulating the
same operations using the frequencydependent cable models developed in 3.2 using
PSCAD/EMTDC.
3.3.1 Measurement SetUp
Fig.3.3 Experimental set up for cable discharge with step.
15
The experimental set up is based on a variable DC source, a current limiting resistor of
1kO, ELFA AC240 comgap (see Table 3.3), and LeCroy Oscilloscope for measuring
waveforms(see Table 3.2). The cable was charged by increasing the variable DC voltage
source until the comgap break down and then the cable is discharged through the comgap.
In other words, a step excitation is applied between core conductor and the sheath when
the gap discharges, which propagates on to the other end of the cable until it is finally
damped by the system resistance. In the set up shown in Fig.3.3 U1 and U2 represent the
input and out put measured points.
The experiment was done on a 620m cable while it was rolled on drum (see Fig. 3.4) and
on cable sections with 242m and 484m lengths of the cable lab set up after the cable had
been cut and installed.The specifications of the cable is given in Table 3.4
Table.3.2 Specifications of LeCroy 9354A Digital Oscilloscopes.
BANDWIDTH 500MHZ
Channels 4
Sample Rate 2GSa/sec
Memory Depth 50K pt/sec
Sensitivity 2 mV/div to 5V/div
Number of Bits 8 8 bits
Input Impedance 1 MOhm/14pF or 50Ohm± 1%
Main time base lowest 1 ns/div
Table.3.3 ELFA AC240 Comgap product specification.
PARAMETER MINIMUM MAXIMUM
DC Breakdown Voltage 425V 
Impulse BreakdownVoltage 800V
Insulation Resistance
10
10
O
Capacitance 1PF
Operational Temperature 40°C +125°C
AC FollowOn Current >300A
16
Fig.3.4 three phase 24 kV 620m Sea cable on drum and its cross section.
Table.3.4 Specification three phase sea cable.
CORE MATERIAL ALUMINIUM SOLID CONDUCTOR
Core cross section 240mm2/phase
Core radius 9mm
Main insulation material Mainly of PE
Screen conductor Copper
Screen cross section 85mm2
Armor material Steel
MATERIAL CROSS
SECTION
AREA
[
2
mm ]
RADIUS
[mm]
THICKNESS
[mm]
Core Aluminum 3x240 9
Main insulation XLPE
6
Outer insulation
22.5
Inner/outer
semiconducting
screen
0.5/1
Sheath conductor Copper 3x98
17
r1
Cable # 1
r2
r3
r4
d [m]
0[m]
Conductor
Insulator 1
Sheath
Insulator 2
The experimental set up in Fig.3.3 was simulated in PSCAD and the cable was modeled
according to the modeling procedure given in section 3.2.The parameters of the cable
model are accordingly given below.
1
r =9mm
2
r =
1
r +Main insulation thickness+ sum of inner and
outer semiconducting screen thickness
2
r =9mm+6mm+0.5mm+1mm=16.5mm
2
2
3
r
A
r
s
+ =
π
=
2
2
) 5 . 16 (
85
mm
mm
+
π
=17.3mm
a =
1
r +thickness of inner semiconducting screen
a =9mm+0.5mm=9.5mm
b=a+main insulation thickness
b =9.5mm+6mm=15.5mm
Corrected permittivity
1 r
ε
) / ln(
) / ln(
1 2
a b
r r
rins
ε = =2.85
r
4
=22.5mm
3.4 Results and Analysis
When a cable is energized with a fast fronted impulse, a traveling wave is likely to be
incepted. One important requirement for the formation of traveling waves is that the
duration of the wave should be significantly less than the time taken to traverse the
medium. Some characteristics inherent to traveling waves in cables and transmission
lines are reflections, propagation delay and damping.
The waveforms in Fig.3.5 and 3.7 show the results of the experimental tests for two
different cable lengths. Fig.3.6. and 3.8 are results of simulations in PSCAD/EMTDC
using the frequency dependent cable model. Subsequent Figs 3.9 and Fig3.10 are “zoom
ins” of these in order to show the pulses more clearly.
The shorter the cable length, the more frequent reflections the cable accommodates and
the higher the frequency of the pulses. This can be observed in Fig 3.9(a) and Fig.3.10
(a) which present the measured waves at the end of a 242m and 484m cables. The
frequency of oscillation for the 242m cable is approximately 154kHz which is almost
double that for the 484m cable which is 76.7kHz. The traveling time of the reflected
waves between the discharging COMgap and the cable open ends also becomes shorter
for the shorter cable since the propagation velocity for the same type of cable is identical
whatever the length is. This is demonstrated by the results in Fig 3.9(a) and Fig.3.10 (a).
The traveling time is equal to a quarter of the propagating pulse cycle. The propagation
delay for waves to reach the open end of the 242m cable has been measured to be 1.634us
18
delay for waves to reach the open end of the 242m cable has been measured to be 1.634us
and is 3.26us for the 484m cable. The cable oscillation frequency indirectly affects the
damping of the propagating transient due to the conductor ac resistance being increased
by the skin effect. Thus, damping of the waves increases with frequency. This can be
seen in the same figure where we have quicker dying of the reflected waves on the 242m
cable than on 484m cable due to higher frequency.
Similar results are obtained by simulating the same operations on the cables using the
frequency dependent model in 3.2. Fig 3.9(b) and Fig.3.10 (b) signify a higher frequency
of oscillation, shorter propagation delay and faster damping for the 242m than for 484m
cable. Comparable figures are also found for the pulse frequency and propagation time
which are in reasonable agreement with those found in the measurements. Measured and
simulated values are given in Table 3.5 for comparison.
Similar results are found by computing the above parameters from the propagation
velocity of the cables. The propagation velocity can be calculated from the inductance L
and capacitance C per meter of the cable as given in (2.3) and L and C have been
calculated from the cable property and geometry (see section 3.3.1) to be 0.121nH/m and
0.269nF/m respectively yielding
v=
nF/m 269 . 0 * nH/m 121 . 0
1
=175m/us .
This gives a traveling time of
v
S
T
p
= (3.9)
where S is the length of the cable yielding
p
T =
us m
m
/ 175
242
=1.38us, for the 242m cable length and
p
T =
us m
m
/ 175
484
=2.76us, for the 484m cable length
It should be noted that the oscillation frequency is slightly higher in the simulations. One
source of this error could be due to the additional capacitance to ground from the
measuring probes which may have decreased the frequency in the measurements as
f
0
=1/2LC where f
0
is the oscillation frequency.
19
100 0 100 200 300 400 500 600 700 800 900
1000
800
600
400
200
0
200
400
600
800
1000
time[us]
v
o
l
t
a
g
e
[
v
]
Charging end L11
Measuring point L21
100 0 100 200 300 400 500 600 700 800 900
1000
800
600
400
200
0
200
400
600
800
1000
time[us]
v
o
l
t
a
g
e
[
v
]
Charging end L11
Measuring point L12
Table.3.5 Comparison of measured and simulated values.
Cable length Oscillation freq.
(Simulation)
[MHz]
Oscillation freq.
(Measurement)
[MHz]
Travel Time
(Simulation)
[us]
Travel Time
(Measurement)
[us]
242m
0.181
0.154
1.4
1.634
484m
0.08962
0.0767
2.789
3.26
Fig.3.5 Cable energizing at L11 and measuring point at L21 (measured).
Fig.3.6 Cable energizing at L11 and measuring point at L21 ( simulated).
20
100 0 100 200 300 400 500 600 700 800 900
1500
1000
500
0
500
1000
1500
time[us]
v
o
l
t
a
g
e
[
v
]
Measuring point L12
Charging end L11
100 0 100 200 300 400 500 600 700 800
1500
1000
500
0
500
1000
1500
time[us]
v
o
l
t
a
g
e
[
v
]
Charging end L11
Measuring point L12
Fig. 3.7 Cable energizing at L11 and measuring point at L12(circuit breaker
end) measured.
Fig.3.8 Cable energizing at L11 and measuring point at L12(circuit breaker
end) simulated.
21
100 50 0 50 100 150 200
1000
0
1000
time[us]
v
o
l
t
a
g
e
[
v
]
Measuring point L12
Charging end L11
100 50 0 50 100 150 200
1000
0
1000
time[us]
v
o
l
t
a
g
e
[
v
]
Charging end L11
Measuring point L12
100 50 0 50 100 150 200
1000
0
1000
time[us]
v
o
l
t
a
g
e
[
v
]
Charging end L11
Measuring point L21
100 50 0 50 100 150 200
1000
500
0
500
1000
time[us]
v
o
l
t
a
g
e
[
v
]
Charging end L11
Measuring point L12
a) Measured b) Simulated
Fig.3.9 Measured and simulated waves for L=484m (Ts=0.02us).
a) Measured b) Simulated
Fig.3.10 Measured and simulated waves for L=242m (Ts=0.02us).
3.5 Conclusions
Of the three frequency dependent cable models provided in PSCAD/EMTDC, the
Frequency Dependent (phase) Model is the best model. It was used with certain
adjustments on data entered to account for the presence of semiconducting layers,
and core and/or sheath stranding.
The results of the experiment show that good agreements have been achieved
between measured and simulated values in the frequency of oscillation and
propagation delay validating the cable model proposed in this regard.
Small disagreements in these parameters are, however, noticed and the authors
believe they can be due to unaccounted capacitance of the measuring cables and
additional stray capacitances and inductances in the real system.
Although it has been found that both simulations and measurements show an
increase in damping of the EM transients with increasing frequency of oscillation
as should be, quicker damping is observed in reality than in the simulations. This
is either the simulation model does have certain parameters estimated incorrectly
or the real system has additional resistance or conductance elements than
considered in the model.
22
Chapter 4
VCB MODELLING
4.1 Introduction
At transmission level, the most modern switching devices used are SF6 breakers, while at
medium voltages VCBs are primarily used [20]. VCBs have very suitable properties like
capability of interruption of both power frequency and high frequency currents, and good
gap breakdown voltage recovery properties. Their ability to interrupt these high
frequency currents, however, may lead to multiple gap breakdowns (reignitions) which
may be the cause for very severe overvoltages under certain network conditions. Many
MV component failures have been found to be caused by TOVs due to VCBs. A good
model that replicates the VCB characteristics like current chopping capability, dielectric
breakdown strength and high frequency current quenching capability is therefore crucial
to study in order to understand the occurrence of these TOVs. This chapter deals with the
general theory of modeling a VCB in PSCAD/EMTDC that takes into account the above
characteristics. A VCB for use in MV networks as wind parks has been modeled based on
[16], [17]. The breaker model is then used in a test circuit to see if it actually
characterizes the original VCB properties stated above.
4.2 Phenomena causing over voltages in VCB
There are four reasons why overvoltages occur in vacuum circuit breakers: current
chopping, multiple reignitions, virtual current chopping and prestrikes [13].
4.2.1 Current chopping
If the contacts of a VCB start opening close to a zero of a power frequency current, the
arcing current across the contacts becomes unstable as to be extinguished. Detailed
treatment of this instability can be found in [18]. This behavior of the breaker to interrupt
a small but non zero current just before the zero of a power frequency current is called
current chopping. The mean value of the chopping current depends on the contact
materials, the level of the load current, and the instant of contact separation. After the
current is chopped and the arc is extinguished, a transient recovery voltage (TRV)
appears across the VCB contacts. The shape and level of the TRV depends on the actual
chopping current and the resulting capacitance, inductance and resistance on the load
23
side. Higher chopping current produces higher TRV as do high inductance on the load
side. (4.1) shows how the chopping voltage (TRV) is related to the inductance and the
chopping current. If the TRV developed becomes greater than the dielectric strength of
the breaker, the gap may reignite.
C
L
i U
ch ch
=
(4.1)
Fig.4.1 TRV waveform and chopped current.
4.2.2 Voltage Escalation
After the extinction of the arc at current zero of the power frequency current, a race
begins between the TRV developed between the breaker contacts and the dielectric
strength of the gap. If the chopping happens very close to current zero, the gap separation
by the time the arc has extinguished is small; the dielectric strength of the gap is,
therefore, accordingly small. The TRV may in this case exceed the gap voltage strength
and the gap may reignite and start conducting. This arcing current is of very high
frequency superimposed on the line frequency current. If the high frequency current is
higher or equal to the power frequency current in magnitude, they may add up to cause
several high frequency arc current zeros through the breaker. The VCB may be able to
interrupt this current at one of the zeros if certain conditions are met. Again, if the TRV
exceeds gap breakdown strength, a restrike occurs causing a substantially high frequency
current. This process may repeat until the gap can withstand the TRV. Each subsequent
restrike produces higher and higher load side over voltages due to the increased and
continuous transfer of energy between the inductance and the capacitance on the load side
24
due to the restrike and extinction respectively, as shown in Fig.4.2. This process of
growth of TRV is termed as voltage escalation.
Fig.4.2 Restrike currents and voltage escalation.
4.2.3 Virtual Current Chopping
In three phase circuits, the flow of the high frequency currents due to ignition in one of
the phases can lead to injection of the current into the other phases through inductive and
capacitive coupling and can lead to forced current zeroes in these phases [16]. This
mechanism is called virtual current chopping when reignitions occur due to these current
zeros. What happens is if the magnitude of the high frequency current that couples to the
other phases is higher than the peak of the line frequency current, the current in the
phases will pass through zero several times before its natural zero. This may cause
several reignitions and interruptions escalating the voltage. Compared to normal current
chopping, virtual current chopping can be much higher, as high as several hundred
amperes compared to 3 or 4A [17].
4.2.4 Prestrikes
While reignitions are for opening breakers, prestrikes are for closing breakers. Before
the contacts come to a physical contact, during closing, the voltage between the contacts
may get higher than the dielectric withstand voltage of the gap which diminishes as the
contacts get closer. In this case, the gap breaks down before the contacts are actually
25
closed and an arcing current is established, and a prestrike is said to have occurred.
Depending on the external network parameters, a high frequency arcing current can occur
with several zero crossings. Multiple prestrikes might happen if the breaker is able to
interrupt some of these and the process repeats. Consequently, multiple high slope
transient step voltages are produced.
4.3 Modeling of VCB
In simulating a network for transient study, the components involved should be
accurately modelled. However obtaining accurate VCB Model which simulates all the
phenomena discussed above is difficult. For one thing these phenomena are statistical in
nature and secondly the authors believe it is beyond the scope of this paper as modelling
all of those phenomena is a thesis work by itself.
The generic model incorporates different stochastic properties inherent to the breaker
operation to control the actual state of the breaker. The properties are as specified in [19]:
• Random nature of arcing time
• Current chopping ability
• The characteristic recovery dielectric strength between contacts when
opening
• High frequency quenching capability at current zero.
The model introduced in this paper is similar to the models on references [16, 17, and
19]. It simulates the following phenomena:
1. Current chopping
2. Dielectric strength
3. High Frequency quenching capability
4.3.1 Current chopping
The actual chopping current is nondeterministic. However earlier research established
different mean chopping levels [19] for different load currents and contact material. The
mean chopping current
CH
I is estimated by (4.2) according to [19]:
q
CH
I I ) . . . (
^
β α ω = (4.2)
where =2f is angular frequency of power supply, Î=Amplitude of the 50Hz current, u, µ,
q are parameters dependent on contact material. For modern VCBs which use Cu/Cr
contact these constants are given by
26
u=6.2x10
16 −
1
) 1 (
−
− = β q =−0.07512
µ=14.3
To account for the statistical nature of the chopping current, the mean chopping current
determined by (4.2) is varied by a normal distribution curve with a standard deviation of
15%.
4.3.2 Dielectric strength
In modelling the breakdown strength of the contact gap, the cold gap breakdown is
considered in this paper as this is the relevant breakdown strength when the preceding arc
current does not exceed several hundred Amperes. The cold withstand voltage
characteristic of the VCB is function of the contact distance. It is strongly dependent on
the speed of contact separation. For short gaps linear variation of dielectric strength with
distance can be assumed [14].
B t t A U
bd
+ − = ) (
0
(4.3)
where t
0
is moment of contact opening, A and B are constants. The constant A is the rate
of rise of dielectric strength. The values of the constants A and B in (4.3) for three
different typical dielectric strength characteristics are shown in Table 4.1, as proposed by
[14]. However those parameters can be adjusted later using experimental data of the
VCB.
The value of the dielectric strength so calculated is assumed as the mean value of a
Gaussian distribution with a standard deviation of 15% as break down phenomena are
statistical in nature.
Table 4.1 Constants for equation (4.3) (Reproduced from [14]).
BREAK DOWN
VOLTAGE TYPE
(BV TYPE)
A(V/MS) B(KV)
High 17 3.4
Medium 13 0.69
Low 4.7 0.69
27
4.3.3 High Frequency Quenching Capability
When the transient recovery voltage developed across the breaker exceeds the dielectric
withstand voltage, reignition occurs. High frequency current superimposed on power
frequency will be conducted through the breaker. VCBs are capable of interrupting this
high frequency current at one of the zero crossings. However, successful interruption
depends on the rate of change of this current at current zero. Interruption of the high
frequency current will not occur if the rate of change at current zero is higher than the
critical value of quenching capability given by (4.4) [14,16,17,19]
D t t C
dt
di
+ − = ) (
0
(4.4)
where t
0
is moment of contact separation, C and D are constants.
The values of the constants for three typical
dt
di
characteristics are given by Table 4.2 as
given in [14].
Table 4.2 Constants for equation (4.3) (Reproduced from [14]).
4.4 Simulation Results
To characterize the various behaviors of the VCB model discussed and validate them, the
test circuit in Fig.3 based on [16] is used. To this end, the effects of arcing time, rate of
rise of dielectric strength, high frequency current quenching capability, and prestrikes
have been considered. Total unsuccessful operation of the breaker where it is unable to
interrupt all current zeros and successful interruption at high frequency current zero are
also demonstrated.
dt
di
TYPE
C(A/MS
2
)
D(A/ MS)
High 0.034 255
Medium 0.31 155
Low 1 190
28
Fig. 4.3 Test Circuit.
4.4.1 Effect of Arcing Time
Arcing time is defined as the time between the initiation of breaker opening to the next
power frequency current zero. In the first case, the breaker is opened at t
0
=0.104s and
the next current zero of the power frequency current comes at t=0.105s. This gives an
arcing time of AT=1ms. Fig.4.4 (left) shows the resulting breaker voltage and current for
this case. The breaker can not break the arc successfully at the first current zero and
successful interruption occurs at the second current zero when the breaker has developed
enough with stand voltage at a bigger gap distance. Fig. 4.4 (right) shows breaker voltage
and current for a larger arcing time of AT=5ms, all other parameters remaining the same.
For this case the beaker is opened at t
0
=0.1sec., 4ms earlier and the breaker has
developed enough gap withstand voltage by the time of the first current zero and thus it is
able to interrupt the arc successfully at this current zero.
The higher the arcing time, the longer the time the breaker gets to develop a bigger and
sufficient dielectric strength to break the arc.
29
Fig.4.4 Unsuccessful (left) and successful (right) interruptions at first current zero for
arcing times of 1ms and 5ms respectively.
4.4.2 Effect of rate of rise of dielectric strength
The rate of rise/fall of the dielectric strength depends on how fast the breaker contacts are
opening or closing respectively. All other parameters remaining constant, if the rate of
rise is increased, the breaker reaches a higher dielectric strength faster and the probability
that the breaker quenches the arc at first power frequency current zero increases. This has
been demonstrated in Fig. 4.5(left) which shows breaker voltage and current for a rate of
rise of A=4.7kV/us in which case there is unsuccessful interruption of arc at the first
current zero. A successful interruption occurs at the following power frequency zero.
Now, if the rate of rise is increased to A=17kV/s as shown in Fig.4.5 (right), the breaker
has successfully interrupted the arc at the first current zero.
30
Fig.4.5 Unsuccessful (left) and successful (right) interruptions at first current zero for
RRDS A=4.7kV/us, and A=17kV/us respectively.
4.4.3 Effect of high frequency quenching capability
The variation of quenching capability has been considered in this case by taking the high
and medium quenching capabilities keeping all other parameters the same for both cases.
As it is shown in Fig.4.6 and Fig.4.7, the higher the quenching capability, the larger the
number of reignitions will be. This is because of the high transient recovery voltage
across the breaker when interrupting currents with high
dt
di
.This is also clearly shown in
the zoomed plots of Fig.4.6 and Fig.4.7 for the two cases. On the other hand, if the
quenching capability is smaller, the current may not be chopped during the high
frequency period, this leads to an increase in conduction period of the high frequency
component (smaller number of reignitions). The latter case might cause failure of
interruption.
31
Fig.4.6 Plots showing the effect of high frequency quenching capability for arcing time of
5ms and high
dt
di
type.
Fig.4.7 Plots showing the effect of high frequency quenching capability for arcing time of
0.005 and medium
dt
di
type.
32
4.4.4 Breaker unsuccessful and high frequency current
successful operation
When a vacuum circuit breaker reignites during opening operation, voltage escalation
occurs and interruption process can terminate in one of the three cases [19].
a. The breaker fails to interrupt the high frequency current, and interruption is
accomplished in one of the next power frequency zeros.
b. The breaker fails to interrupt and may cause harm to itself and/or the equipments
connected to.
c. The breaker can successfully interrupt at one of the high frequency current zeros.
The first case is shown in Figures (4.4, 4.5, 4.6 and 4.7). Case two and three are
illustrated in Fig.4.8 and Fig.4.9.Transition from case two to three was achieved by
increasing the breakdown voltage when the breaker is fully open from 20kV in case two
to 26kV in three.
Fig.4.8 Unsuccessful breaker operation.
33
Fig4.9 Successful interruption of high frequency current.
4.4.5 Prestrikes
During closing operation of a vacuum circuit beaker, due to the diminishing dielectric
strength of the gap as the contact separation decreases, the gap may break and an arc is
established before real galvanic contact occurs. For the same reasons as for reignition,
during opening operation, a number of prestrikes can happen in the gap during closing.
The effect of pre strikes is illustrated in Fig.4.10, where very steep breaker step voltages
are shown to have developed, due to the high frequency current quenching.
Fig.4.10 Pre strike during contact closing.
34
4.5 Conclusions
Interruption of small but non zero power frequency inductive currents and
subsequent high frequency currents due to re ignitions and prestrikes by
VCBs causes generation of very steep TOVs.
A statistical model of the VCB based on modeling of current chopping,
dielectric strength and high frequency current quenching capability was used
in a generic test circuit to characterize properties of an actual VCB and
expected results were found.
35
Chapter 5
TRANSFORMER MODELING Background and
Theory
5.1 Introduction
In the previous sections it was shown that the transients caused by the switching action of
the VCB have a wide range of frequencies. During the prestrike and restrike of a VCB for
example highfrequency oscillations with short rise time are produced. Consequently the
transformer model proposed in this work should adequately represent the transformer
over a wide range of frequencies from 50/60Hz to MHz. However, the transformer
representation used in PSCADE/EMTDC does not take into account the high frequency
behavior of the transformer like the other devices modeled in previous sections. Many
papers have tried to deal with this problem. A model to simulate the high frequency
behavior of a power transformer is presented in [22], and [23].
A number of high frequency transformer models have been proposed by different authors.
Internal winding models and terminal models are the two broad categories.
Internal winding models
Very fast TOVs voltages can generate a voltage oscillation inside the transformer
windings. Current chopping and reignition of a VCB give rise to fast electromagnetic
transients in power networks, which may result in high frequency oscillations and
overvoltage stresses on the terminals of network components. The effects of such
transients on transformer windings are particularly important. Most of the time, the
greatest problem is the internal resonance which occurs when the frequency of the input
surge is equal to some of the resonance frequencies of the transformer. These
overvoltages are characterized by a very short rise time. Most of the time, resonant
overvoltages can cause a flashover from the windings to the core or between the turns.
The inter turn insulation is particularly vulnerable to highfrequency oscillation, and
therefore, the study of the distribution of interturn overvoltages is of essential interest
[21].
The internal winding model is generally applicable if calculating interturn voltages in
windings of the transformer during transient conditions are of interest in the study.
Besides at the design stage, an accurate calculation of overvoltages which would stress
winding sections is very important for economical use of the insulation material. On the
other hand, in HV and EHV networks, prediction of possible voltage stresses at critical
points is required for efficient insulation coordination and protection. These models
36
assumed that the winding configuration, geometry, material constants and other
construction details are known.
Terminal models
In the majority of problems related to power system transients, transformers are
considered as system components. Transformer terminal models are based on the
simulation of the frequency and/or time domain characteristics at the terminals of the
transformer. A number of studies dealing with modelling of the power system equipment
based on terminal measurements have been published, example [22], [23].External
measurements can be performed either in the frequency or time domain. Frequency
domain data consist of terminal impedance or admittance characteristics, whereas time
domain data is basically acquired from impulse voltage tests.
The model proposed in this thesis belongs to terminal model as the interest is on the
transient over voltages at the terminals associated due to interaction of the different
components in the system, for example switching operation of the VCB.
5.2 Wide Band Modelling of Power Transformers
The highfrequency behaviour of power transformers is characterized by several
resonance points due to inductive and capacitive effects from the windings, tank, and
core. This behaviour should be included in any overvoltage study where the high
frequency characteristics of the transformer
is of significance (e.g., transferred over voltages and resonant over voltages)[23].
In this thesis a method for the development of a wide band frequency dependent black
box model of a transformer is provided.
This model is based on experimental determination of the admittance matrix in a wide
frequency range and subjecting the measured admittance matrix to an approximation with
a rational function by the method of vector fitting. The rational function so obtained can
then be realized in to an RLC net work for time domain simulation in PSCAD/EMTDC.
The following sections are devoted to describing the modelling techniques used to obtain
the model.
5.2.1 Measurements
Consider a transformer with n –terminals as shown in Fig5.1. The voltage and current at
the terminals can be expressed as
37
) ( ) ( ) ( s V s Y s I = (5.1)
where Y(s) is the admittance matrix, V(s) is terminal voltage vector and I(s) is terminal
current vector all in frequency domain. Y(s) is an n by n symmetrical matrix while I(s)
and V(s) are vectors of length n.
Fig. 5.1 Nterminal Transformer Model.
(5.1) can be expanded as
(
(
(
(
(
(
(
¸
(
¸
(
(
(
(
(
(
(
¸
(
¸
=
(
(
(
(
(
(
(
¸
(
¸
n nn n n
n
n
n
V
V
V
Y Y Y
Y Y Y
Y Y Y
I
I
I
.
.
.
...
.
.
.
.
.
.
.
.
.
...
...
.
.
.
2
1
2 1
2 22 21
1 12 11
2
1
(5.2)
It can be seen from (5.1) that applying a voltage
j
V to terminal j and zero voltage to the
remaining terminal will produce the
th
j column of ) (S Y , where the element
ij
Y can be
found by taking the ratio
i
I to
j
V .This is demonstrated for the first column of ) (S Y .
Apply
1
V to terminal 1 and short circuit all other terminals, (5.2) can be reduced to
1 11 1
V Y I = (5.3)
1 21 2
V Y I = (5.4)
.
.
.
1
V Y I
nn n
= (5.5)
From (5.3), (5.4) and (5.5)
.
.
j
v
2
i
j
i
.
.
Transformer
1
i
n
i
1
v
2
v
n
v
38
1
1
11
V
I
Y = ,
1
2
21
V
I
Y = ,
1
1
V
I
Y
i
i
= ,
1
1
V
I
Y
n
n
= (5.6)
The same procedure can be used to find all the columns of Y. In general the procedure
used to measure the admittance is depicted in Fig.5.2.
Fig.5.2 Measuring element of the
th
j column of Y (
j
i
ij
V
I
Y = ,
j
j
jj
V
I
Y = ).
The measurements were performed on three test transformer. A single phase transformer
at Chalmers University of Technology and two three phase transformers at ABB
Corporate Research.
In both the three phase transformers, the high voltage winding are connected in delta and
the low voltage windings are connected in wye.
The admittance matrix for the test transformers has been measured using the measuring
technique outlined in Fig.5.2. This matrix has a size of 2x2 and 6x6 for the single and
three phase test transformers respectively; hence (5.7) in its simplified form for the
single and three phase test transformers can be re written by (5.8) and (5.9). The terminal
identification in these equations is according to Fig.5.3.
a) Single phase transformer b) TX1 c) TX2
Fig.5.3 Test transformers.
r Transforme
.
.
.
.
j
v
i
i
j
i
.
.
1
n
i
j
1
2
3
4
5
6
V 410
KV 5 . 20
1
2
3
4
5
6
V 690 KV 20
1 2
V 220 V 127
39
(
¸
(
¸
(
¸
(
¸
=
(
¸
(
¸
2
1
22 21
12 11
2
1
V
V
Y Y
Y Y
I
I
(5.7)
(
(
(
(
(
(
(
(
¸
(
¸
(
(
(
(
(
(
(
(
¸
(
¸
=
(
(
(
(
(
(
(
(
¸
(
¸
6
3
3
3
2
1
66 35 64 63 62 61
56 35 54 53 12 51
46 35 44 43 42 41
36 35 34 33 32 31
26 25 24 23 22 21
16 15 14 13 12 11
6
5
4
3
2
1
5
V
V
V
V
V
V
Y Y Y Y Y Y
Y Y Y Y Y
Y Y Y Y Y Y
Y Y Y Y Y Y
Y Y Y Y Y Y
Y Y Y Y Y Y
I
I
I
I
I
I
(5.8)
Consider the situation that the transformer terminals are divided into two groups, denoted
by H and L, to represent the High voltage and Low voltage terminals respectively. This
partitioning of the terminal or winding simplifies the problem. (5.7) and (5.8) can then be
written in the form
(
¸
(
¸
(
¸
(
¸
=
(
¸
(
¸
H
L
HH HL
LH LL
H
L
V
V
Y Y
Y Y
I
I
(5.9)
(
¸
(
¸
(
¸
(
¸
=
(
¸
(
¸
L
H
LL LH
HL HH
L
H
V
V
Y Y
Y Y
I
I
(5.10)
Where
L
I ,
L
V and
H
I ,
H
V are the current and voltage at the low and high voltage
terminals and the Sub matrixes
HH
Y ,
HL
Y ,
LH
Y ,
LL
Y are of size 1x1 and 3x3 for the single
and three phase transformer respectively as it can be seen from (5.7) and (5.8) .
It is also interesting to mention the symmetry of the sub matrix. This is due to the
symmetry of the windings.
If the high voltage (H) terminals are opencircuited, we can easily get the voltage ratio
from high to low by putting 0 =
H
I in (5.9) and (5.10).
HL
HH
L
H
HL
Y Y
V
V
V
1 −
− = =
, for 0 =
H
I (5.11)
Similarly if the low voltage (L) terminals are opencircuited the voltage ratio from low to
high can be obtained by setting 0 =
L
I in (5.9) and (5.10).
LH
LL
H
L
LH
Y Y
V
V
V
1 −
− = =
, for 0 =
L
I (5.12)
40
It can be easily noted from (5.11) and (5.12)
HL
V ,
LH
V are matrices of size 1 by 1and 3 by
3 for a single phase and three phase transformer respectively.
The voltage ratios from Low to High and from High to Low can be measured using the
set up given in Fig.5.4 and Fig.5.5 which demonstrates this clearly.
Fig.5.4 Measuringelement of the first column of
LH
V (
1 ,
4 ,
41 ,
H
L
LH
V
V
V =
,
1 ,
5 ,
51 ,
H
L
LH
V
V
V =
,
1 ,
6 ,
61 ,
H
L
LH
V
V
V =
).
Fig.5.5 Measuring element of the first column of
HL
V (
4 ,
1 ,
14
L
H
HL
V
V
V = ,
4 ,
2 ,
24
L
H
HL
V
V
V = ,
4 ,
3 ,
34
L
H
HL
V
V
V = ).
5.2.2 Rational Approximation of Frequency Responses by
Vector Fitting
This section briefly reviews the method of rational approximation by vector fitting. The
method is used for the fitting of measured or calculated frequency domain responses into
a rational function that approximates the response very well. It can be employed both for
single as well as for several frequency responses of a system. The method was developed
by Björn Gustavsen et al based on reformulation of the Sanathanan–Koerner iteration and
has been found to be very suitable for fitting network equivalents and transformer
responses. Some examples where this method can be used are:
• Network equivalent for portion of a large network
1
2
3
4
5
6
H
L
1 , H
V
4 , L
V
5 , L
V
6 , L
V
1
2
3
4
5
6
H
L
3 , H
V
4 , L
V
1 , H
V
2 , H
V
41
• Highfrequency transformer model based on measured frequency responses
• Transfer of voltages from transformer terminals to internal points in winding
• Transmission line modeling
The method is later applied to fit the admittance matrix of power transformers in coming
sections.
The Vector Fitting Algorithm
Generally speaking, an approximation of a given order can be obtained by fitting a given
set of frequency dependent data to a ratio of polynomials as shown in below [27].
N
N
N
N
s b s b s b s b
s a s a s a s a
+ + + + +
+ + + + +
≈
... b
... a
f(s)
3
3
2
2 1 0
3
3
2
2 1 0
(5.13)
To write this non linear equation as a linear problem of the type Ax=b with the
coefficients as unknowns, both sides should be multiplied with the denominator. This,
however, produces a badly scaled and illconditioned problem as the columns of A are
multiplied with different powers of s. This imposes a limitation on the applicability of the
approximation to very lower order only. A more efficient method is therefore required.
Of the many fitting algorithms developed to date, Vector fitting has been found to be very
efficient to model transformer responses and network equivalents.
The idea of vector fitting is to find a rational function approximation which is sum of
partial fractions from the given frequency response data. This is done by replacing a set
of starting poles with an improved set of poles via an iterative pole relocation method
based on least squares approximation of linear problems [27]. The starting poles are so
chosen to cover the frequency range of interest and should be linearly or logarithmically
distributed over the whole range. It is recommended to choose complex conjugate pole
pairs for functions with distinct resonance peaks like transformer responses as starting
poles. For smooth functions real poles are sufficient.
To demonstrate the method of vector fitting, consider the rational function approximation
e s d
a s
c
s f
N
n n
n
. ) (
1
+ +
−
=
¿
=
(5.14)
where f(s) is a frequency dependent transfer function of a single response or a matrix of
several responses, c
n
and
a
n
are the residues and poles respectively which can be real or
conjugate pairs and d and e are real. N is the order of approximation. The intention is to
approximate the response f(s) by the terms on the right hand side of the equation. This
requires computation of the unknowns c
n
, a
n
, d and e. An approximation of f(s) in a
given frequency range is therefore achieved this way. Since one of the unknowns a
n
in
(5.14) is in the denominator, the problem is a non linear one. Therefore, the problem
should be linearized before it is solved. The algorithm of vector fitting tackles the
42
problem in two steps. In the first step, the poles are identified and in the second step
residues are identified.
1. Step 1
Here, a set of starting poles
n
a
are chosen for f(s) and (5.14) is multiplied by an
unknown function (s) which is also approximated with a rational function with the same
starting poles.
(s) ≈ 1
~
1
+
−
¿
=
N
n n
n
a s
c
(5.15)
An augmented problem can thus be written
(
¸
(
¸
(s)
(s) f(s)
σ
σ
≈
(
(
(
(
¸
(
¸
+
−
+ +
¿
¿
=
=
1
~
.
a  s
c
1
1 n
n
N
n n
n
N
n
a s
c
e s d
(5.16)
To have a linear problem in the unknowns c
n,
n
c
~
,
d and e, the second row of (5.16) is
multiplied by f(s) giving
e s d
a s
c
N
n n
n
.
1
+ +
−
¿
=
≈


.

\

+
−
¿
=
1
~
1
N
n n
n
a s
c
f(s) (5.17)
( f)
fit
(s)
=
fit
(s) f(s) (5.18)
is a linear problem of the type Ax=b in the frequency range of interest with the above
stated unknowns in the solution vector x. For a given frequency
k
S
k k
b x A = (5.19)
where A
k
=[
2 1
1
...
1
a s a s
k k
− −
1 s
k
1
) (
a s
s f
k
k
−
−
…
1
) (
a s
s f
k
k
−
−
] (5.20)
x=[ c
1….
c
N
d e
1
~
c …
n
c
~
]
T
(5.21)
b
k
=f(s
k
) (5.22)
From (5.18), we have
43
f(s) =
) (
) ( ) (
s
s f
fit
fit
σ
σ
(5.23)
Since both numerator and denominator of (5.23) have the same poles( s 
n
a ), the ratio in
(5.23) can be rewritten as single fraction instead of sum of partial fractions
f(s) =e
∏
∏
=
+
=
−
−
N
n
n
N
n
n
z s
z s
1
1
1
~
(5.24)
It is clearly seen from (5.24) that the zeros of
fit
(s) are equal to the poles of f(s). Thus by
calculating the zeroes from the linear equation of (5.15), a good set of poles for the fitted
f(s) is obtained.
2. Step 2
Now the identification of the residues of the function f(s) remains in order to have a
complete fit of its rational approximation. If the zeroes of
fit
(s) calculated in step 1 are
used in the original problem in (5.14) as new poles, we thus have a linear problem where
the residues can be easily solved giving the unknowns c
n
, d and e.
For more complete and deeper treatment of how the vector fitting algorithm is
formulated, the reader is referred to [27].
Passivity enforcement
The objective of the rational approximation of the frequency dependent response f(s) in
(5.14) using the vector fitting algorithm is to have an equivalent model in the form of a
realized network of passive elements when f(s) is an admittance or an impedance
function, Y(s) or Z(s). The equivalent model thus obtained is linear, hence the term
Passive Linear Time invariant component. If this component is to be included in a
transient time domain simulation, it should behave a passive nature in that it should only
absorb active power for any set of applied voltages and not produce.
Time domain simulations involving a fitted admittance or impedance may sometimes
lead to an unstable time domain simulation due to non passivity. There are also cases
where the model could be stable and still non passive in which case unstable simulations
arise when the model is used in an interconnection with other components. It is, therefore,
important to check that the equivalent network contains no passivity violations over the
whole frequency range and outside the frequency range as well. If violations exist, they
should be removed. This procedure of removing passivity violations is termed Passivity
Enforcement.
44
A requirement for the fitted admittance Y(s) to be passive is for the real part of the
admittance to be Positive Definite, which means all eigenvalues of G should be real
positive. A general procedure to enforce passivity to the network is to calculate a
correction for the approximated rational function so that it becomes Positive Definite
[28]. This corrective perturbation should be as minimal as possible since it causes a
fitting error of the actual data. One way to achieve minimum correction is to have very
accurate measurement of the response. A formal formulation of the PD criterion is given
below [28].
Passivity Criterion
Consider a component defined by an admittance matrix Y(s) in the frequency domain
I=Y v (5.25)
where I and v are current and voltage complex vectors. The absorbed power is therefore
given as
P=Re {v*Y v} =Re {v*(G+jB)v}= Re {v*Gv} (5.26)
where * means transpose of conjugate of v.
It follows from (5.26) that the real power P will always be positive if all eigenvalues of G
are positive, entailing the PD criterion.
For further and detailed reading on passivity enforcement the reader is referred to [29].
5.2.3 Time Domain Implementation
The basic idea behind the new transformer model is to produce an equivalent network
whose admittance matrix matches the admittance matrix of the original transformer over
the frequency range of interest. The next question is how we get this equivalent network.
This equivalent network can be obtained by realization of the fitted admittance matrix
Y
fit
(s).
The procedure to obtain the network equivalent is given in the APPENDIX and
summarized here. Consider the admittance matrix Y of a transformer obtained using the
measurement procedure in section 5.2.1 and fitted with a rational function using the VF
method given in section 5.2.2 creating a matrix Yfit(s) in which the elements are
ij ij
N
m
m
mij
ij fit e s d
a s
r
s Y . ) (
1
+ +
−
=
¿
=
(5.27)
The network has branches between all nodes and ground and between all nodes. Branches
between node and ground are given by:
45
¿
=
=
n
j
ij fit ii s Y y
1
) ( (5.28)
while the branches between node i and node j are
ij fit ij s Y y ) ( − = (5.29)
where n is the size of the matrix which is 2 for single and 6 for the three phase
transformers.
Consider now a single branch y(s) of the network (node to ground or between two nodes)
fitted with real and complex conjugate poles
e s d
ja a s
jr r
ja a s
jr r
ja a s
jr r
ja a s
jr r
a s
r
a s
r
s y . ...
) " (
" '
) " ' (
" '
) " ' (
" '
) " ' (
" '
... ) (
2 2
2 2
2 2
2 2
1 1
1 1
1 1
1 1
2
2
1
1
+ + +
+ −
+
+
− −
−
+
+ −
+
+
− −
−
+ +
−
+
−
=
(5.30)
The synthesis of (5.30) is an RLC network of parallel branches as shown in Fig.5.6.
where
d
R
1
0 = , e C = 0 (5.31)
The real pole is represented by the RL circuit as
r
Lr
1
= ,
r
a
Rr
−
= (5.32)
while the complex conjugate pair is given by RLCG network, where
' 2
1
r
Lc = , ( ) ( ) ' " " ' ' 2 a a r a r L L R c c c − + = (5.33)
)) ' ' ' ' ' ' ( 2 ' ' ' (
1
2 2
a r a r R a a L
C
c c
c
+ + +
= , ( ) " " ' ' 2 a r a r C L G c c c + − = (5.34)
46
Fig.5.6 Basic structure of the synthesized electrical network for a single element branch.
...
0
R
0
C
r
R
r
L
c
R
c
L
c
C
c
G
...
Constant term s.e term
Real Pole
Complex conjugate pole
47
Chapter 6
TRANSFORMER MODELINGExperiments and
Results
6.1 Introduction
In this chapter the measurement setup designed and instrumentations used to measure the
admittance matrix and voltage ratio over a wider frequency range is describes in detail
.The measurement and fitting result along with their analysis is discussed and finally the
model is validated by comparing the measured results with the corresponding calculated
and PSCAD/EMTDC simulated results.
6.2 Measurement Setup
The admittance matrix and voltage ratios for the transformers were measured in
frequency domain
using a network analyzer.The measurement set up as configured to measure one element
of the admittance matrix and voltage ratio is outlined in Fig.5.1 and Fig.5.2 respectively.
The set up consists of the test transformer (a single phase 127/220V and two three phase
transformer), the network analyzer to measure the admittance and the voltage ratio over a
wide frequency range, current probe for the current measurement and shielded cables
which connect the output port of the network analyzer to the transformer terminal for
excitation and from the transformer terminal to the input ports of the network analyzer for
the purpose of voltage or/and current measurements.
Fig.6.1 Experimental setup to measure element
31
Y of the admittance matrix.
48
Fig.6.2 Experimental setup to measure
1 ,
6 ,
61 ,
H
L
LH
V
V
V = element of the admittance matrix.
Measurement Objects
As it was mentioned earlier the measurements and hence modeling has been done on
three transformers but the measurements done on the single phase transformer was an
important step to the modeling of two three phase transformers which represent the wind
mill (TX1) and the load side transformers of three phase cable systems that has been built
at ABB corporate research to emulate a section of an off shore collection wind park
model in order to study fast transients that can occur in an off shore wind park.
Windmill Transformer (TX1)
Transformer TX1 (Fig.6.3) in the experimental set up is the representation for one of the
windmill transformers in the off shore wind park. It is a 1.25MVA, 410V/20.5kV step up
transformer with low voltage side connected in Y and high voltage side in A. The
transformer data as it is shown in the name plate data is given in Table 6.1.
Load side Transformer (TX2)
The load side transformer TX2 (Fig.6.4) is a 1.0MVA, 690 V/20kV distribution
transformer with A connected high voltage winding and Y connected low voltage
winding. The specification of this transformer is given in Table 6.2.
49
Fig.6.3 Windmill Transformer (TX1). Fig.6.4 Load side Transformer (TX2).
Table 6.1 Transformer TX1 data. Table 6.2 Transformer TX2 data.
Network Analyzer
Two different network analyzers have been used for measurement. This is because the
measurements for the single phase transformer were done at Chalmers while for the two
three phase transformers they were done at ABB Corporate Research. The MS4630B
network analyzer was used to measure the admittance and voltage ratio of the single
phase transformer at Chalmers. This network analyzer has a bandwidth of 10Hz to
300MHz, two output (A and B) and three input (R, TA and TB) channels. The Bode 100
network analyzer which was used for the two threephase transformers measurements has
a bandwidth of 10 to 40MHz, one output and two input ports. For both admittance and
voltage ratio measurements one output channel which is for the excitation voltage and
two input channel; one for voltage and one for current measurement for the case of
admittance or both for voltage measurement were required as the admittance and voltage
NO. 5210667 1997
CTMU 24 HM 1000 IEC 76
1000 kVA Dyn11 50 Hz
20000 ± 2´2.5 % V 28.87 A
690 V 836.7 A
U 125AC50/AC8 kV ONAN
Zk 5.1 % 2670 kg
Pk 10260 W 505 kg
Po 1223 W
NO. 5214242 1999
CTMU 24 HA 1250 (ABB) IEC 76
1250 kVA Dyn11 50 Hz
20500 ± 2´2.5 % V 35.2 A
410 V 1760 A
U 125AC50/AC8 kV ONAN
Zk 5.4 % 2885 kg
Pk 11971 W 605 kg
Po 1455 W
50
ratios were measured as the ratio of current to voltage or two voltages, respectively. The
specifications of the network analyzers are given in Table 6.3.
Fig.6.5 MS4630B Network Analyzer. Fig.6.6 Bode 100 Network Analyzer.
Table 6.3 Technical data of the bode 100 network analyzer.
CHARACTERISTIC MS4630B BODE 100
Frequency range 10Hz to 300MHz 10 Hz to 40 MHz
Output impedance 50 O 50 O
Out put Connector BNC BNC
Wave form Sinusoidal signal Sinusoidal signal
Output voltage Output A 0 to +21 dBm
Output B –6 to +15 dBm
0.01 to 1Vrms into 50 O
load
–27 dBm to 13 dBm
Input Channel R, TA, TB CH 1 , CH 2
Input impedance 50 O or 1 MO 50 O or 1 MO
Connectors BNC BNC
Receiver bandwidth
3, 10, 30, 100, 500 Hz, 1,
2, 3, 4, 5, 10, 20 kHz
10 Hz, 30 Hz, 100 Hz, 300
Hz, 1 kHz,3 kHz
SweepingNumberof
measurement points:
11, 21, 51, 101, 251, 501,
1001
10 to 16501 (freely
selectable)
Current Probe
As the admittance elements were measured as the ratio of current to voltage, current
probes of model pearson (Fig.6.7) with sensitivity of 1V/1A for the singlephase
transformer and 0.1V/1A for two threephase transformer measurements has been used.
Fig.6.7 current probes of model Pearson.
51
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
A
d
m
i
t
t
a
n
c
e
[
S
]
Diag. elements of YLL
OffDiag. elements of YLL
Diag elements of YHH
Diag. elements of YHL and YLH
OffDiag. elements of YHL and YLH
OffDiag. elements of YHH
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
M
a
g
n
i
t
u
d
e
[
S
]
Diagonal elements of YLL
OffDiag elements of YLL
Diagonal elements of YHL,YLH
OffDiag. elements of YHH
Diagonal elements of YHH
OffDiag.elements of YHL,YLH
6.3 Measurement Results
Fig.6.8 and 6.9 show the magnitudes of the 36 elements of the admittance matrix Y for
TX2 and TX1 obtained by direct measurement in four sets of frequency sweeps, two with
801 and 401 sampling points for TX2 and two with 1601 and 801 for TX1. These raw
data contain inaccuracies due to noise and other factors as is reflected clearly in the
figures. The reproducibility or repeatability of the measurements is, however,
indisputably validated in the 4 sets of measurements. The data in Fig. 6.8 a) and 6.8 b)
obtained for TX2 are almost the same for the two sets of measurements, with 401 and 801
sampling points respectively. The same is also observed in Fig. 6.9 a) and 6.9b) for TX1,
with 801 and 1601 sampling points respectively. The credibility of the data is further
enhanced by the large difference in the element sizes which is expected due to the high
voltage ratio of the transformers ( ≈50).
a) 401 frequency points b) 801 frequency points
Fig.6.8 Magnitude plot of admittance matrix (raw data) for TX2.
Reproducibility of the data, however, does not always guarantee an accurate data set
especially when the cause for the inaccuracy persists. This problem has been tackled by
exploiting the symmetry of the transformers and thus of the admittance matrix. Due to
this symmetry, several elements of the admittance matrix are identical or closely so.
Therefore, bad measurements or noisy measured admittance elements have been replaced
with better measured data of another symmetric element. A summary of elements
expected to be identical or closely so is given below where Y has been partitioned into
four sub matrices as shown in
Y=
(
¸
(
¸
LL LH
HL HH
Y Y
Y Y
(6.1)
52
10
1
10
2
10
3
10
4
10
5
10
6
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
M
a
g
n
i
t
u
d
e
[
S
]
Diagonal elements of YLL
Diagonal elements of YHH
OffDiag. elements of YHH
OffDiag. elements of YHL,YLH
Diagonal elements of YHL,YLH
OffDiag. elements of YLL
10
1
10
2
10
3
10
4
10
5
10
6
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
A
d
m
i
t
t
a
n
c
e
[
S
]
Diag. elements of YLL
OffDiag. elements of YLL
OffDiag. elements of YHL and YLH
Diag. elements of YHL and YLH
Diag elements of YHH
OffDiag. elements of YHH
10
1
10
2
10
3
10
4
10
5
10
6
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
M
a
g
n
i
t
u
d
e
[
S
]
10
1
10
2
10
3
10
4
10
5
10
6
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
A
d
m
i
t
t
a
n
c
e
[
S
]
i) Diagonal elements of Y
HH
ii) Diagonal elements of Y
LL
iii) Diagonal elements of Y
HL
and Y
LH
iv) Offdiagonal elements of Y
HH
v) Offdiagonal elements of Y
LL
vi) Offdiagonal elements of Y
HL
and
Y
LH
a) 801 frequency points b) 1601 frequency points
Fig.6.9 Measured admittance elements (Raw data) for TX1.
Neater and more accurate data is accordingly found with this modification. The new data
set is given in Fig. 6.10 and Fig. 6.11.
a) 401 frequency points b) 801 frequency points
Fig.6.10 Measured admittance elements modified using symmetry of the sub matrices for
TX2.
53
10
1
10
2
10
3
10
4
10
5
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
M
a
g
n
i
t
u
d
e
[
S
]
10
1
10
2
10
3
10
4
10
5
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
A
d
m
i
t
t
a
n
c
e
a) 801 frequency points b) 1601 frequency points
Fig.6.11 Measured admittance elements modified using symmetry of the sub matrices for
TX1.
6.4 Rational Approximation of Admittance Matrices of
The Transformers
This section discusses the results of rational approximation of the measured admittance
matrices of the power transformers by curve fitting of the frequency response measured at
the terminals. The method used is based on a series of papers on curve fitting and open
Matlab source codes developed by Björn Gustavsen from SINTEF. Matrix curve fitting is
used as the transformer is a multiterminal device [25]. All the files and Matlab codes on
curve fitting can be downloaded from [26].
The frequencydependent responses are obtained via measurements as discrete functions
of frequency. The method applied assumes the transformer as a linear timeinvariant and
frequencydependent black box with a certain number of terminals. The frequency
response measured is a set of currentvoltage ratios over a wide band of frequencies. This
gives a threedimensional admittance matrices of size nxnxNs where n is the number of
terminals(nodes) and Ns is the number of measurement points (frequency points). Matrix
curve fit is then applied to approximate the elements of the admittance matrix by a
rational approximation.
The rational function f(s) given in (5.14) in this case becomes an nxn admittance matrix
as all transformers are multi terminal devices with n terminals. A Matrix curve fit stacks
all elements of the matrices into a single column vector h(s) before the elements are
fitted.
54
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency [Hz]
M
a
g
n
i
t
u
d
e
[
p
.
u
.
]
Original
Approximation
10
1
10
2
10
3
10
4
10
5
10
6
700
600
500
400
300
200
100
0
100
200
300
Frequency [Hz]
P
h
a
s
e
a
n
g
l
e
[
d
e
g
]
Original
Approximation
10
3
10
4
10
5
10
6
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Frequency [Hz]
E
ig
e
n
v
a
lu
e
s
o
f
G
Before pass. enf.
After pass. enf.
The curve fitting algorithm was used to approximate all elements of the admittance
matrices with a common set of poles. Passivity was also enforced for all elements to
ensure stable time simulation.
The original admittance elements are drawn together with their rational approximations
for comparison to see how good the fitting is for different orders of approximations.
Eigen values of Re{Y} have also been drawn before and after passivity is enforced on all
elements for different number and range of frequencies to see the effect of passivity on
the error of approximation.
Fig. 6.12 shows the rational approximation of both magnitude and phase of the 36
elements of TX2 with an order of approximation of N=20, 3 iterations and passivity over
801 frequency points in the range from 1kHz and 1 MHz. Although the fit in the low
frequency segment up to around 10 kHz seems ok, we have a very poor approximation at
high frequencies. The total rmserror of approximation in this case is equal to 0.013563.
The eigenvalues of Re{Y} before and after passivity enforcement for this case is given in
Fig. 6.13.
a) Magnitude b) Phase
Fig.6.12 Rational approximation of Y for TX2 with N=20.
Fig.6.13 Eigenvalues of Re {Y} for TX2 before and after passivity, N=20 for 1kHz to
1MHz.
55
10
1
10
2
10
3
10
4
10
5
10
6
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency [Hz]
M
a
g
n
i
t
u
d
e
[
p
.
u
.
]
Original
Approximation
10
1
10
2
10
3
10
4
10
5
10
6
700
600
500
400
300
200
100
0
100
200
300
Frequency [Hz]
P
h
a
s
e
a
n
g
l
e
[
d
e
g
]
Original
Approximation
10
3
10
4
10
5
10
6
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Frequency [Hz]
E
i
g
e
n
v
a
l
u
e
s
o
f
G
Before pass. enf.
After pass. enf.
In principle, a perfect fit can be obtained by increasing the order indefinitely high. In
practical applications, however, one often wants to find a low order approximation as
higher order gives larger element size of the resulting network, higher simulation time
and requires very big computational resource. In our case a better fit has been achieved
by increasing the order of approximation to N=100 with the same passivity enforcement
as the previous one. The resulting approximation is shown in Fig.6.14 where it is seen
that the approximation has been improved throughout the whole frequency range. The
rmserror in this case is 0.0113. The Eigen values of Re{Y} before and after passivity
enforcement is given in Fig. 6.15.
a) Magnitude b) Phase
Fig.6.14 Rational approximation of Y for TX2 with N=100.
Fig.6.15 Eigenvalues of Re {Y} for TX2 before and after passivity, N=100 for 1kHz to
1MHz.
56
10
2
10
0
10
2
10
4
10
6
10
8
1
0
1
2
3
4
5
6
7
Frequency [Hz]
E
i
g
e
n
v
a
l
u
e
s
o
f
G
Before pass. enf.
After pass. enf.
To show the effect of passivity on the fitting error, Y(s) was approximated with order of
approximation N=100 as in Fig. 6.14 but with an increased range of frequencies
enforced. Fig. 6.16 gives the Eigen values of Y(s) before and after passivity is enforced
over the frequency range 0.1Hz to 10MHz. The error of approximation is increased from
0.0113 to 0.015133. The increase in fitting error is because of perturbation of more
frequency points while enforcing passivity in the later case. To decrease the fitting error
due to passivity, therefore, very accurate measurements should be made so that there are
no passivity violations inside the frequency range of interest or outside. A compromise is
therefore made between stability and fitting error.
Fig.6.16 Eigenvalues of Re {Y} for TX2 before and after passivity, N=100 for 0.1Hz to
10MHz.
Fig.6.17 and Fig.6.18 give the rational approximation and eigenvalues of Re{Y(s)}
respectively of TX1 for N=60 and passivity enforcement over 0.1kHz to 10MHz on 1601
points. The rmserror is 0.023634. A reduced order of approximation was used due to
limitation of computational memory since the number of measurement points to be
approximated in this case was 1601 which is double that for TX2.
57
10
1
10
2
10
3
10
4
10
5
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
Frequency [Hz]
M
a
g
n
i
t
u
d
e
[
p
.
u
.
]
Original
Approximation
10
1
10
2
10
3
10
4
10
5
10
6
500
400
300
200
100
0
100
200
300
Frequency [Hz]
P
h
a
s
e
a
n
g
l
e
[
d
e
g
]
Original
Approximation
10
2
10
0
10
2
10
4
10
6
10
8
1
0
1
2
3
4
5
6
7
Frequency [Hz]
E
i
g
e
n
v
a
l
u
e
s
o
f
G
Before pass. enf.
After pass. enf.
a) Magnitude b) Phase
Fig.6.17 Rational approximation of Y for TX1 with N=60
Fig.6.18 Eigenvalues of Re {Y} for TX1 before and after passivity, N=60 for 0.1Hz to
10MHz.
Finally, approximations and eigenvalues with orders 20 and 100 are plotted for the small
transformer at Chalmers as shown in Fig. 6.19, Fig. 6.20, Fig.6.21 and Fig.6.22
respectively. Deviations between original data and approximations are also shown in the
same plots. Far better fits are achieved for this transformer since the number of elements
to be fitted with a common set of poles in this case is only 4.
58
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
5
10
4
10
3
10
2
10
1
10
0
Frequency [Hz]
M
a
g
n
i
t
u
d
e
[
p
.
u
.
]
Original
Approximation
Deviation
Y12 and Y21
Y22
Y11
10
1
10
2
10
3
10
4
10
5
10
6
10
7
1800
1600
1400
1200
1000
800
600
400
200
0
200
Frequency [Hz]
P
h
a
s
e
a
n
g
l
e
[
d
e
g
]
Original
Approximation
Y11 and Y22
Y12 and Y21
10
2
10
0
10
2
10
4
10
6
10
8
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Frequency [Hz]
E
i
g
e
n
v
a
l
u
e
s
o
f
G
Before pass. enf.
After pass. enf.
a) Magnitude b) Phase
Fig.6.19 Rational approximation of Y for small transformer with N=20.
Fig.6.20 Eigenvalues of Re {Y} for small transformer before and after passivity, N=20
for 0.1Hz to 10MHz.
59
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
6
10
5
10
4
10
3
10
2
10
1
10
0
Frequency [Hz]
M
a
g
n
i
t
u
d
e
[
p
.
u
.
]
Original
Approximation
Deviation
Y11
Y22
Y12 and Y21
10
1
10
2
10
3
10
4
10
5
10
6
10
7
2000
1500
1000
500
0
500
Frequency [Hz]
P
h
a
s
e
a
n
g
l
e
[
d
e
g
]
Original
Approximation
Y11 and Y22
Y12 and Y21
10
2
10
0
10
2
10
4
10
6
10
8
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Frequency [Hz]
E
i
g
e
n
v
a
l
u
e
s
o
f
G
Before pass. enf.
After pass. enf.
a) Magnitude b) Phase
Fig.6.21 Rational approximation of Y for small transformer with N=100.
Fig.6.22 Eigenvalues of Re {Y} for small transformer before and after passivity, N=100
for 0.1Hz to 10MHz.
60
6.5 Model Validation
In this section the validity of the transformer models has been checked by comparing the
voltage ratio measured directly on the transformer terminals with the calculated and
simulated by the model in PSCAD/EMTDC. The voltage ratio calculated was obtained
from the measured admittance using (5.11) and (5.12). The accuracy of this model greatly
depends on the accuracy of the measured admittance and hence it is very important to the
check if these measurements are accurate enough to give the accuracy requirements to the
model to be developed using this measurements. One way to validate the transformer
model is therefore to validate the measured admittance matrix without further doing
timedomain implementation in PSCAD/EMTDC as this model was formulated in terms
of its admittance matrix.
Transformer TX2
The PSCAD/EMTDC simulated result for TX2 was obtained by the model developed for
an admittance matrix Y measured with 801 number of frequency samples and fitted with
an approximating rational function of order 100. Fig.6.23 to Fig.6.26 compares the
measured elements of voltage ratio from low to high (
LH
V ) with the corresponding
elements calculated and simulated. The calculated ratio is seen to closely match the
measured ratio except at lower frequencies. The reason for this will be explained later in
section (6.6). Similarly very good agreement has been obtained between the calculated
and simulated ratio as it can be seen in the same figures. It is important to note here that
as the model developed is not directly from the measured admittance but from rational
approximation of the measured admittance, the small deviation between the calculated
and simulated are the result of fitting errors otherwise the calculated and simulated curves
fall on the same curve provided that we can get an ideal perfect fitting. In practical
applications, one often wants to find low order approximation as higher order gives larger
element size of the resulting network, higher computer memory and higher simulation
time. With the computer we had for example it was impossible to exceed the order from
130 or 80 for number of samples 801 and 1601 respectively.
A similar observation is made for the voltage ratio elements from high to low (
HL
V ) when
comparing measured and calculated quantities, see Fig.6.27 to Fig.6.29.
61
10
1
10
2
10
3
10
4
10
5
10
6
10
2
10
1
10
0
10
1
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
simulated
10
1
10
2
10
3
10
4
10
5
10
6
10
2
10
1
10
0
10
1
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
simulated
10
1
10
2
10
3
10
4
10
5
10
6
10
2
10
1
10
0
10
1
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
simulated
10
1
10
2
10
3
10
4
10
5
10
6
10
2
10
1
10
0
10
1
Frequency[Hz]
V
o
l
t
a
g
e
r
a
t
i
o
calculated
measured
simulated
Fig.6.23 Voltage ratio low to high
1 ,
4 ,
41 ,
H
L
LH
V
V
V = . Fig.6.24 Voltage ratio low to high
2 ,
5 ,
52 ,
H
L
LH
V
V
V = .
Fig.6.25Voltage ratio low to high
3 ,
6 ,
63 ,
H
L
LH
V
V
V = . Fig.6.26 Voltage ratio low to high
1 ,
6 ,
61 ,
H
L
LH
V
V
V = .
62
10
1
10
2
10
3
10
4
10
5
10
6
10
2
10
1
10
0
10
1
10
2
10
3
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
10
1
10
2
10
3
10
4
10
5
10
6
10
2
10
1
10
0
10
1
10
2
10
3
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
10
1
10
2
10
3
10
4
10
5
10
6
10
4
10
2
10
0
10
2
10
4
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
Fig.6.27 Voltage ratio high to low
4 ,
1 ,
14 ,
L
H
HL
V
V
V = . Fig.6.28 Voltage ratio high to low
5 ,
2 ,
25 ,
L
H
HL
V
V
V = .
Fig.6.29 Voltage ratio high to low
6 ,
3 ,
36 ,
L
H
HL
V
V
V = .
Transformer TX1
Similar plots have been made for transformer TX1 except the admittance matrix and
voltage ratios were measured for 1601 number of samples and the simulated results are
for the model obtained with an order of approximation of 60 for the measured admittance.
Reasonably good agreement between calculated and simulated results are obtained for the
voltage ratios
LH
V as it is shown in the plots of Fig.6.30 to Fig.6.32.Compared to the
results obtained for TX2,it is seen that a relatively large deviation between the measured
and calculated (simulated) is observed for TX1. This is due to the higher error on the
measured admittance for TX1 than TX2.This was observed during the measurements of
the admittance for TX1 where more elements have to be replaced using symmetry of the
63
10
1
10
2
10
3
10
4
10
5
10
6
10
3
10
2
10
1
10
0
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
simulated
10
1
10
2
10
3
10
4
10
5
10
6
10
3
10
2
10
1
10
0
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
simulated
10
1
10
2
10
3
10
4
10
5
10
6
10
3
10
2
10
1
10
0
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
time simulation
matrix. This can also be observed by comparing the admittance matrix plots for TX2 and
TX1 shown in Fig.6.8 and Fig.6.9.
Fig.6.30 Voltage ratio low to high
1 ,
4 ,
41 ,
H
L
LH
V
V
V = . Fig.6.31 Voltage ratio low to high
2 ,
5 ,
521 ,
H
L
LH
V
V
V = .
Fig.6.32 Voltage ratio low to high
3 ,
6 ,
63 ,
H
L
LH
V
V
V = .
64
10
1
10
2
10
3
10
4
10
5
10
6
10
2
10
1
10
0
10
1
10
2
10
3
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
10
1
10
2
10
3
10
4
10
5
10
6
10
2
10
0
10
2
10
4
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
10
1
10
2
10
3
10
4
10
5
10
6
10
1
10
0
10
1
10
2
10
3
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
Fig.6.33 Voltage ratio high to low
4 ,
1 ,
14 ,
L
H
HL
V
V
V = . Fig.6.34 Voltage ratio high to low
5 ,
2 ,
25 ,
L
H
HL
V
V
V = .
Fig.6.35 Voltage ratio high to low
6 ,
3 ,
36 ,
L
H
HL
V
V
V = .
Single phase Transformer
Fig.6.36 and Fig.6.37 compare the measured and calculated (using (5.11) and (5.12))
values of voltage ratios. The calculated ratios are seen to closely match the measured
ratios with some discrepancy. According to [23] the condition with one or more terminals
of the transformer open is the worst case scenario because a 1% measurement error in the
admittance matrix could cause as high as 100% error in the calculated voltage ratio,
hence the errors obtained are reasonable.
65
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
3
10
2
10
1
10
0
10
1
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
10
1
10
2
10
3
10
4
10
5
10
6
10
7
10
3
10
2
10
1
10
0
Frequency[Hz]
V
o
l
t
a
g
e
R
a
t
i
o
calculated
measured
Fig.6.36 Voltage ratio high to low
2 ,
1 ,
12 ,
H
L
LH
V
V
V = . Fig.6.37 Voltage ratio high to low
1 ,
2 ,
21 ,
L
H
HL
V
V
V = .
6.6 Accuracy of the Model
The models developed for TX2 and TX1 are not accurate at lower frequencies. This can
be seen from the comparison between measured and calculated results of the preceding
sections, see Fig.5.27 to Fig.5.39. This is especially observed for TX2 in plots of
Fig.5.27 to Fig.5.33.This error is due to the ungrounded windings of the delta connected
high voltage winding for both transformers.
The inaccurate result at low frequencies for transformers with ungrounded windings is
due to the inaccuracy in the zero sequence component of the admittance matrix at low
frequency corresponding to the ungrounded winding as the capacitive coupling to ground
of an ungrounded winding approaches zero with decreasing frequency, thereby causing a
nearsingular condition for the blocks in the admittance matrix Y associated to
ungrounded windings(
HH
Y ,
HL
Y ,
LH
Y ) and hence the zerosequence component of
diagonal and off diagonal blocks associated with ungrounded windings(
HH
Y ,
HL
Y ,
LH
Y )
becomes very small because the current is too small to be accurately determined as the
only connection to ground is via capacitors.
The sensitivity of the model to errors in measurements and the rational approximation
was also found to be very high when one or more terminals were opencircuited on one or
both sides of the transformer. Thus, a transformer model may, in general, produce
accurate results in one situation and inaccurate results in a different situation.
66
6.7 Conclusion
This chapter demonstrated how a linear wide band frequency dependent model of a
power transformer has been developed. A dedicated measurement setup to measure the
admittance matrix over a wider frequency range has been designed. Vector fitting was
adapted to subject the measured admittance matrix to rational function approximation.
Network realization of the approximating rational function to its equivalent electrical
network generated given in a Matlab generated file was used later to implement the
model in a PSCAD script for time domain simulation.
From the admittance matrix measurements taken at two different numbers of samples, the
measurement setup was shown to give reproducible results. Comparison of results
between measured, calculated and simulated results show good agreement. The model
was however found to be sensitive to measurement errors especially for a transformer
with ungrounded windings. The errors were magnified at lower frequencies and during
open circuit conditions.
67
Chapter 7
CONCLUSIONS AND FUTURE WORK
7.1 Conclusions
In power systems that comprise an interconnection of cables, VCBs and transformers as
in a wind farm, very severe high frequency overvoltages may occur due to the switching
action of the VCB. Characterization of these overvoltages and the ultimate protection
from them requires accurate modeling of the above components.
The first part of this report dealt with cable modeling and experimental validations. It has
been studied that the following conclusions can be drawn:
o With certain adjustments that take into account the semiconducting layers and
core and sheath stranding, the Frequency Dependent (Phase) Model provided by
PSCAD/EMTDC gives a fairly very good model of a real cable. This model
replicates high frequency behaviors like skin effect, reflections, EM propagation.
o The results of the experiment show that good agreements have been achieved
between measured and simulated values in the frequency of oscillation and
propagation delay. Small disagreements in these parameters are, however, noticed
and the authors believe they can be due to unaccounted capacitance of the
measuring cables and additional stray capacitances and inductances in the real
system. Although it has been found that both simulations and measurements show
an increase in damping of the EM transients with increasing frequency of
oscillation as it should be, quicker damping is observed in reality than in the
simulations.
VCB modeling and characterization of phenomena as current chopping, dielectric with
stand and high frequency quenching capability makes the second part of this thesis. The
following conclusions are inferred:
o The statistical nature of the arcing time, chopping current, dielectric strength, and
quenching capability of a VCB requires that a stochastic model be developed for
it.
o Although the VCB has good current interruption characteristics, its ability to
interrupt high frequency current zeros causes escalation of steep front
68
overvoltages on the load side due to reignitions or prestrikes specially when
inductive currents are interrupted. This has been demonstrated by using a
stochastic VCB model in a generic test circuit.
o The generic test circuit was able to replicate actual VCB characteristics and
expected results were obtained by looking into the effects of arcing time, rate of
rise of dielectric strength, and high frequency quenching capability.
Finally, a wide band frequency dependent model of a power transformer was developed.
This model is suitable for the study of transients induced due to VCB switching or other
types which contain a very wide range of frequencies. The following points have been
noted:
o Since the model is a linear black box equivalent based on measurements on the
terminals as opposed to a detailed model of the internal parts of a transformer, it
can only be used in transient studies involving overvoltages at the terminals, surge
transfer and how it interacts with other external components during these events.
o The transformer model is a linear passive network equivalent of RLC elements.
Therefore, non linearity and core saturation of the transformers have not been
accounted for. These characteristics are very pronounced for low frequency
overvoltages. However, the model can work well for voltages significantly lower
than the transformer rated voltage. In addition at relatively high frequencies where
the core excitation flux varies very rapidly, these effects are small and can be
neglected. Therefore, for high frequency transients, the model can be a sufficient
representation.
o It has been tried to validate the model developed by studying voltage transfer
from the delta side to the wye side and vise versa. Although good agreements
were obtained between voltage ratios calculated from the measured admittance
matrix, simulated voltage ratios and those directly measured as function of
frequency, there are discrepancies at lower frequencies. These errors are due to
the ungrounded windings of the delta side which produces very small zero
sequence components of elements associated with this winding at low
frequencies.
7.2 Future Work
A lot could be done to get an improved and more accurate model of the components
especially the transformer.
69
Cable Model
The cable model simulates most of the characteristics of a real cable and this was verified
to measurements on real cable. How ever the damping of the real cable was found to be
higher than the PSCAD/EMTDC model besides the model doesn’t take in to account the
presence of the armour. Hence more work could be done to get much more accurate
model in this regard.
VCB Model
The VCB model in this project simulates most of the phenomena of a real breaker very
accurately and hence the authors feel it is good enough for its purpose. However a more
accurate model can still be obtained by considering more aspects that were not dealt in
this work.
Transformer Model
It was seen that the transformer model developed in this work is not accurate at lower
frequencies; hence more work should be done to get a refined and robust model by taking
in to account a lot of aspects that were not taken in to consideration on this work. The
following is suggested:
Improving the measurement set up
The accuracy of this model depends on the accuracy of the measured admittance matrix
and hence accurate measurement is required as the model is very sensitive to error in
measurement.
One way to improve the accuracy of the measured data is by designing Special
measurement setup dedicated to measure the admittance matrix and voltage ratio which
allow easily reconnection and accurate measurement. One problem faced in this work
was the measurement set up was so simple that it doesn’t allow easily reconnection.
Reconnection has to be made every time a single element of the admittance matrix or
voltage ratio was measured. For the three phase transformer with 36 admittance matrix
and 9 voltage ratio elements for example 45 different connections was made. This was
time consuming not mentioning its effect on the accuracy of the data set due to bad
connection and inconsistency in the length of the connection cable used. This was
reflected on the measurements obtained on the admittance and voltage ratio. Besides the
simple connection cables used has greatly contributed to the measurement error. This was
also seen during measurements where some elements of the admittance matrix and
especially voltage ratio were very noisy. To deal with this problem the cables to be used
in future work should be shielded.
70
Measuring the Zero sequence Admittance
In accuracy of the model at lower frequencies for a transformer with ungrounded
windings like the transformers we dealt with can be overcome by explicit measurement
and modelling of the zero sequence system. Blocks of Y, which correspond to
ungrounded windings (
LH HL HH
Y Y Y , , ), are modified by removal of their zero sequence
components as the zero sequence component of the admittance matrix is inaccurate due to
small zero sequence current. A new zero sequence system has to be obtained by dedicated
measurements and the two functions can then be approximated with rational functions
which can be combined to produce the final model. In this work we have tried to measure
the zero sequence admittance though we didn’t succeed to get accurate measurement.
71
REFERENCES
[1] D. Chapman, “The Cost of Poor Power Quality”, Power Quality Application
Guide, Section 2.1, Copper Development Association of UK, November 2001.
[2] Vic Smith, Venthanar Ilango, Sarath Perera, Vic Gosbell, Duane Robinson,
“Transient Overvoltages On The Electricity Supply Network – Classification, causes and
propagation,” Power Quality and Reliability Center, University of Wollongong.
[3] A. Greenwood,”Electrical Transients in Power Systems”, 2nd Edition, Wiley
Interscience.
[4] Pritindra chowdhuri, “Electromagnetic Transients in Power systems”, 2nd
Edition.
[5] D.J.Clare. “Failures of encapsulated transformers for converter winders at Oryx
Mine”. Elektron magazine,March 1991, pp.2427.
[6] Van Craenenbroeck T, De Ceuster J, Marly J P, De Herdt H, Brouwers B, Van
Dommelen D, 2000,"Experimental and Numerical Analysis of Fast Transient Phenomena
in Distribution Transformers", Proc. IEEE/PES ,Winter Meeting, Singapore, CDROM
(6P).
[7] W. Sweet, _Danish wind turbines take unfortunate turn,_ Spectrum, IEEE, vol.
41,Issue 11, pp. 3034, Nov. 2004.
[8] J.Larssen, H.Soerensen, E.Christiansen, S. Naef, P.Völund “Experiences from
Middelgrunden 40 MW Offshore Wind Farm,” Copenhagen Offshore Wind, 2628 Oct.
2005.
[9] B. Gustavsen, “Panel session on data for modeling system transients: Insulated
Cables,” in Proc. IEEE Power Engineering Soc. Winter Meeting, 2001.
[10] B. Gustavsen, J. A. Martinez, and D. Durbak, “Parameter Determination for
System Transients Part II: Insulated Cables,” IEEE Transactions on power delivery, vol.
20, no. 3, July 2005.
[11] PSCAD manual.
[12] Janko KosmaE and Peter Zunko, “A Statistical Vacuum Circuit Breaker model for
simulation of transient over voltages,” IEEE Transactions on Power Delivery, Vol. 10,
No. 1, January 1995.
[13] “Over voltages – Measurement and Statistical Simulation”, Industrial Group,
TAVRIDA ELECTRIC.
72
[14] Mietek T. Gliiikowslti , hdoises R. Gutierrez and Dieter Brauii “Voltage
escalation and re ignition behaviour of vacuum generator Circuit Breakers during Load
Shedding,” IEEE Transactions on Power Delivery, Vol. 12, No. 1, January 1997.
[15] M. Popov, L. van der Sluis “Comparison of two vacuum circuit breaker arc
models for small Inductive Current Switching,”
[16] Helmer J., Lindmayer M. “Mathematical Modeling of the High Frequency
Behaviour of Vacuum Interrupters and Comparison with measured transients in power
systems,” XVIIth International Symposium on Discharges and Electrical Insulation in
vacuum, Berkley, USA, 1996.
[17] B. Kondala Rao, and Gopal Gajjar,”Development and Application of Vacuum
Circuit Breaker Model in Electromagnetic Transient Simulation,” IEEE Transactions on
Power Delivery 2006.
[18] R P P Smeets, “Stability of lowcurrent vacuum arcs,” J. Phys. D: Appl. Phys.1 9
(1986) 575587. Printed in Great Britain.
[19] S.M. Wong, L.A. Snider and E.W.C. Lo, “Over voltages and Reignition
behaviour of Vacuum Circuit Breaker”, International Conference on Power System
Transients – IPST 2003.
[20] S. Giere, H. C. Kärner, H. Knobloch, switching capability of double and single
break vacuum interrupters – experiments on real high highvoltage demonstrationtubes,
Technical University of Braunschweig, Germany.
[21] Marjan Popov, Lou van der Sluis, Gerardus C. Paap,and Hans De Herdt
“Computation of Very Fast Transient Overvoltages in Transformer Windings”, IEEE
TRANSACTIONS ON POWER DELIVERY, VOL. 18, NO. 4, OCTOBER 2003.
[22] A. Morched (SM), L. Mad (M), J. Ottevangers” A High Frequency Transformer
Model for the EMTP”, IEEE Transactions on Power Delivery. Vol. 8, No. 3. July 1993.
[23] Bjørn Gustavsen ,”Wide Band Modeling of Power Transformers”, IEEE
TRANSACTIONS ON POWER DELIVERY, VOL. 19, NO. 1, JANUARY 2004.
[24] B. Gustavsen and A. Semlyen, "Rational approximation of frequency domain
responses by Vector Fitting", IEEE Trans. Power Delivery, vol. 14, no. 3, July 1999, pp.
10521061.
[25] B. Gustavsen,” Rational approximation of frequency dependent admittance
Matrices", IEEE Trans. PWRD, vol. 17, no. 4, Oct. 2002, 10931098.
[26] http://www.energy.sintef.no/produkt/VECTFIT/index.asp
73
[27] B. Gustavsen, “Rational approximation of frequency domain responses by vector
fitting”, IEEE Transactions on power delivery, Vol. 14, No.3, July 1999.
[28] B.Gustavsen,”Enforcing Passivity for Admittance Matrices Approximated by
Rational Functions”, IEEE transactions on power systems,vol.16, No. 1,February 2001.
[29] Stefano GrivetTalocia, “ Passivity enforcement via perturbation of Hamiltonian
Matrices”, E transactions on circuits and systems—I: regular papers, vol. 51, no. 9,
september 2004.
[30] Maialen Boyra,” Transient Overvoltages in Cable Systems”, Master of Science
Thesis, Göteborg, Sweden, 2007.
[31] Daniel Mireanu, ” Transient Overvoltages in Cable Systems”, Master of Science
Thesis, Göteborg, Sweden, 2007.
74
APPENDICES
Appendix A. Implementation of 3phase Transformer
model in PSCAD/EMTDC
In this appendix the procedure for implementing the transformer model developed in this
thesis in transient time simulation software PSCAD/EMTDC is demonstrated.
The out put of the main Matlab source code is either a state space equivalent or a file of
branch cards of an equivalent RLC network of the transformer. The later can be used
directly in PSCAD/EMTDC or other time simulation packages to build the transformer
component. A branch based interface was used to define the branches of the equivalent
network. Branch design is accomplished by specifying the type and size of passive
elements (i.e. lumped R, L and/or C), and to which electrical connections they fall
between. The component definition is written as a script in the format
$<TO> $<FROM> [<R>] [<L>] [<C>]
where $<TO> and $<FROM> refer to the branch end nodes and [<R>] , [<L>] ,
[<C>] to the passive element values in Ohm, Henry and micro Farad respectively.
Nested page modules are used to build the model as the number of branches is too big to
implement as a single component. Therefore each one of the big branches between two
nodes or between a node and ground is modeled as sub modules which are then
connected to form the whole network.
The steps are summarized as follows:
1. Create a new component page module in the circuit page with 3 connections on the
left, 3 on the right and one at the bottom, ground to model the whole equivalent network
as in Fig. A.1.
2. Double click on the newly created page module and create components corresponding
to each one of the big branches. Connect these blocks between corresponding nodes
(connections) referred in 1 as shown in Fig. A.2.
75
Fig. A.1 Transformer component page module.
3. Enter component definitions for each one of the components in 2 as script. A typical
script page would look like as in Fig. A.3.
76
Fig. A.2 Components corresponding to internode branches.
77
Fig. A.3 Component definition as script page.
78
Appendix B. Network Realization
Consider the multiterminal network shown in Fig.B.1 the admittance matrix for this
network can be computed as
A diagonal element
ii
Y of the admittance matrix is obtained as the sum of admittances for
all branches connected to terminal i, including branches to ground,i.e.,
¿
≠ =
+ =
n
i j j
ij i ii
y y Y
, 1
(1)
The offdiagonal elements are the negative of the admittances connecting buses i and j
i.e.,
ij ij
y Y − = (2)
Where
i
y denotes the admittance of a branch connected between terminal i to ground and
ij
y admittance connected between terminal i and j.
Given the admittance matrix we can also compute the admittance of the branches
connected between terminals and a terminal to ground.
From (2) the admittance of the branch connected between terminal i and j may be written
as
ij ij
Y y − = (3)
Similarly the branch between terminal i to ground can be computed from (1)
¿
≠ =
− =
n
i j j
ij ii i
y Y y
, 1
(4)
If the value of
ij
y from (2) is replaced in to (4) then
¿
≠ =
+ =
n
i j j
ij ii i
Y Y y
, 1
(5)
We may write (5) in compact form according to:
¿
=
=
n
j
ij i
Y y
1
(6)
79
Given the admittance matrix of a device in frequency domain the corresponding network
can easily be obtained connecting branches between terminals and terminal to ground by
realizing the admittances obtained from (3) and (6). This gives a network of the form
shown in Fig.B.1.
Fig.B.1 Multi terminal network equivalent of a device.
Fundamentals of Network Synthesis
A network is represented by a rational function in sdomain. This rational function is
called network function. Network functions are expressed in terms of the complex
frequency variable s, resulting from the application of Laplace transformation to time
domain network equations. This network functions are either impedance or admittance.
Given the impedance or admittance function then it possible to find the network
representation.
As a passive net work is an inter connection of the basic passive elements; resistor,
capacitor, and inductor; let’s start by driving the network function of this elements which
is the basic for synthesis of a complex network including the network we are trying to
find for the transformer model.
The voltage and current at the terminals of these elements in time domain t and frequency
domain s is given in Table B.1 is given by
1
y
n
y
2
y
12
y
2
1
n
n
y
2 n
y
1
80
Table B.1 Voltage and current relation ship of passive elements
RESISTOR CAPACITOR INDUCTOR
Relation ship
between voltage
and current in time
domain
) ( ) (
) ( ) (
t Gv t i
t Ri t v
=
=
dt
t dv
C t i
v dx x i
C
t v
t
) (
) (
) 0 ( ) (
1
) (
0
=
+ =
}
−
}
−
+ =
=
t
i dt t v
L
t i
dt
t di
L t v
0
) 0 ( ) (
1
) (
) (
) (
Relation ship
between voltage
and current in
frequency domain
) ( ) (
) ( ) (
S GV S I
S RI S V
=
=
)] 0 ( ) ( [ ) (
) 0 (
) (
1
) (
−
−
− =
+ =
v S SV C S I
S
v
S I
SC
s V
S
i
S V
SL
s I
i S SI L S V
) 0 (
) (
1
) (
)] 0 ( ) ( [ ) (
−
−
+ =
− =
The impedance and admittance functions are defined as the ratio of voltage to current
transform and current to voltage transformer.
) (
) (
) (
S I
S V
S Z = (7)
) (
) (
) (
S V
S I
S Y = (8)
If the initial conditions in Table.B.1 are zero then the impedance and admittance
functions of the elements can easily be computed (7) and (8).
Table B.2 Network function of the basic elements
RESISTOR CAPACITOR INDUCTOR
Impedance
Z(S)
R
S I
S V
S Z = =
) (
) (
) (
CS S I
S V
S Z
1
) (
) (
) ( = =
LS
S I
S V
S Z = =
) (
) (
) (
Admittance
Y(S)
G
S V
S I
S Y = =
) (
) (
) ( CS
S V
S I
S Y = =
) (
) (
) (
LS S V
S I
S Y
1
) (
) (
) ( = =
Suppose Y(s) represent the admittance function of one of the branches of nterminal
network representation of the Fig.B.1 and assume that it can be written in partial fraction
expansion of the form
¿
+ +
−
=
N
e S d
a S
r
S Y
1
. ) ( (9)
e s d
ja a s
jr r
ja a s
jr r
ja a s
jr r
ja a s
jr r
a s
r
s Y . ...
) " (
" '
) " ' (
" '
) " ' (
" '
) " ' (
" '
) (
2 2
2 2
2 2
2 2
1 1
1 1
1 1
1 1
+ + +
+ −
+
+
− −
−
+
+ −
+
+
− −
−
+
−
=
(10)
81
e s d
a a S a s
a r a r S r
a a S a s
a r a r S r
a s
r
s Y . ...
" ' ' 2
" " 2 ' ' 2 ' 2
" ' ' 2
" " 2 ' ' 2 ' 2
) (
2
2
2
2 2
2
2 2 2 2 2
2
1
2
1 1
2
1 1 1 1 1
+ + +
+ + −
− −
+
+ + −
− −
+
−
= (11)
The network function in (11) can be realized in to its corresponding network. Each part
represent admittance of an element or combination of elements and the final network can
be obtained by connecting these branches in parallel as admittance in parallel sum up.
We can further simplify (12) by
..... ) ( ) ( ) ( ) ( ) ( ) (
5 4 3 2 1
+ + + + + = S Y S Y S Y S Y S Y S Y (12)
where
d s Y = ) (
1
e S S Y . ) (
2
=
a S
r
S Y
−
= ) (
3
Now let’s realize each section starting from the simplest ) (
1
s Y
From Table.B.1
d s Y = ) (
1
,represents admittance of a resistor with a resistance value of
d
R
1
=
S e S Y . ) (
2
= ,represents admittance of a capacitor with a capacitance value of e C =
One can write ) (
3
S Y as
) (
1
1
1
) (
3
3
S Z
r
a
S
r
S Y =
−
= , where
r
a
S
r
S Z − =
1
) (
3
is an impedance function representing
a series combination of an inductor with an inductance value of
r
L
1
= and a resistor with
a resistance value of
r
a
R
−
= as impedances in series sum up.
)
'
' '
1 ( ' '
) ' ' ' ' 2 ' ' 2
)
'
' '
1 ( ' '
' 2
1
' 2
'
' 2
' ' ' '
.
2
1
1
" ' ' 2
" " 2 ' ' 2 ' 2
) (
2
1
2
1 2
1 1 1
2
1
2
1 2
1
1
1
2
1 1
2
1
2
1 1
2
1 1 1 1 1
4
r
r
a
a r a r
S
r
r
a
r
r
a
r
r a
S
r
a a S a s
a r a r S r
S Y
+
−
−
+
+ − +
=
+ + −
− −
=
(13)
The above equation can be written as
82
) ( ) (
1
) (
5 4
4
S Z S Z
S Y
+
=
) (
4
S Z and ) (
5
S Z are in series and each can be realized as follows
) (
4
S Z is a series combination of an inductor with a resistor whose values can be
computed as follows
' 2
'
' 2
' ' ' '
.
2
1
) (
1
1
2
1 1
4
r
a
r
r a
S
r
S Z − + = (14)
(14) can be written as:
R LS S Z + = ) (
4
(15)
Equating (14) and (15) gives the values of L and R as
1
2
1
r
L = ,
' 2
'
' 2
' ' ' '
1
1
2
1
r
a
r
r a
R − =
)
'
' '
1 ( ' '
) ' ' ' ' 2 ' ' 2
)
'
' '
1 ( ' '
' 2
1
) (
2
1
2
1 2
1 1 1
2
1
2
1 2
1
5
r
r
a
a r a r
S
r
r
a
r
S Z
+
−
−
+
= (16)
Similarly if (16) is rearranged as
G CS
S Z
+
=
1
) (
5
(17)
Then ) (
5
S Z can be realized as a parallel combination of a capacitor and a resistor whose
values can be computed by equating (16) and (17).
)
'
' '
1 ( ' '
' 2
2
1
2
1 2
1
r
r
a
r
C
+
= ,
)
'
' '
1 ( ' '
) ' ' ' ' 2 ' ' 2
2
1
2
1 2
1 1 1
r
r
a
a r a r
G
+
−
− =
The final network will then look like:
83
Fig.B.2 Network realization of Y(s).
...
0
R
0
C
r
R
r
L
c
R
c
L
c
C
c
G
...
Constant term s.e term
Real Pole
Complex conjugate pole
THESIS FOR THE DEGREE OF MASTERS OF SCIENCE
Analysis of Transients in Wind Parks: Modeling of System Components and Experimental Verification
Abey Daniel & Samson Gebre
Performed at: ABB Corporate Research Västerås,Sweden Advisor: Ambra Sannino ABB Corporate Research Västerås,Sweden Examinar: Torbjörn Thiringer Division of Electric Power Engineering Department of Energy & Environment Chalmers University of Technology
Department of Electric Power Engineering Division of Energy and Environment CHALMERS UNIVERSITY OF TECHNOLOGY Göteborg, Sweden, 2008
Analysis of Transients in Wind Parks: Modeling of System Components and Experimental Verification ABEY DANIEL & SAMSON GEBRE
© ABEY DANIEL & SAMSON GEBRE, 2008
Department of Energy and Environment Chalmers University of Technology SE412 96 Göteborg Sweden Telephone + 46 (0)31772 1000 Email: abiy1882@yahoo.com
gebrsa@yahoo.com
Chalmers Reproservice Göteborg,Sweden 2008
An RLC equivalent network is realized and implemented in the time simulation software PSCAD/EMTDC. Although a lot has been done and fairly good models of cables and VCBs have been developed. i . The curve fitting algorithm is used to approximate the admittance matrix.ABSTRACT This thesis deals with analysis of transients and experimental verifications in wind parks. Critical as it is in transient studies. Secondly. it seems that a unified wide band model of transformers which works well for broad range of frequencies has been the most difficult to develop. a statistical model which emulates the occurrence of reignitions. In the first part of the thesis. a terminal model of a transformer which represents its wide frequency response is developed based on experimental measurement of admittance matrix over a wide range of frequencies. damping and reflection is used to model the cables using PSCAD/EMTDC. a frequency dependent model which represents high frequency effects like wave propagation. The model is also validated experimentally. vacuum circuit breakers and transformers which make any branch in a wind park. Models are also validated where possible. modeling of such a transformer has been extensively treated in this thesis. Finally. prestrikes and current chopping behaviors of the VCB during opening or closing operations is dealt with. The focus is on accurate modeling of system components such as cables.
We would like to express our deepest gratitude to our supervisor Ambra Sannino for providing us with the opportunity to work in an excellent academic environment and for her brilliant ideas. Regrettably we will not be able to recall all of them. who is responsible for helping us with the modeling part and his constant guidance. Last but not least. we are deeply indebted to our friends and families for their support and advice. especially we are extremely grateful to Henrik Breder. He was always there to listen and to give advice.Yuriy Serdyuk. suggestion and valuable input to this work. generosity. guidance and inspiration. Leif Hederström.Dierk Bormann. excellent guidance and for the care with which he reviewed the thesis report.assistance. and advice we received. ii . for the warm welcome . for the valuable discussions. We want to thank Nilanga Abeywickrama. Josef Medvegy. We are deeply grateful to our examiner Torbjörn Thiringer. Special thanks goes to Tarik Abdulahovic. We would also like to acknowledge all of the other people who have helped and encouraged us. but here is a good start. We have learnt a great deal from those who have worked with us over the months and gratefully acknowledge our debt to them.PREFACE This thesis work was conducted at ABB corporate research.
........... 1 1....1 Current chopping ................4.........................3 Testing Model .............................................................2 Voltage Escalation ..............................2 Dielectric strength....2........................................................................... 30 4........................................................................................................2 Insulation and Semiconducting screen.........................................................2 Overview of earlier work ................................................... 14 3................................................................1 Core Parameters .................. 14 3...........4 Breaker unsuccessful and high frequency current successful operation....................................................................................................................................................................... 29 4....1 Introduction.......1 Measurement SetUp .................4 Traveling Waves ..........................................3........... 25 4....................................4......5 Conclusions..................................1 Introduction.4 Simulation Results .. 12 3.................... 28 4... 22 4.........................4 Results and Analysis ......................3 High Frequency Quenching Capability ................................................5 Prestrikes ...2 Cable Parameter Selection in PSCAD/EMTDC ...................................................................2....................................1 Current chopping ............................................. 10 3....................................................................4 Prestrikes.................. 33 4......... 5 2...............CONTENTS ABSTRACT................................................................. 17 3......... 10 3.......2 Effect of rate of rise of dielectric strength ..............................................................................................................................................................................................3.........................4 Thesis outline .................. 11 3...1 Introduction................................................. 27 4......... 2 2 TRANSIENT OVERVOLTAGES .....5 Conclusions....... 7 3 CABLE MODELING ................................2......................................... i PREFACE...................................... 32 4..............................................................................................2 Phenomena causing over voltages in VCB.................................................................................................................................................... 25 4.........................................................1 Effect of Arcing Time...............................3 Shunt Capacitor Switching Transients................ 34 5 TRANSFORMER MODELING Background and Theory .....3 Sheath conductor.................................................................................... 22 4..................................2.................................. 14 3................................................................. 27 4...3 Effect of high frequency quenching capability............3.................... ii CONTENTS................ 2 1....................3 Aim of this thesis ...............................2........................................................................................................................................... 24 4................................................................................................................................................................. 1 1........................................................................................................ 23 4...............................................................................................2............................................................................................ iii ABBREVIATIONS ................................................3 Modeling of VCB ............................... 22 4.............................................4............................................................................................................. 24 4..2 Classification................................................. 2 1............................ 35 iii ..... 26 4.......1 Introduction.....................................................................3 Virtual Current Chopping ..........................4........ 21 4 VCB MODELLING...............................................................................................3.............. 4 2................................................. 4 2........................2................................ v 1 INTRODUCTION... 35 5.......4.......................................................................................................................................... 13 3.. 22 4......1 Problem description ................................................... 4 2...............
.................................................... 67 7............................1 Conclusions......................................................................2.2 Rational Approximation of Frequency Responses by Vector Fitting................................................... 36 5....................................2.................................................5 Model Validation ..................1 Introduction...... 53 6..................... 67 7......................................................................................................................................... 74 Appendix A..........................1 Measurements ..................................2...........................................2 Wide Band Modelling of Power Transformers..............2 Measurement Setup.. 47 6.......... 51 6. 66 7 CONCLUSIONS AND FUTURE WORK ......................6 Accuracy of the Model.................................................................................. 60 6...................................... 74 Appendix B................ Network Realization...2 Future Work ....................................................................... 78 iv ............ 47 6.................................................7 Conclusion ................ 71 APPENDICES ..................4 Rational Approximation of Admittance Matrices of The Transformers .......................................................5............................................................................................ 36 5................ 44 6 TRANSFORMER MODELINGExperiments and Results................................3 Time Domain Implementation ............................................ 40 5...... 65 6........3 Measurement Results ................................................ Implementation of 3phase Transformer model in PSCAD/EMTDC................................................................................................................................................. 47 6............................................... 68 REFERENCES.............
ABBREVIATIONS VCB BIL MV WP EMTP PSCAD EMTDC TOV EM RRDS Vacuum Circuit Breaker Basic Insulation Level Medium Voltage Wind Park Electro Magnetic Transient Program Power System CAD Electromagnetic Transient with DC Transient Overvoltage Electromagnetic Rate of Rise of Dielectric Strength v .
1 Problem description Nowadays due to the development of modern circuit breakers. Worldwide many transformer insulation failures have been reported caused by switching transients [5]. where all transformers and generators had to be moved to shore for maintenance. high voltage transients are suspected to have contributed to the failure [7]. Transient overvoltages and other power quality disturbances cause billions of dollars of losses each year due to damage to equipment and loss of production [1]. One example could be Denmark’s Horn Rev. The experimental set up is also useful to study large cable systems such as MV undergrounding and industrial systems. 1 . while those transformers had previously passed all the standard tests and complied to all quality requirements. Due to the unpredictable threat of TOV in WPs and large MV systems it is worth studying fast switching TOV in wind parks and MV systems experienced due to the unavoidable switching action of VCB. These switching transient over voltages have the potential to result in large financial losses. Especially the vacuum circuit breaker (VCB). The use of a cable network and VCB in medium voltage (MV) networks could be a source of insulation failure due to the fast voltage transients caused by multiple restrikes of VCB.Chapter 1 INTRODUCTION 1. These phenomena can occur in both distribution and transmission networks. one of the world’s biggest off shore wind farms built to date. may cause switching overvoltages with fast rise time and thus degrading the insulation and eventual failure due to the cumulative effect of voltage transients of even less magnitude than the BIL. The problem is generally associated with the fast transient over voltages produced by restrikes and prestrikes during the opening or closing of a switching device. To analyze these phenomena. the switching overvoltage has become a more sever problem. to some extent caused by the fact that the current interruption capabilities of these breakers have improved significantly [3]. Fast switching. vacuum and SF6 for example. including one feeder with one turbine under test and the adjacent tower and adjacent feeder to reproduce a unique wind farm in a lab. [6]. There is a similarity between wind parks (WP) and industrial MV networks in this regard as both use VCB as switching device and consists of a cable network. Hence there is a concern of transient overvoltages (TOV) in wind parks as well. an experimental setup was designed in ABB Corporate Research with threephase cable representing a part of a wind park. which shows a high ability of interrupting high frequency currents of several hundreds of kHz.
2 Overview of earlier work A number of investigations have been done related to this work. and supply side and load side transformers. The two systems are then simulated and. are discussed in detail and modeled in PSCAD/EMTDC. but the interest in this section lies on overview of the two works by Daniel Mireanu and Maialen Boyra as this work is the continuation of these two. The first work provides an analysis of TOVs and their propagation in cable systems [31]. vacuum circuit breaker and transformer which comprise a branch in any wind park.4 Thesis outline Chapter 1 gives a general background of the thesis. In the second work. Particularly. The model derived in this chapter proposes a method to account for the effects of stranding of the conductor and semi conducting layers for use in PSCAD/EMTDC. Chapter 3 presents modelling of a power cable which is to be used later in the simulation of the whole system. As the work is a continuation of [30] and [31]. one that enables representing the transformer by using a single and unified terminal model for a very wide band of frequencies.3 Aim of this thesis The aim of this work is to study and characterize high frequency transients in wind parks. [30] and [31]. TOVs in cable systems are characterized through both simulations in PSCAD/EMTDC and laboratory experiments which characterize a section of wind power plant [30]. Chapter 2 presents an overview of TOVs. where available. a particular interest is to refine and improve models of the aforementioned components. a different approach is used to model the transformer than those used in [30] and [31]. The laboratory system was modeled. 1. two essential components in any MV cable system. An overview of previous literature by is presented. The aim of this thesis work is defined as well. Comparison of results obtained between the model and measurement data during cable discharging is also included. It is not the intention of this section to summarize these valuable works. their causes and how they propagate through the system.1. the results are compared to measured data for verification. This investigation is performed through the use of two systems: an industrial steel mill electrical system and a wind farm. how they are classified. A single phase laboratory set up was used which includes two single phase cables. A crucial part is accurate modeling of the cable. 2 . Cables and vacuum circuit breakers. Analysis of measurement and simulation results were also made for different loading conditions. a VCB in between. 1.
A transformer model compatible with electromagnetic transient programs (EMTP) can then be obtained by network realization of the rational function. The admittance matrix experimentally determined is subjected to a rational function approximation by vector fitting. Chapter 7 gives conclusions of the thesis work and suggests future work. The model simulates most of the effects (phenomena) of a VCB. Results obtained from model testing are also included in this chapter. Chapter 5 and 6 propose the modelling of a transformer by measuring the admittance matrix in a wide frequency domain. 3 .Chapter 4 Describes the VCB model treated in this paper.
The overvoltage developed in this case is limited by the impedance of the system seen by the lightning current. 4].2 Classification TOVs are broadly classified as impulsive and oscillatory. a switching operation or a fault) or network stimuli (e.1 Introduction TOVs in electrical transmission and distribution networks result from the unavoidable effects of lightning strikes and network switching operations. (2) Oscillatory transients Switching operations within the network are a major cause of oscillatory overvoltage transients. Lightning is associated with the discharge of a large current. a lightning strike). 2. Such operations include switching of circuit breakers to clear faults.Chapter 2 TRANSIENT OVERVOLTAGES 2.g. (e. A brief explanation and analysis of transients can be found in references [3. This situation usually causes power system faults due to insulation failure which in turn causes a supply interruption and voltage sags throughout the distribution network. Its duration is in the range of microseconds to milliseconds. A TOV can be defined as the response of an electrical network to a sudden change in network conditions. switching 4 . either intended or accidental. (1) Impulsive transients The most common cause of impulsive transients is lightning. This current can reach up to 200kA. In this chapter. This overvoltage can reach several MV. the fundamental concepts of TOVs with especial emphasis on switching transients is presented to facilitate understanding of the next chapters as the transients dealt with in this work are of this type caused by switching operation of VCBs. Lighting current and its corresponding voltage rises to its peak in a time which varies from less than a microsecond to 10 or 20 microseconds and then decays in a few hundred microseconds [3]. Impulsive transients are characterized by their rise and decay time and peak values.g. A transient is a natural part of the process by which the power system moves from one steady state condition to another [2].
The faster a switch tries to interrupt a current.e. as every electrical system can withstand a maximum level of voltage. An example of a capacitor switching waveform is shown in Fig. However. the system maximum voltage strength is reached causing a discharge and a temporary or a permanent damage. They are switched in and out depending on the level of support needed. a voltage is developed by the system inductance to oppose the change in current.u. the voltage of the bus bar to which it is attached will momentarily collapse as the voltage across a capacitor can not change instantaneously. Capacitor switching transients can be magnified to quite high values as they pass to a lower voltage level if certain network conditions are met. and to reduce energy costs by improving the power factor of the system.e. This can occur when transients originating in the medium voltage (MV) distribution network move into the low voltage (LV) network and there are power factor correction (PFC) capacitors present at LV consumer’s installations. These capacitors are implemented in the system in order to control the system voltage. which is manifested by the generation of a counter EMF in the inductance in such a direction as to keep the magnetic flux (and the current) in the inductance constant. the current) will be opposed by the inductance. The energy originally stored in the inductance is transferred back and forth until it is finally damped by the system resistance. the circuit recovery voltage causes the dielectric between the switch contacts to break down and capacitor current is reestablished). Any effort to change the magnetic flux (i.3 Shunt Capacitor Switching Transients Shunt capacitors are used extensively in power transmission and distribution systems as a means of supplying reactive power for voltage support.When current flows through an inductance. to increase the power transfer capability. the higher the resulting switching surge voltage is.1 which shows an MV/LV network supplied by an MV feeder 5 . This is followed by an oscillatory transient recovery voltage which can be as high as 2p. In a system with large physical extension. such as a power transmission line where the inductance and the capacitance are distributed. 2. to reduce equipment loading. Therefore. capacitance and resistance) associated with an electric network .2. thus transferring the energy from the inductance to the capacitance. When a discharged capacitor is energized. Dangerously high voltages can be generated by multiple reflections at the end of the line. all electrical systems have capacitances. Such a situation is given in Fig.2. The same effect can occur when a capacitor is switched off and a restrike occurs (i. the current through the inductance flows through the capacitance. In practice. A switching overvoltage or as they are some times called switching surge is generated due to the interaction of the inherent elements(inductance.of utility capacitor banks and switching of distribution feeders to rearrange the network for maintenance or construction. when a circuit breaker tries to interrupt a current. When the current is interrupted in an electrical system. 2. Theoretically this voltage can be infinite if the current through the inductor is snapped off instantaneously. traveling voltage and current waves are generated during a switching operation. it produces magnetic flux.
The overvoltage associated with this phenomena can reach to a peak value up to 4 times the corresponding power frequency system voltage. L 1C 1 ≅ L 2 C BRK C2 2 (2. This condition is mathematically represented by P+jQ LV Bus L2 L1 BRK MV Bus #1 #2 R1 C1 Fig.with inductance L1. The series combination of the outgoing MV feeder and the MV/LV transformer has an inductance of L2.2. C1 is the capacitance of switched capacitor.1 Power system network used in PSCAD/EMTDC to illustrate magnification of capacitor switching transients. A switched capacitor C1 is connected to the MV bus bar. Magnification occurs when the predominant frequency of the switching transient is approximately equal to the resonant frequency of the LV system.1) where L1 is the system source inductance seen from MV bus bar. L2 the inductance of stepdown transformer and associated MV feeder and C2 is the capacitance of LV PFC capacitor. 6 . The PFC capacitor C2 is connected to the LV bus bar.
.0 time[s] 0...0 2. Higher frequency transients will be more affected by the stray and distributed inductance and capacitance of the system than lower frequency transients. as well as transformer and generator windings are distributed.. The ratio of the amplitudes of the voltage and the current waves on a 7 .40 0.u] 10.0 Curent Voltage I[p. A characteristic of a circuit with distributed parameters is its ability to support traveling transient waves of voltage and current. cables.50 . Fig. The current and voltage waves travel in both directions from the point of excitation or disturbance. ..10 0.00 0.0 3.4 Traveling Waves The parameters of transmission lines.. . The influence of the distributed parameters on the propagation of TOV depends on the frequency content of the waves.0 2.u] 0.0 1.2.0 1.30 0.2 Voltage and current resulting from capacitor switching simulated.20 0.0 V[p.0 10.Over voltage and current transients due to shunt capacitor switching 3. 2.
It is near the speed of light (3x108 m/s) for overhead lines. there is a strict proportionality between the two. where the characteristic impedance changes.transmission line or cable is called the characteristic impedance Z 0 of the line and is for a lossless line given by Z0 = L . When a wave arrives at a point of discontinuity. Reflection and refraction of traveling waves When voltage and current waves propagate in transmission lines or cables. The velocity of propagation of the waves for a lossless line is given by v= 1 . m/s LC (2. respectively.4) 8 . possibly causing quite high voltages at some points. one reflected back superimposed on the incident wave and another transmitted beyond the discontinuity.3) and depends on the medium of propagation. to keep this proportionality. The magnitudes of the reflected and refracted voltage waves at a junction point with characteristic impedances Z a and Z b on the incident side and refractive side. an underground cable or transformer. C (2.2) where L and C are the distributed inductance and capacitance respectively of the line or cable. can be quantified as V2 = Zb − Za V1 Z a + Zb (2. two new wave pairs are generated. Like all other waves. traveling waves initiated by disturbances in power systems also have classic wave characteristics like reflection and refraction. Typical values of the characteristic impedance vary between 300 to 500 for overhead lines and for cables from 30 to 60 [2]. The consequence of these discontinuities could be the production of complex sets of traveling waves which add and subtract. The amplitudes of the reflected and refracted waves are such that the voltages to current proportionalities are preserved for each. and between one half and two thirds of this value for underground cables [2]. The proportionality constant is the characteristic impedance of the line or cable. which could be an open end.
b Z a + Zb Z a + Zb refractive coefficient.V3 = 2Z b V1 Za + Zb (2.6) (2. The reflected and transmitted currents are given by V2 Za V I3 = − 3 Zb I2 = − (2.7) 9 . V2 is the reflected wave and V3 is the refracted Z − Za 2Z b is called the reflection coefficient and is the (transmitted) wave.5) where V1 is the incident wave.
it may not be good if the conductors are not ideally transposed. etc. A model is developed in PSCAD/EMTDC which suitably works not only at power frequency but also at higher frequencies. R. etc. inductance and resistance per meter are well accounted for should be used to include high frequency consequences such as skin effect. iii) Frequency dependent(phase) model This model also represents well the frequency dependence of the distributed parameters over a wide range of frequencies. ii) Frequency dependent (mode) model This model represents well the frequency dependence of the distributed parameters over a wide range of frequencies. Thus. at a wider range of frequencies. It can be sufficient to model a power cable by using simple RLC network to study its response at a single frequency such as steady state. for higher frequencies. However. PSCAD/EMTDC provide three options for modeling of a power cable [11]: i) Bergeron model This model treats the cable as an infinite number of represents accurately only one frequency. The objective is to find a close representation (model) of a real cable that best depicts its response. reflections. Both single phase and three phase cables have been modeled. These models are later to be used to simulate and study effects of TOVs in real power networks during switching operations. but uses a constant transformation of internal matrices. The frequency dependence of sections of L. and C. frequencydependent models where the frequency dependence of the distributed cable parameters such as capacitance.1 Introduction This chapter treats modeling of a real power cable in PSCAD/EMTDC. etc. It 10 . skin effect. The model can then be used to study the effect.Chapter 3 CABLE MODELING 3. of different network operations (both normal and abnormal) in a real medium voltage power network of which the cable is a part. earth faults. The focus is on high frequency effects like EM transient propagation.
Both Z and Y are functions of frequency .2 Cable Parameter Selection in PSCAD/EMTDC The basic parameters used to represent cables and transmission lines are series impedance matrix Z and shunt admittance matrix Y [9]. Z (ω ) = R (ω ) + jωL Y (ω ) = G (ω ) + jωC (3. Fig. There is a major difference. therefore. shunt conductance and shunt capacitance respectively.2) where R.3. semiconductor and sheath. C are series resistance. In PSCAD/EMTDC the cable parameters are calculated based on the geometry and material property of the cable specified by the user. series inductance. As it is clearly shown in Fig. This model has been used in our work. G. on the geometric representation of the actual cable from that used in PSCAD/EMTDC as the later does not account for the presence of a semiconducting screen that is present in the actual cable and it assumes that the core conductor is a homogeneous solid conductor which is different from the core conductor made up of many strand conductors in the actual cable. This model is.1 PSCAD/EMTDC requires the geometric parameters for the core.1 below shows the difference between the actual cable and its representation in PSCAD/EMTDC. L. [10]. the best of all three. EMTPtype programs including PSCAD/EMTDC have dedicated support routines for calculating an electrical representation of cable system in terms of the series impedance matrix Z and admittance Y [9].1) (3.3. an 11 .3.1 Actual cable versus PSCAD/EMTDC representation. [10]. however. Cable # 1 0[m] d [m] Conductor Insulator 1 Sheath Insulator 2 r1 r2 r3 r4 Fig.internal transformation matrices is well taken care of. All the commonly used programs for simulation of transients. An actual cable however includes a conductor. 3. The situation is further complicated by the fact that some of the geometrical dimensions of the cable are not given by the manufacturers.
is the resistivity of the core material. Accordingly the user needs to decide how to represent the following cable features which PSCAD/EMTDC fails to represent. (3.2. The actual cross section area of the core given by the manufacturer is different from the area of circle obtained using the radius of the conductor given by a manufacturer for stranded core design.2.3) Ac where ´ is the corrected resistivity to account for the space between the strands. It is obvious that the area of the core needed for resistance calculation in PSCAD/EMTDC is obtained based on the radius specified.4) 12 . r 1 is the radius of the conductor given by manufacturer. π .5) l where l is cable length.1 Core Parameters The main parameters of the core to be given to PSCAD/EMTDC are the resistivity and radius. Resistivity of commonly used core materials is given in Table 3. Stranded conductors can be modelled in PSCAD/EMTDC in two different ways. 3. the corrected resistivity can alternatively be calculated as πr12 ρ ' = RDC (3. The inner radius in this case is given by r in = r1 − 2 Ac If the manufacturer provides the DC resistance for the core RDC. The second way is to model it as a hollow cylinder whose cross section is the difference of areas between two circles with area equal to Ac as shown by Fig.inner semiconducting screen insulator.3. The first way is to model it as a solid conductor by increasing the resistivity by a factor equal to the inverse of the fill factor as given by the formula below πr12 ρ'= ρ (3. 1) 2) 3) Core stranding Inner and outer semiconductor layers The wire screen The following sections are dedicated to obtaining the cable parameters for simulation in PSCAD/EMTDC.1. an outer semiconducting screen sheath among others. and Ac is the efficient cross section area of the core. An attempt will be made to represent the above features. hence there is a need to make some form of correction to account for the fill factor as to give the correct resistance of the conductor.
1 Resistivity of commonly used core material. This data conversion procedure is summarized as follows. and increasing the permittivity proportionally to leave the capacitance unaltered. Table 3. [10]. As explained in [9] and [10]. Unfortunately.2 Alternative approach of dealing with stranding of core. 1) Calculate r 2 . If the capacitance is unknown. the permittivity is corrected as ε r1 = ε rins ln(r2 / r1 ) ln(b / a ) (3.m] 1.854E −12 .7) 13 .3. MATERIAL COPPER ALUMINIUM −8 [ .82 E −8 3.72E 2. These must therefore be introduced by a modification of the input data. none of the EMTPtype programs for instance PSCAD/EMTDC in our case allows the user to directly specify the semiconducting layers [9]. this is done by allowing the insulation to extend between the core conductor and sheath conductor. 2) Calculate the corrected permittivity r1 as ε r1 = C ln(r2 / r1 ) 2π 0 (3. r 1 plus the sum of the thickness of the semi conducting screens and the main insulation. This radius is given as outer radius of the insulation to PSCAD/EMTDC.Cable # 1 0[m] d [m ] Conductor Insulator 1 Sheath Insulator 2 rin r1 r2 r3 r4 Fig.6) where C is the cable capacitance and 0 =8.2.2 Insulation and Semiconducting screen The main insulation of highvoltage cables is always sandwiched between two semi conductive layers.
8) The other input data that needs to be specified to PSCAD/EMTDC that is worth mentioning is the resistivity of the sheath material.3 Sheath conductor PSCAD/EMTDC allows the user to specify the sheath conductor. 3.2.where a and b are the inner and outer insulation radii respectively and ε rins is the permittivity of the insulation material. the best way to model it is with a tubular conductor having a cross section area equal to the total wire area A s . 3. damping (due to skin effect and proximity effect) and reflections.1. as it is the case with the cable we are dealing with.3. The measured waveforms are compared with results obtained by simulating the same operations using the frequencydependent cable models developed in 3.2 using PSCAD/EMTDC. 3. Materials commonly used as sheath conductor along with their resistivity is given in Table 3.1 Measurement SetUp Fig. When the sheath conductor is made of wire screens. The cables at the wind cable lab were charged to 800V and discharged through a COMgap to excite a very steep fronted voltage at one end and measure the transients at the other end for different lengths and arrangements of the cable system. inner sheath radius of r 2 and outer radius r 3 given by r3 = As π + r22 (3.3 Experimental set up for cable discharge with step.3. 14 .3 Testing Model The cable models developed in the foregoing sections are tested to characterize different frequency dependent effects like EM propagation.
4) and on cable sections with 242m and 484m lengths of the cable lab set up after the cable had been cut and installed. The cable was charged by increasing the variable DC voltage source until the comgap break down and then the cable is discharged through the comgap.3).3.2 Specifications of LeCroy 9354A Digital Oscilloscopes.3 U1 and U2 represent the input and out put measured points. ELFA AC240 comgap (see Table 3. and LeCroy Oscilloscope for measuring waveforms(see Table 3. BANDWIDTH 500MHZ Channels 4 Sample Rate 2GSa/sec Memory Depth 50K pt/sec Sensitivity 2 mV/div to 5V/div Number of Bits 8 8 bits Input Impedance 1 MOhm/14pF or 50Ohm± 1% Main time base lowest 1 ns/div Table.The specifications of the cable is given in Table 3. The experiment was done on a 620m cable while it was rolled on drum (see Fig. a current limiting resistor of 1k . In the set up shown in Fig. which propagates on to the other end of the cable until it is finally damped by the system resistance. PARAMETER MINIMUM DC Breakdown Voltage 425V Impulse BreakdownVoltage Insulation Resistance 10 10 Capacitance Operational Temperature 40°C AC FollowOn Current >300A MAXIMUM 800V 1PF +125°C 15 .3.2).The experimental set up is based on a variable DC source. 3. a step excitation is applied between core conductor and the sheath when the gap discharges. In other words.3 ELFA AC240 Comgap product specification.4 Table.3.
5/1 Copper 3x98 THICKNESS [mm] Core Main insulation Outer insulation Inner/outer semiconducting screen Sheath conductor Aluminum XLPE 16 . Table.5 0.3.4 three phase 24 kV 620m Sea cable on drum and its cross section.Fig.4 Specification three phase sea cable.3. CORE MATERIAL ALUMINIUM SOLID CONDUCTOR Core cross section 240mm2/phase Core radius 9mm Main insulation material Mainly of PE Screen conductor Copper Screen cross section 85mm2 Armor material Steel MATERIAL CROSS SECTION AREA [ mm 2 ] 3x240 RADIUS [mm] 9 6 22.
The traveling time of the reflected waves between the discharging COMgap and the cable open ends also becomes shorter for the shorter cable since the propagation velocity for the same type of cable is identical whatever the length is.5 and 3.6.5mm 3.5mm) 2 =17. r1 =9mm outer semiconducting screen thickness r2 =9mm+6mm+0.The experimental set up in Fig.3.3. The traveling time is equal to a quarter of the propagating pulse cycle.4 Results and Analysis When a cable is energized with a fast fronted impulse.8 are results of simulations in PSCAD/EMTDC using the frequency dependent cable model.10 are “zoom ins” of these in order to show the pulses more clearly.5mm r2 = r1 +Main insulation thickness+ sum of inner and Cable # 1 d [m] 0[m] r3 = As π + r22 = 85mm 2 π + (16. The waveforms in Fig. a traveling wave is likely to be incepted.3 was simulated in PSCAD and the cable was modeled according to the modeling procedure given in section 3.3.10 (a) which present the measured waves at the end of a 242m and 484m cables. This is demonstrated by the results in Fig 3.85 Corrected permittivity ε r1 = ε rins ln(b / a ) r 4 =22. Some characteristics inherent to traveling waves in cables and transmission lines are reflections. One important requirement for the formation of traveling waves is that the duration of the wave should be significantly less than the time taken to traverse the medium.5mm b=a+main insulation thickness b =9.5mm+1mm=16.2. The propagation delay for waves to reach the open end of the 242m cable has been measured to be 1.5mm=9. propagation delay and damping.5mm+6mm=15. the more frequent reflections the cable accommodates and the higher the frequency of the pulses. Subsequent Figs 3. The shorter the cable length.9(a) and Fig.3.7kHz.9 and Fig3.3. The frequency of oscillation for the 242m cable is approximately 154kHz which is almost double that for the 484m cable which is 76.10 (a).5mm ln(r2 / r1 ) =2. This can be observed in Fig 3.9(a) and Fig. Fig.The parameters of the cable model are accordingly given below.3mm Conductor Insulator 1 Sheath Insulator 2 r1 r2 r3 r4 a = r1 +thickness of inner semiconducting screen a =9mm+0.7 show the results of the experimental tests for two different cable lengths.634us 17 . and 3.
1) to be 0. Comparable figures are also found for the pulse frequency and propagation time which are in reasonable agreement with those found in the measurements. for the 484m cable length 175m / us (3. The cable oscillation frequency indirectly affects the damping of the propagating transient due to the conductor ac resistance being increased by the skin effect. S v where S is the length of the cable yielding Tp = 242m =1.9(b) and Fig. damping of the waves increases with frequency. Similar results are obtained by simulating the same operations on the cables using the frequency dependent model in 3.121nH/m and 0.269nF/m respectively yielding v= This gives a traveling time of 1 0. One source of this error could be due to the additional capacitance to ground from the measuring probes which may have decreased the frequency in the measurements as f0=1/2 LC where f0 is the oscillation frequency.3.2. Thus. Measured and simulated values are given in Table 3.3) and L and C have been calculated from the cable property and geometry (see section 3.10 (b) signify a higher frequency of oscillation. shorter propagation delay and faster damping for the 242m than for 484m cable.634us and is 3.5 for comparison.26us for the 484m cable.9) Tp = Tp = It should be noted that the oscillation frequency is slightly higher in the simulations.38us.269nF/m =175m/us .3. 18 . Similar results are found by computing the above parameters from the propagation velocity of the cables.76us. The propagation velocity can be calculated from the inductance L and capacitance C per meter of the cable as given in (2.121nH/m * 0. for the 242m cable length and 175m / us 484m =2.delay for waves to reach the open end of the 242m cable has been measured to be 1. This can be seen in the same figure where we have quicker dying of the reflected waves on the 242m cable than on 484m cable due to higher frequency. Fig 3.
634 3.789 Travel Time (Measurement) [us] 1. 19 .26 1000 800 600 400 200 0 200 400 600 800 1000 100 0 100 200 300 400 time[us] 500 600 700 800 900 Charging end L11 Measuring point L21 vo g [v] lta e Fig.Table.08962 0. 1000 800 600 400 v lta e ] o g [v 200 0 200 400 600 800 1000 100 0 100 200 300 400 time[us] 500 600 700 800 900 Charging end L11 Measuring point L12 Fig.4 484m 0.6 Cable energizing at L11 and measuring point at L21 ( simulated).3.5 Comparison of measured and simulated values.0767 2. Travel Time (Simulation) (Measurement) (Simulation) [MHz] [MHz] [us] 242m 0.5 Cable energizing at L11 and measuring point at L21 (measured).3. Cable length Oscillation freq.154 1. Oscillation freq.3.181 0.
8 Cable energizing at L11 and measuring point at L12(circuit breaker end) simulated.7 Cable energizing at L11 and measuring point at L12(circuit breaker end) measured. 1500 Charging end L11 Measuring point L12 1000 500 vo g [v] lta e 0 500 1000 1500 100 0 100 200 300 400 time[us] 500 600 700 800 Fig.1500 Measuring point L12 Charging end L11 1000 500 vo g [v] lta e 0 500 1000 1500 100 0 100 200 300 400 time[us] 500 600 700 800 900 Fig. 20 . 3.3.
the Frequency Dependent (phase) Model is the best model.02us).02us). quicker damping is observed in reality than in the simulations. It was used with certain adjustments on data entered to account for the presence of semiconducting layers. This is either the simulation model does have certain parameters estimated incorrectly or the real system has additional resistance or conductance elements than considered in the model.9 Measured and simulated waves for L=484m (Ts=0. Although it has been found that both simulations and measurements show an increase in damping of the EM transients with increasing frequency of oscillation as should be.5 Conclusions Of the three frequency dependent cable models provided in PSCAD/EMTDC. 21 .10 Measured and simulated waves for L=242m (Ts=0. 3. The results of the experiment show that good agreements have been achieved between measured and simulated values in the frequency of oscillation and propagation delay validating the cable model proposed in this regard. 1000 0 1000 100 50 0 50 time[us] 100 150 200 Measuring point L12 Charging end L11 v o lt a g e [ v ] Charging end L11 Measuring point L12 v o lta g e [v ] 1000 0 1000 100 50 0 50 time[us] 100 150 200 a) Measured b) Simulated Fig. however. Small disagreements in these parameters are.3.3.1000 v o lta g e [v ] 0 v oltage[v ] Charging end L11 Measuring point L21 1000 500 0 500 1000 100 50 0 50 time[us] Charging end L11 Measuring point L12 1000 100 50 0 50 time[us] 100 150 200 100 150 200 a) Measured b) Simulated Fig. and core and/or sheath stranding. noticed and the authors believe they can be due to unaccounted capacitance of the measuring cables and additional stray capacitances and inductances in the real system.
1 Current chopping If the contacts of a VCB start opening close to a zero of a power frequency current. and good gap breakdown voltage recovery properties. 4.2 Phenomena causing over voltages in VCB There are four reasons why overvoltages occur in vacuum circuit breakers: current chopping. the arcing current across the contacts becomes unstable as to be extinguished. A VCB for use in MV networks as wind parks has been modeled based on [16]. 4.2. The mean value of the chopping current depends on the contact materials. VCBs have very suitable properties like capability of interruption of both power frequency and high frequency currents. dielectric breakdown strength and high frequency current quenching capability is therefore crucial to study in order to understand the occurrence of these TOVs. inductance and resistance on the load 22 . [17]. After the current is chopped and the arc is extinguished. This chapter deals with the general theory of modeling a VCB in PSCAD/EMTDC that takes into account the above characteristics. and the instant of contact separation. the level of the load current. a transient recovery voltage (TRV) appears across the VCB contacts. may lead to multiple gap breakdowns (reignitions) which may be the cause for very severe overvoltages under certain network conditions. This behavior of the breaker to interrupt a small but non zero current just before the zero of a power frequency current is called current chopping. The shape and level of the TRV depends on the actual chopping current and the resulting capacitance. Many MV component failures have been found to be caused by TOVs due to VCBs. A good model that replicates the VCB characteristics like current chopping capability. Detailed treatment of this instability can be found in [18]. however.1 Introduction At transmission level. multiple reignitions. Their ability to interrupt these high frequency currents.Chapter 4 VCB MODELLING 4. virtual current chopping and prestrikes [13]. The breaker model is then used in a test circuit to see if it actually characterizes the original VCB properties stated above. while at medium voltages VCBs are primarily used [20]. the most modern switching devices used are SF6 breakers.
the dielectric strength of the gap is.2.1 TRV waveform and chopped current. the gap may reignite. if the TRV exceeds gap breakdown strength.2 Voltage Escalation After the extinction of the arc at current zero of the power frequency current. (4. Again. This process may repeat until the gap can withstand the TRV.1) U ch = ich C Fig. therefore. a race begins between the TRV developed between the breaker contacts and the dielectric strength of the gap. they may add up to cause several high frequency arc current zeros through the breaker.side. Each subsequent restrike produces higher and higher load side over voltages due to the increased and continuous transfer of energy between the inductance and the capacitance on the load side 23 . The TRV may in this case exceed the gap voltage strength and the gap may reignite and start conducting.4. This arcing current is of very high frequency superimposed on the line frequency current. If the high frequency current is higher or equal to the power frequency current in magnitude. The VCB may be able to interrupt this current at one of the zeros if certain conditions are met. Higher chopping current produces higher TRV as do high inductance on the load side. a restrike occurs causing a substantially high frequency current. 4. If the chopping happens very close to current zero. L (4.1) shows how the chopping voltage (TRV) is related to the inductance and the chopping current. the gap separation by the time the arc has extinguished is small. accordingly small. If the TRV developed becomes greater than the dielectric strength of the breaker.
the voltage between the contacts may get higher than the dielectric withstand voltage of the gap which diminishes as the contacts get closer. 4. during closing. the current in the phases will pass through zero several times before its natural zero.4 Prestrikes While reignitions are for opening breakers. This may cause several reignitions and interruptions escalating the voltage. Fig.2.4. Compared to normal current chopping.2. Before the contacts come to a physical contact. the flow of the high frequency currents due to ignition in one of the phases can lead to injection of the current into the other phases through inductive and capacitive coupling and can lead to forced current zeroes in these phases [16]. This mechanism is called virtual current chopping when reignitions occur due to these current zeros.3 Virtual Current Chopping In three phase circuits. virtual current chopping can be much higher. the gap breaks down before the contacts are actually 24 . In this case. This process of growth of TRV is termed as voltage escalation. 4.4. as high as several hundred amperes compared to 3 or 4A [17].2 Restrike currents and voltage escalation.2. as shown in Fig. What happens is if the magnitude of the high frequency current that couples to the other phases is higher than the peak of the line frequency current. prestrikes are for closing breakers.due to the restrike and extinction respectively.
Dielectric strength 3. Multiple prestrikes might happen if the breaker is able to interrupt some of these and the process repeats. It simulates the following phenomena: 1. Consequently. q are parameters dependent on contact material. the components involved should be accurately modelled. High Frequency quenching capability 4.closed and an arcing current is established.2) according to [19]: I CH = (ω. . Î=Amplitude of the 50Hz current. 17.2) where =2 f is angular frequency of power supply.1 Current chopping The actual chopping current is nondeterministic.α .3. For modern VCBs which use Cu/Cr contact these constants are given by 25 . I . . and a prestrike is said to have occurred. However obtaining accurate VCB Model which simulates all the phenomena discussed above is difficult. Depending on the external network parameters. However earlier research established different mean chopping levels [19] for different load currents and contact material. The properties are as specified in [19]: • • • • Random nature of arcing time Current chopping ability The characteristic recovery dielectric strength between contacts when opening High frequency quenching capability at current zero. The generic model incorporates different stochastic properties inherent to the breaker operation to control the actual state of the breaker. The mean chopping current I CH is estimated by (4. and 19]. 4. multiple high slope transient step voltages are produced.3 Modeling of VCB In simulating a network for transient study. For one thing these phenomena are statistical in nature and secondly the authors believe it is beyond the scope of this paper as modelling all of those phenomena is a thesis work by itself.β ) q ^ (4. The model introduced in this paper is similar to the models on references [16. a high frequency arcing current can occur with several zero crossings. Current chopping 2.
However those parameters can be adjusted later using experimental data of the VCB.2x10 −16 q = (1 − β ) −1 =−0. as proposed by [14].7 B(KV) 3. BREAK DOWN VOLTAGE TYPE (BV TYPE) High Medium Low A(V/ S) 17 13 4.1. The constant A is the rate of rise of dielectric strength.69 26 .1 Constants for equation (4. The cold withstand voltage characteristic of the VCB is function of the contact distance. A and B are constants. the mean chopping current determined by (4.3) for three different typical dielectric strength characteristics are shown in Table 4. The values of the constants A and B in (4.3) (Reproduced from [14]).4 0. It is strongly dependent on the speed of contact separation.3 To account for the statistical nature of the chopping current. the cold gap breakdown is considered in this paper as this is the relevant breakdown strength when the preceding arc current does not exceed several hundred Amperes.3.3) where t 0 is moment of contact opening.2 Dielectric strength In modelling the breakdown strength of the contact gap. 4. The value of the dielectric strength so calculated is assumed as the mean value of a Gaussian distribution with a standard deviation of 15% as break down phenomena are statistical in nature.=6.69 0. U bd = A(t − t0 ) + B (4. For short gaps linear variation of dielectric strength with distance can be assumed [14].2) is varied by a normal distribution curve with a standard deviation of 15%. Table 4.07512 =14.
31 1 D(A/ 255 155 190 S) 4. C and D are constants.4. Interruption of the high frequency current will not occur if the rate of change at current zero is higher than the critical value of quenching capability given by (4. reignition occurs. To this end. the test circuit in Fig. and prestrikes have been considered. Table 4.3 based on [16] is used. However. di TYPE dt High Medium Low C(A/ S 2 ) 0. rate of rise of dielectric strength. Total unsuccessful operation of the breaker where it is unable to interrupt all current zeros and successful interruption at high frequency current zero are also demonstrated.2 Constants for equation (4.3. 27 . di characteristics are given by Table 4. successful interruption depends on the rate of change of this current at current zero. High frequency current superimposed on power frequency will be conducted through the breaker. high frequency current quenching capability.4) [14.4) dt where t 0 is moment of contact separation. VCBs are capable of interrupting this high frequency current at one of the zero crossings. the effects of arcing time.4 Simulation Results To characterize the various behaviors of the VCB model discussed and validate them.17.16.034 0.19] di = C (t − t 0 ) + D (4.3 High Frequency Quenching Capability When the transient recovery voltage developed across the breaker exceeds the dielectric withstand voltage.2 as The values of the constants for three typical dt given in [14].3) (Reproduced from [14]).
In the first case. all other parameters remaining the same. the longer the time the breaker gets to develop a bigger and sufficient dielectric strength to break the arc. 28 .1 Effect of Arcing Time Arcing time is defined as the time between the initiation of breaker opening to the next power frequency current zero.4. the breaker is opened at t 0 =0.4.4 (left) shows the resulting breaker voltage and current for this case. Fig.Fig.1sec.4 (right) shows breaker voltage and current for a larger arcing time of AT=5ms. For this case the beaker is opened at t 0 =0.3 Test Circuit. This gives an arcing time of AT=1ms.104s and the next current zero of the power frequency current comes at t=0.. 4ms earlier and the breaker has developed enough gap withstand voltage by the time of the first current zero and thus it is able to interrupt the arc successfully at this current zero. Fig.105s. The breaker can not break the arc successfully at the first current zero and successful interruption occurs at the second current zero when the breaker has developed enough with stand voltage at a bigger gap distance. 4. 4. 4. The higher the arcing time.
4. Now. 29 . A successful interruption occurs at the following power frequency zero. if the rate of rise is increased. 4. if the rate of rise is increased to A=17kV/s as shown in Fig. All other parameters remaining constant.5(left) which shows breaker voltage and current for a rate of rise of A=4. the breaker reaches a higher dielectric strength faster and the probability that the breaker quenches the arc at first power frequency current zero increases.5 (right).4 Unsuccessful (left) and successful (right) interruptions at first current zero for arcing times of 1ms and 5ms respectively. This has been demonstrated in Fig.4.2 Effect of rate of rise of dielectric strength The rate of rise/fall of the dielectric strength depends on how fast the breaker contacts are opening or closing respectively.Fig. the breaker has successfully interrupted the arc at the first current zero.4. 4.7kV/us in which case there is unsuccessful interruption of arc at the first current zero.
and A=17kV/us respectively.4. On the other hand.4.5 Unsuccessful (left) and successful (right) interruptions at first current zero for RRDS A=4.4. the current may not be chopped during the high frequency period. this leads to an increase in conduction period of the high frequency component (smaller number of reignitions). This is because of the high transient recovery voltage di .7. The latter case might cause failure of interruption. if the quenching capability is smaller.7 for the two cases.3 Effect of high frequency quenching capability The variation of quenching capability has been considered in this case by taking the high and medium quenching capabilities keeping all other parameters the same for both cases.6 and Fig. the higher the quenching capability.4.Fig.This is also clearly shown in across the breaker when interrupting currents with high dt the zoomed plots of Fig.4. 30 . the larger the number of reignitions will be. As it is shown in Fig. 4.4.6 and Fig.7kV/us.
4.6 Plots showing the effect of high frequency quenching capability for arcing time of di type.4.Fig.005 and medium type. 5ms and high dt Fig. dt 31 .7 Plots showing the effect of high frequency quenching capability for arcing time of di 0.
5.8 and Fig. The breaker fails to interrupt and may cause harm to itself and/or the equipments connected to.7). Fig. and interruption is accomplished in one of the next power frequency zeros. Case two and three are illustrated in Fig.6 and 4.9. 4. The breaker fails to interrupt the high frequency current.4.8 Unsuccessful breaker operation. a. b. The breaker can successfully interrupt at one of the high frequency current zeros. 4. 32 .4 Breaker unsuccessful and high frequency current successful operation When a vacuum circuit breaker reignites during opening operation.4.4.4.Transition from case two to three was achieved by increasing the breakdown voltage when the breaker is fully open from 20kV in case two to 26kV in three.4. c. voltage escalation occurs and interruption process can terminate in one of the three cases [19].4. The first case is shown in Figures (4.
33 .10.4. For the same reasons as for reignition. Fig.5 Prestrikes During closing operation of a vacuum circuit beaker. due to the high frequency current quenching.10 Pre strike during contact closing. where very steep breaker step voltages are shown to have developed. a number of prestrikes can happen in the gap during closing. the gap may break and an arc is established before real galvanic contact occurs.4. 4.Fig4. due to the diminishing dielectric strength of the gap as the contact separation decreases. The effect of pre strikes is illustrated in Fig.9 Successful interruption of high frequency current. during opening operation.4.
34 . A statistical model of the VCB based on modeling of current chopping. dielectric strength and high frequency current quenching capability was used in a generic test circuit to characterize properties of an actual VCB and expected results were found.5 Conclusions Interruption of small but non zero power frequency inductive currents and subsequent high frequency currents due to re ignitions and prestrikes by VCBs causes generation of very steep TOVs.4.
The inter. However. an accurate calculation of overvoltages which would stress winding sections is very important for economical use of the insulation material. Consequently the transformer model proposed in this work should adequately represent the transformer over a wide range of frequencies from 50/60Hz to MHz. the study of the distribution of interturn overvoltages is of essential interest [21]. The effects of such transients on transformer windings are particularly important. During the prestrike and restrike of a VCB for example highfrequency oscillations with short rise time are produced. the greatest problem is the internal resonance which occurs when the frequency of the input surge is equal to some of the resonance frequencies of the transformer. A model to simulate the high frequency behavior of a power transformer is presented in [22].1 Introduction In the previous sections it was shown that the transients caused by the switching action of the VCB have a wide range of frequencies. the transformer representation used in PSCADE/EMTDC does not take into account the high frequency behavior of the transformer like the other devices modeled in previous sections. A number of high frequency transformer models have been proposed by different authors. resonant overvoltages can cause a flashover from the windings to the core or between the turns. Internal winding models Very fast TOVs voltages can generate a voltage oscillation inside the transformer windings. and [23]. Most of the time. Besides at the design stage. Most of the time. and therefore. On the other hand.Chapter 5 Theory TRANSFORMER MODELING Background and 5. prediction of possible voltage stresses at critical points is required for efficient insulation coordination and protection. These models 35 . The internal winding model is generally applicable if calculating interturn voltages in windings of the transformer during transient conditions are of interest in the study. which may result in high frequency oscillations and overvoltage stresses on the terminals of network components. Many papers have tried to deal with this problem. Current chopping and reignition of a VCB give rise to fast electromagnetic transients in power networks. in HV and EHV networks.turn insulation is particularly vulnerable to highfrequency oscillation. These overvoltages are characterized by a very short rise time. Internal winding models and terminal models are the two broad categories.
2.. transferred over voltages and resonant over voltages)[23].assumed that the winding configuration. [23]. The voltage and current at the terminals can be expressed as 36 . 5. 5.g. Transformer terminal models are based on the simulation of the frequency and/or time domain characteristics at the terminals of the transformer. This model is based on experimental determination of the admittance matrix in a wide frequency range and subjecting the measured admittance matrix to an approximation with a rational function by the method of vector fitting. The following sections are devoted to describing the modelling techniques used to obtain the model. tank. example [22]. material constants and other construction details are known.1. Terminal models In the majority of problems related to power system transients. for example switching operation of the VCB. whereas time domain data is basically acquired from impulse voltage tests.1 Measurements Consider a transformer with n –terminals as shown in Fig5. and core. In this thesis a method for the development of a wide band frequency dependent black box model of a transformer is provided.External measurements can be performed either in the frequency or time domain. A number of studies dealing with modelling of the power system equipment based on terminal measurements have been published.2 Wide Band Modelling of Power Transformers The highfrequency behaviour of power transformers is characterized by several resonance points due to inductive and capacitive effects from the windings. geometry. transformers are considered as system components. The rational function so obtained can then be realized in to an RLC net work for time domain simulation in PSCAD/EMTDC. This behaviour should be included in any overvoltage study where the highfrequency characteristics of the transformer is of significance (e. The model proposed in this thesis belongs to terminal model as the interest is on the transient over voltages at the terminals associated due to interaction of the different components in the system. Frequency domain data consist of terminal impedance or admittance characteristics.
I (s) = Y (s)V (s) (5. . Y(s) is an n by n symmetrical matrix while I(s) and V(s) are vectors of length n.4) (5. (5..1 Nterminal Transformer Model. = . . In Y 11 Y 21 . Y n1 Y 12 Y 22 .1) can be expanded as I1 I2 .3). (5. I n = YnnV1 From (5.2) can be reduced to I 1 = Y11V1 I 2 = Y21V1 .. ... . Y nn V1 V2 . . Yn2 . where the element Yij can be found by taking the ratio I i to V j .2) It can be seen from (5. Y1 n Y2n .4) and (5. .. . .. 5.1) that applying a voltage V j to terminal j and zero voltage to the remaining terminal will produce the j th column of Y ( S ) . . Apply V1 to terminal 1 and short circuit all other terminals. .. (5.5) (5.This is demonstrated for the first column of Y ( S ) .1) where Y(s) is the admittance matrix.. v1 v2 vj i1 i2 i j Transformer vn in Fig. .5) 37 . . . . .3) (5. V(s) is terminal voltage vector and I(s) is terminal current vector all in frequency domain. Vn (5.
1 127V 2 2 0V 2 20.7) in its simplified form for the single and three phase test transformers can be re written by (5.5. i ii ij Transformer j n Fig. 38 .2 Measuring element of the j th column of Y ( Yij = Ij Ii . Yi1 = i .5.. In general the procedure used to measure the admittance is depicted in Fig. .5 KV 410 V 1 20KV 690 V 4 5 1 4 5 2 3 2 3 6 6 a) Single phase transformer b) TX1 c) TX2 Fig. hence (5. Yn1 = n V1 V1 V1 V1 (5. the high voltage winding are connected in delta and the low voltage windings are connected in wye. The terminal identification in these equations is according to Fig.5.9).2.8) and (5.. Vj Vj The measurements were performed on three test transformer. .3 Test transformers.2.5. The admittance matrix for the test transformers has been measured using the measuring technique outlined in Fig.6) The same procedure can be used to find all the columns of Y.5. This matrix has a size of 2x2 and 6x6 for the single and three phase test transformers respectively.Y11 = I I I1 I .3. 1 vj . In both the three phase transformers. Y jj = ). A single phase transformer at Chalmers University of Technology and two three phase transformers at ABB Corporate Research.. Y21 = 2 .
11) Similarly if the low voltage (L) terminals are opencircuited the voltage ratio from low to high can be obtained by setting I L = 0 in (5.10) Where I L . for I H = 0 VL (5. YLL are of size 1x1 and 3x3 for the single and three phase transformer respectively as it can be seen from (5.12) 39 . denoted by H and L. for I L = 0 VH (5.I1 I2 I1 I2 I3 I4 I5 I6 Y 11 Y 21 Y 31 Y 41 Y 51 Y 61 = Y 12 Y 22 Y 32 Y 42 5 12 Y 62 Y 11 Y 21 Y 12 Y 22 Y 13 Y 23 Y 33 Y 43 Y 53 Y 63 V1 V2 Y 14 Y 24 Y 34 Y 44 Y 54 Y 64 Y 15 Y 25 Y 35 Y 35 Y 35 Y 35 Y 16 Y 26 Y 36 Y 46 Y 56 Y 66 V1 V2 V3 V3 V3 V6 (5.10).10).8) can then be written in the form IL IH IH IL = Y LL Y HL Y HH Y LH Y LH Y HH Y HL Y LL VL VH VH VL (5. we can easily get the voltage ratio from high to low by putting I H = 0 in (5. YHL .9) = (5. VHL = VH = −Y −1 HH YHL . VLH = VL = −Y −1 LLYLH . VH are the current and voltage at the low and high voltage terminals and the Sub matrixes YHH .9) and (5. This is due to the symmetry of the windings.8) . to represent the High voltage and Low voltage terminals respectively. This partitioning of the terminal or winding simplifies the problem. (5. YLH .7) and (5. If the high voltage (H) terminals are opencircuited.8) Consider the situation that the transformer terminals are divided into two groups.9) and (5.7) and (5.7) = (5. It is also interesting to mention the symmetry of the sub matrix. VL and I H .
2 VH . 6 Fig.5 V H . VH .1 VH .51 = V H .2 2 H VH. 4 5.4 and Fig.5.5 6 VL . VL. It can be employed both for single as well as for several frequency responses of a system.5. V HL 24 = . VLH are matrices of size 1 by 1and 3 by 3 for a single phase and three phase transformer respectively. V HL 34 = ). VLH .5. 4 .5.3 VHL ( V HL14 = .It can be easily noted from (5. V .4 V.1 VL. V LH .6 VH . 4 VL.1 1 H L 4 2 3 VL.12) VHL .4 5V L. 41 = V L. 4 V L.1 ). Some examples where this method can be used are: • Network equivalent for portion of a large network 40 .5 which demonstrates this clearly.4 Measuringelement of the first column of VLH ( V LH .1 1 3 H L4 5 VL.11) and (5.2 Rational Approximation of Frequency Responses by Vector Fitting This section briefly reviews the method of rational approximation by vector fitting.1 .61 = VL . The voltage ratios from Low to High and from High to Low can be measured using the set up given in Fig.5 Measuring element of the first column of V H . The method is used for the fitting of measured or calculated frequency domain responses into a rational function that approximates the response very well.3 H 6 Fig.2. The method was developed by Björn Gustavsen et al based on reformulation of the Sanathanan–Koerner iteration and has been found to be very suitable for fitting network equivalents and transformer responses.
N is the order of approximation. The Vector Fitting Algorithm Generally speaking. A more efficient method is therefore required. Since one of the unknowns an in (5.14) is in the denominator. This requires computation of the unknowns. The idea of vector fitting is to find a rational function approximation which is sum of partial fractions from the given frequency response data.. d and e. consider the rational function approximation f ( s) = cn + d + s. however.14) where f(s) is a frequency dependent transfer function of a single response or a matrix of several responses. an approximation of a given order can be obtained by fitting a given set of frequency dependent data to a ratio of polynomials as shown in below [27]. An approximation of f(s) in a given frequency range is therefore achieved this way. The intention is to approximate the response f(s) by the terms on the right hand side of the equation. produces a badly scaled and illconditioned problem as the columns of A are multiplied with different powers of s. f(s) ≈ a 0 + a1 s + a 2 s 2 + a3 s 3 + . For smooth functions real poles are sufficient.e n =1 s − a n N (5. To demonstrate the method of vector fitting. It is recommended to choose complex conjugate pole pairs for functions with distinct resonance peaks like transformer responses as starting poles.. The starting poles are so chosen to cover the frequency range of interest and should be linearly or logarithmically distributed over the whole range. Of the many fitting algorithms developed to date. Vector fitting has been found to be very efficient to model transformer responses and network equivalents.13) To write this non linear equation as a linear problem of the type Ax=b with the coefficients as unknowns. This.. an. cn and an are the residues and poles respectively which can be real or conjugate pairs and d and e are real.. the problem should be linearized before it is solved. the problem is a non linear one.• Highfrequency transformer model based on measured frequency responses • Transfer of voltages from transformer terminals to internal points in winding • Transmission line modeling The method is later applied to fit the admittance matrix of power transformers in coming sections. The algorithm of vector fitting tackles the 41 .cn . + a N s N b 0 + b1 s + b2 s 2 + b3 s 3 + . Therefore. This is done by replacing a set of starting poles with an improved set of poles via an iterative pole relocation method based on least squares approximation of linear problems [27]. both sides should be multiplied with the denominator. + bN s N (5. This imposes a limitation on the applicability of the approximation to very lower order only.
problem in two steps. In the first step, the poles are identified and in the second step residues are identified.
1. Step 1
Here, a set of starting poles a n are chosen for f(s) and (5.14) is multiplied by an unknown function (s) which is also approximated with a rational function with the same starting poles. ~ cn (s) ≈ +1 n =1 s − a n
N
(5.15)
An augmented problem can thus be written cn + d + s.e f(s) σ (s) n =1 s  a n ≈ N ~ σ (s) cn +1 n =1 s − a n
N
(5.16)
~ To have a linear problem in the unknowns cn, cn , d and e, the second row of (5.16) is multiplied by f(s) giving
cn + d + s.e ≈ n =1 s − a n
N
~ cn + 1 f(s) n =1 s − a n
N
(5.17) (5.18)
( f)fit(s) =
fit(s)
f(s)
is a linear problem of the type Ax=b in the frequency range of interest with the above stated unknowns in the solution vector x. For a given frequency S k
Ak x = bk
where
(5.19)
1
2
Ak= [
sk
1 1 ... − a1 sk − a
sk
− f ( s k ) − f ( sk ) … sk − a1 s k − a1
]
(5.20)
~ ~ x= [ c1…. cN d e c1 … cn bk=f(sk) From (5.18), we have
]T
(5.21) (5.22)
42
f(s) =
(σf ) fit ( s )
σ fit ( s )
(5.23)
Since both numerator and denominator of (5.23) have the same poles( s  an ), the ratio in (5.23) can be rewritten as single fraction instead of sum of partial fractions
f(s) =e
∏s − z
n =1 N n =1
N +1
n
z ∏ s − ~n
(5.24)
It is clearly seen from (5.24) that the zeros of fit(s) are equal to the poles of f(s). Thus by calculating the zeroes from the linear equation of (5.15), a good set of poles for the fitted f(s) is obtained.
2. Step 2
Now the identification of the residues of the function f(s) remains in order to have a complete fit of its rational approximation. If the zeroes of fit(s) calculated in step 1 are used in the original problem in (5.14) as new poles, we thus have a linear problem where the residues can be easily solved giving the unknowns cn, d and e. For more complete and deeper treatment of how the vector fitting algorithm is formulated, the reader is referred to [27].
Passivity enforcement
The objective of the rational approximation of the frequency dependent response f(s) in (5.14) using the vector fitting algorithm is to have an equivalent model in the form of a realized network of passive elements when f(s) is an admittance or an impedance function, Y(s) or Z(s). The equivalent model thus obtained is linear, hence the term Passive Linear Time invariant component. If this component is to be included in a transient time domain simulation, it should behave a passive nature in that it should only absorb active power for any set of applied voltages and not produce. Time domain simulations involving a fitted admittance or impedance may sometimes lead to an unstable time domain simulation due to non passivity. There are also cases where the model could be stable and still non passive in which case unstable simulations arise when the model is used in an interconnection with other components. It is, therefore, important to check that the equivalent network contains no passivity violations over the whole frequency range and outside the frequency range as well. If violations exist, they should be removed. This procedure of removing passivity violations is termed Passivity Enforcement.
43
A requirement for the fitted admittance Y(s) to be passive is for the real part of the admittance to be Positive Definite, which means all eigenvalues of G should be real positive. A general procedure to enforce passivity to the network is to calculate a correction for the approximated rational function so that it becomes Positive Definite [28]. This corrective perturbation should be as minimal as possible since it causes a fitting error of the actual data. One way to achieve minimum correction is to have very accurate measurement of the response. A formal formulation of the PD criterion is given below [28].
Passivity Criterion
Consider a component defined by an admittance matrix Y(s) in the frequency domain
I=Y v
(5.25)
where I and v are current and voltage complex vectors. The absorbed power is therefore given as
P=Re {v*Y v} =Re {v*(G+jB)v}= Re {v*Gv}
where * means transpose of conjugate of v.
(5.26)
It follows from (5.26) that the real power P will always be positive if all eigenvalues of G are positive, entailing the PD criterion. For further and detailed reading on passivity enforcement the reader is referred to [29].
5.2.3 Time Domain Implementation
The basic idea behind the new transformer model is to produce an equivalent network whose admittance matrix matches the admittance matrix of the original transformer over the frequency range of interest. The next question is how we get this equivalent network. This equivalent network can be obtained by realization of the fitted admittance matrix Yfit(s). The procedure to obtain the network equivalent is given in the APPENDIX and summarized here. Consider the admittance matrix Y of a transformer obtained using the measurement procedure in section 5.2.1 and fitted with a rational function using the VF method given in section 5.2.2 creating a matrix Yfit(s) in which the elements are
rmij (5.27) + dij + s.eij m =1 s − am The network has branches between all nodes and ground and between all nodes. Branches between node and ground are given by: Yfit ( s )ij =
N
44
33) (5.32) while the complex conjugate pair is given by RLCG network.6. where 1 . + d + s. where Lc = Cc = 1 . Rc = 2 Lc (Lc (r ' a'+ r" a") − a') 2r ' 1 2 2 (5.30) The synthesis of (5.yii = n j =1 Yfit ( s )ij (5.. Consider now a single branch y(s) of the network (node to ground or between two nodes) fitted with real and complex conjugate poles y (s) = r1 r r1 '− jr1" r1 '+ jr1" r2 '− jr2 " r2 '+ jr2 " + 2 + . Gc = −2 LcCc (r ' a'+ r" a") 45 .29) where n is the size of the matrix which is 2 for single and 6 for the three phase transformers.34) Lc (a' + a' ' +2 Rc (r ' a'+ r ' ' a' ' )) . C0 = e d The real pole is represented by the RL circuit as R0 = (5. Rr = r r (5.28) while the branches between node i and node j are yij = −Yfit ( s )ij (5..31) Lr = −a 1 ..5.e s − a1 s − a2 s − (a1 '− ja1" ) s − (a1 '+ ja1" ) s − (a2 '− ja2 " ) s − (a2 + ja2 " ) (5.. + + + + + .30) is an RLC network of parallel branches as shown in Fig.
Constant term
s.e term
Real Pole
Complex conjugate pole
Rr
Rc
R0
C0
...
Cc
Lc Gc
...
Lr
Fig.5.6 Basic structure of the synthesized electrical network for a single element branch.
46
Chapter 6 TRANSFORMER MODELINGExperiments
and
Results
6.1 Introduction
In this chapter the measurement setup designed and instrumentations used to measure the admittance matrix and voltage ratio over a wider frequency range is describes in detail .The measurement and fitting result along with their analysis is discussed and finally the model is validated by comparing the measured results with the corresponding calculated and PSCAD/EMTDC simulated results.
6.2 Measurement Setup
The admittance matrix and voltage ratios for the transformers were measured in frequency domain using a network analyzer.The measurement set up as configured to measure one element of the admittance matrix and voltage ratio is outlined in Fig.5.1 and Fig.5.2 respectively. The set up consists of the test transformer (a single phase 127/220V and two three phase transformer), the network analyzer to measure the admittance and the voltage ratio over a wide frequency range, current probe for the current measurement and shielded cables which connect the output port of the network analyzer to the transformer terminal for excitation and from the transformer terminal to the input ports of the network analyzer for the purpose of voltage or/and current measurements.
Fig.6.1 Experimental setup to measure element Y31 of the admittance matrix.
47
Fig.6.2 Experimental setup to measure V LH , 61 =
VL,6 V H ,1
element of the admittance matrix.
Measurement Objects
As it was mentioned earlier the measurements and hence modeling has been done on three transformers but the measurements done on the single phase transformer was an important step to the modeling of two three phase transformers which represent the wind mill (TX1) and the load side transformers of three phase cable systems that has been built at ABB corporate research to emulate a section of an off shore collection wind park model in order to study fast transients that can occur in an off shore wind park.
Windmill Transformer (TX1)
Transformer TX1 (Fig.6.3) in the experimental set up is the representation for one of the windmill transformers in the off shore wind park. It is a 1.25MVA, 410V/20.5kV step up and high voltage side in . The transformer with low voltage side connected in transformer data as it is shown in the name plate data is given in Table 6.1.
Load side Transformer (TX2)
The load side transformer TX2 (Fig.6.4) is a 1.0MVA, 690 V/20kV distribution transformer with connected high voltage winding and connected low voltage winding. The specification of this transformer is given in Table 6.2.
48
Table 6.4 Load side Transformer (TX2). This network analyzer has a bandwidth of 10Hz to 300MHz.5 % V 690 V U 125AC50/AC8 kV Zk 5. For both admittance and voltage ratio measurements one output channel which is for the excitation voltage and two input channel. NO. one output and two input ports. TA and TB) channels. 5210667 CTMU 24 HM 1000 1000 kVA Dyn11 20000 ± 2´2.6. The MS4630B network analyzer was used to measure the admittance and voltage ratio of the single phase transformer at Chalmers.4 % Pk 11971 W Po 1455 W 1999 IEC 76 50 Hz 35. one for voltage and one for current measurement for the case of admittance or both for voltage measurement were required as the admittance and voltage 49 . two output (A and B) and three input (R.7 A ONAN 2670 kg 505 kg Network Analyzer Two different network analyzers have been used for measurement.3 Windmill Transformer (TX1).5 % V 410 V U 125AC50/AC8 kV Zk 5.1 Transformer TX1 data.1 % Pk 10260 W Po 1223 W 1997 IEC 76 50 Hz 28.2 Transformer TX2 data.Fig. This is because the measurements for the single phase transformer were done at Chalmers while for the two three phase transformers they were done at ABB Corporate Research.2 A 1760 A ONAN 2885 kg 605 kg Fig. Table 6.6. 5214242 CTMU 24 HA 1250 (ABB) 1250 kVA Dyn11 20500 ± 2´2. NO. The Bode 100 network analyzer which was used for the two threephase transformers measurements has a bandwidth of 10 to 40MHz.87 A 836.
100.1V/1A for two threephase transformer measurements has been used. respectively.5 MS4630B Network Analyzer. TA. 10. 4. Table 6. 300 2.6. 100 Hz. Fig.7) with sensitivity of 1V/1A for the singlephase transformer and 0.6.7 current probes of model Pearson. 20 kHz Hz.6. 500 Hz. current probes of model pearson (Fig.3.3 kHz 11. 50 .ratios were measured as the ratio of current to voltage or two voltages. 10. 30 Hz.6. The specifications of the network analyzers are given in Table 6. 1. 5.01 to 1Vrms into 50 load –27 dBm to 13 dBm R. Fig.3 Technical data of the bode 100 network analyzer. 21. 10 to 16501 (freely 1001 selectable) Current Probe As the admittance elements were measured as the ratio of current to voltage. TB CH 1 . 251. Fig. 501. 30. CH 2 50 or 1 M 50 or 1 M BNC BNC 3. CHARACTERISTIC MS4630B Frequency range 10Hz to 300MHz Output impedance 50 Out put Connector BNC Wave form Sinusoidal signal Output voltage Output A 0 to +21 dBm Output B –6 to +15 dBm Input Channel Input impedance Connectors Receiver bandwidth SweepingNumberof measurement points: BODE 100 10 Hz to 40 MHz 50 BNC Sinusoidal signal 0.6 Bode 100 Network Analyzer. 10 Hz. 51. 3. 101. 1 kHz.
YLH OffDiag. The data in Fig. This problem has been tackled by exploiting the symmetry of the transformers and thus of the admittance matrix.8 a) and 6.3 Measurement Results Fig. elements of YHH 5 5 6 Diagonal elements of YHH OffDiag.8 and 6.6. elements of YHL and YLH 10 10 6 Diag.9 a) and 6. The reproducibility or repeatability of the measurements is. Reproducibility of the data. 6. elements of YLL 0 0 OffDiag.6.8 b) obtained for TX2 are almost the same for the two sets of measurements. Therefore. Due to this symmetry. several elements of the admittance matrix are identical or closely so. two with 801 and 401 sampling points for TX2 and two with 1601 and 801 for TX1. elements of YHH Diag elements of YHH 1 1 2 2 Magnitude[S] Admittance[S] 3 3 4 4 OffDiag.1) . The same is also observed in Fig. These raw data contain inaccuracies due to noise and other factors as is reflected clearly in the figures. 1 1 10 10 10 10 10 10 10 10 10 Diagonal elements of YLL OffDiag elements of YLL Diagonal elements of YHL. elements of YHL and YLH 1 2 3 4 5 6 7 7 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 10 10 10 10 Frequency[Hz] 10 10 a) 401 frequency points b) 801 frequency points Fig. indisputably validated in the 4 sets of measurements. with 401 and 801 sampling points respectively.6. however. 6. bad measurements or noisy measured admittance elements have been replaced with better measured data of another symmetric element.elements of YHL.9 show the magnitudes of the 36 elements of the admittance matrix Y for TX2 and TX1 obtained by direct measurement in four sets of frequency sweeps. elements of YLL OffDiag.YLH 10 10 10 10 10 10 10 Diag. with 801 and 1601 sampling points respectively.8 Magnitude plot of admittance matrix (raw data) for TX2. does not always guarantee an accurate data set especially when the cause for the inaccuracy persists. however. The credibility of the data is further enhanced by the large difference in the element sizes which is expected due to the high voltage ratio of the transformers ( ≈ 50).9b) for TX1. A summary of elements expected to be identical or closely so is given below where Y has been partitioned into four sub matrices as shown in Y= YHH YLH Y HL YLL 51 (6.
elements of YHH 10 4 OffDiag. elements of YHL and YLH 10 5 10 5 Diag. elements of YHH 10 1 OffDiag. Neater and more accurate data is accordingly found with this modification.6. 6.9 Measured admittance elements (Raw data) for TX1. elements of YLL Diag elements of YHH OffDiag.10 Measured admittance elements modified using symmetry of the sub matrices for TX2.10 and Fig. elements of YHL and YLH 10 6 10 1 10 2 10 3 10 Frequency[Hz] 4 10 5 10 6 10 6 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 a) 401 frequency points b) 801 frequency points Fig. elements of YLL Diagonal elements of YHL.6.YLH 1 Diag.11. The new data set is given in Fig.i) ii) iii) iv) v) vi) 10 10 1 Diagonal elements of YHH Diagonal elements of YLL Diagonal elements of YHL and YLH Offdiagonal elements of YHH Offdiagonal elements of YLL Offdiagonal elements of YHL and YLH 10 10 1 0 0 10 Magnitude[S] 1 10 Admittance[S] 1 1 10 2 10 10 2 10 3 3 10 4 10 4 10 5 10 2 3 4 5 6 5 10 6 10 10 10 10 Frequency[Hz] 10 10 10 6 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 a) 801 frequency points b) 1601 frequency points Fig.YLH 10 Admittance[S] 1 Magnitude[S ] 10 2 10 10 2 10 3 3 Diagonal elements of YHH 10 4 OffDiag. 10 1 10 Diagonal elements of YLL OffDiag. elements of YHL. 6. elements of YLL 10 0 10 0 OffDiag. 52 .
The method applied assumes the transformer as a linear timeinvariant and frequencydependent black box with a certain number of terminals.6. Matrix curve fit is then applied to approximate the elements of the admittance matrix by a rational approximation. The frequencydependent responses are obtained via measurements as discrete functions of frequency. 53 . 6. All the files and Matlab codes on curve fitting can be downloaded from [26]. The rational function f(s) given in (5. The method used is based on a series of papers on curve fitting and open Matlab source codes developed by Björn Gustavsen from SINTEF. A Matrix curve fit stacks all elements of the matrices into a single column vector h(s) before the elements are fitted.10 1 10 1 10 0 10 0 10 Magnitude[S] 1 10 Admittance 1 10 2 10 2 10 3 10 3 10 4 10 4 10 5 10 1 5 10 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 a) 801 frequency points b) 1601 frequency points Fig. The frequency response measured is a set of currentvoltage ratios over a wide band of frequencies.14) in this case becomes an nxn admittance matrix as all transformers are multi terminal devices with n terminals. This gives a threedimensional admittance matrices of size nxnxNs where n is the number of terminals(nodes) and Ns is the number of measurement points (frequency points).11 Measured admittance elements modified using symmetry of the sub matrices for TX1. Matrix curve fitting is used as the transformer is a multiterminal device [25].4 Rational Approximation of Admittance Matrices of The Transformers This section discusses the results of rational approximation of the measured admittance matrices of the power transformers by curve fitting of the frequency response measured at the terminals.
10 10 10 10 10 10 10 10 10 1 300 200 100 0 Phase angle [deg] Original Approximation 1 0 1 Magnitude [p.12 shows the rational approximation of both magnitude and phase of the 36 elements of TX2 with an order of approximation of N=20. Fig.013563. The original admittance elements are drawn together with their rational approximations for comparison to see how good the fitting is for different orders of approximations. After pass.1 0.4 0. 6. Passivity was also enforced for all elements to ensure stable time simulation.35 0. 54 .05 3 10 4 5 6 Before pass.] 2 100 200 300 400 500 3 4 5 6 600 3 4 5 6 7 Original Approximation 10 2 10 10 2 10 10 Frequency [Hz] 10 10 700 1 10 10 10 Frequency [Hz] 3 4 10 5 10 6 a) Magnitude b) Phase Fig. The eigenvalues of Re{Y} before and after passivity enforcement for this case is given in Fig. The total rmserror of approximation in this case is equal to 0. N=20 for 1kHz to 1MHz. 10 10 Frequency [Hz] 10 Fig. 3 iterations and passivity over 801 frequency points in the range from 1kHz and 1 MHz. 6.u. Although the fit in the low frequency segment up to around 10 kHz seems ok.13 Eigenvalues of Re {Y} for TX2 before and after passivity. enf. 0.25 0.13.3 E e v lu s o G ig n a e f 0.05 0 0.12 Rational approximation of Y for TX2 with N=20.6. we have a very poor approximation at high frequencies. enf.15 0.45 0. Eigen values of Re{Y} have also been drawn before and after passivity is enforced on all elements for different number and range of frequencies to see the effect of passivity on the error of approximation.6.The curve fitting algorithm was used to approximate all elements of the admittance matrices with a common set of poles.2 0.
In principle. The Eigen values of Re{Y} before and after passivity enforcement is given in Fig. N=100 for 1kHz to 1MHz. The rmserror in this case is 0.4 Eigenvalues of G 0. 0.6.15 Eigenvalues of Re {Y} for TX2 before and after passivity. one often wants to find a low order approximation as higher order gives larger element size of the resulting network. The resulting approximation is shown in Fig.6. a perfect fit can be obtained by increasing the order indefinitely high.0113.2 0. 55 .15. however. In our case a better fit has been achieved by increasing the order of approximation to N=100 with the same passivity enforcement as the previous one.5 Before pass. In practical applications. 0.] 1 0 Phase angle [deg] 100 200 300 400 500 Original Approximation 1 10 2 10 3 10 4 10 5 600 3 4 5 6 10 6 Original Approximation 10 2 10 10 2 10 10 Frequency [Hz] 10 10 700 1 10 10 10 Frequency [Hz] 3 4 10 5 10 6 a) Magnitude b) Phase Fig. higher simulation time and requires very big computational resource. 1 10 300 200 100 10 0 10 Magnitude [p.14 where it is seen that the approximation has been improved throughout the whole frequency range.6. enf.14 Rational approximation of Y for TX2 with N=100. enf. After pass. 6.3 0.1 0 0.u.6 0.1 3 10 10 4 10 Frequency [Hz] 5 10 6 Fig.
17 and Fig. Fig. very accurate measurements should be made so that there are no passivity violations inside the frequency range of interest or outside.14 but with an increased range of frequencies enforced.1Hz to 10MHz. 6. 56 .1Hz to 10MHz. The increase in fitting error is because of perturbation of more frequency points while enforcing passivity in the later case.015133. N=100 for 0.023634. Fig. 7 6 5 Eigenvalues of G 4 3 2 1 0 1 2 10 Before pass.0113 to 0.18 give the rational approximation and eigenvalues of Re{Y(s)} respectively of TX1 for N=60 and passivity enforcement over 0. The error of approximation is increased from 0. The rmserror is 0.6. A reduced order of approximation was used due to limitation of computational memory since the number of measurement points to be approximated in this case was 1601 which is double that for TX2. enf.6. enf.1kHz to 10MHz on 1601 points. A compromise is therefore made between stability and fitting error. 6. therefore. Y(s) was approximated with order of approximation N=100 as in Fig.To show the effect of passivity on the fitting error. 10 0 10 10 Frequency [Hz] 2 4 10 6 10 8 Fig.16 gives the Eigen values of Y(s) before and after passivity is enforced over the frequency range 0.16 Eigenvalues of Re {Y} for TX2 before and after passivity. After pass. To decrease the fitting error due to passivity.6.
10 0 10 10 Frequency [Hz] 2 4 10 6 10 8 Fig. Finally.u.17 Rational approximation of Y for TX1 with N=60 7 6 5 Eigenvalues of G 4 3 2 1 0 1 2 10 Before pass.6.10 1 300 200 100 Phase angle [deg] Original Approximation 1 10 0 10 Magnitude [p. N=60 for 0.19.] 1 0 100 200 300 10 2 10 3 10 4 400 500 1 10 10 5 Original Approximation 10 2 10 10 2 10 10 Frequency [Hz] 3 4 10 5 10 6 10 10 Frequency [Hz] 3 4 10 5 10 6 a) Magnitude b) Phase Fig. Deviations between original data and approximations are also shown in the same plots. 6. enf.20. enf.18 Eigenvalues of Re {Y} for TX1 before and after passivity. 6.6.6.22 respectively. Far better fits are achieved for this transformer since the number of elements to be fitted with a common set of poles in this case is only 4. Fig.6.21 and Fig. approximations and eigenvalues with orders 20 and 100 are plotted for the small transformer at Chalmers as shown in Fig.1Hz to 10MHz. 57 . After pass. Fig.
5 Eigenvalues of G 0.u.10 0 10 M agnitude [p.1 2 10 0 2 4 6 8 Before pass. N=20 for 0. enf. enf.2 0.6 0.] 1 Original Approximation Deviation Phase angle [deg] 200 0 200 400 Y11 and Y22 600 800 1000 1200 1400 1600 Original Approximation 10 2 10 2 Y12 and Y21 Y22 Y11 Y12 and Y21 10 3 10 4 10 5 10 1 10 2 10 3 a) Magnitude 10 10 Frequency [Hz] 4 5 10 6 10 7 1800 1 10 10 3 10 10 Frequency [Hz] 4 5 10 6 10 7 b) Phase Fig. 10 10 10 Frequency [Hz] 10 10 Fig. After pass. 0.1Hz to 10MHz. 58 .3 0.6.20 Eigenvalues of Re {Y} for small transformer before and after passivity.19 Rational approximation of Y for small transformer with N=20.6.1 0 0.4 0.
6.6 0.u.22 Eigenvalues of Re {Y} for small transformer before and after passivity.1 0 0. After pass.4 0. enf.1 2 10 0 2 4 6 8 Before pass.5 0. 59 .1Hz to 10MHz. enf. N=100 for 0.10 0 500 Y11 Y22 Y12 and Y21 10 1 0 P has e angle [deg] Magnitude [p. 0.6.3 0.] 10 2 500 Y11 and Y22 Y12 and Y21 10 3 10 4 1000 10 5 10 6 Original Approximation Deviation 1 1500 Original Approximation 10 10 Frequency [Hz] 4 5 10 10 2 10 3 10 6 10 7 2000 1 10 10 2 10 3 10 10 Frequency [Hz] 4 5 10 6 10 7 a) Magnitude b) Phase Fig.21 Rational approximation of Y for small transformer with N=100.2 0. Eigenvalues of G 10 10 10 Frequency [Hz] 10 10 Fig.
Fig. The accuracy of this model greatly depends on the accuracy of the measured admittance and hence it is very important to the check if these measurements are accurate enough to give the accuracy requirements to the model to be developed using this measurements. the small deviation between the calculated and simulated are the result of fitting errors otherwise the calculated and simulated curves fall on the same curve provided that we can get an ideal perfect fitting.29.6.6. The voltage ratio calculated was obtained from the measured admittance using (5. With the computer we had for example it was impossible to exceed the order from 130 or 80 for number of samples 801 and 1601 respectively. Similarly very good agreement has been obtained between the calculated and simulated ratio as it can be seen in the same figures. The reason for this will be explained later in section (6. 60 .6.11) and (5.5 Model Validation In this section the validity of the transformer models has been checked by comparing the voltage ratio measured directly on the transformer terminals with the calculated and simulated by the model in PSCAD/EMTDC.12). higher computer memory and higher simulation time.23 to Fig.6.26 compares the measured elements of voltage ratio from low to high ( VLH ) with the corresponding elements calculated and simulated. one often wants to find low order approximation as higher order gives larger element size of the resulting network.6. One way to validate the transformer model is therefore to validate the measured admittance matrix without further doing timedomain implementation in PSCAD/EMTDC as this model was formulated in terms of its admittance matrix.27 to Fig. Transformer TX2 The PSCAD/EMTDC simulated result for TX2 was obtained by the model developed for an admittance matrix Y measured with 801 number of frequency samples and fitted with an approximating rational function of order 100. A similar observation is made for the voltage ratio elements from high to low ( VHL ) when comparing measured and calculated quantities. In practical applications. see Fig.6). It is important to note here that as the model developed is not directly from the measured admittance but from rational approximation of the measured admittance. The calculated ratio is seen to closely match the measured ratio except at lower frequencies.
6 V H .10 1 calculated measured simulated 10 1 V olta ge R atio Voltage Ratio 10 0 10 0 calculated measured simulated 10 1 10 1 10 2 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 10 2 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 Fig.23 Voltage ratio low to high V LH .6 V H .6. 63 = VL.52 = V L . Fig. 41 = VL. 10 1 calculated measured simulated Voltage Ratio 10 1 10 1 Voltage ratio 10 0 calculated measured simulated 10 0 10 1 10 2 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 10 2 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 Fig.5 VH .24 Voltage ratio low to high V LH .1 .1 .6.6.6.26 Voltage ratio low to high V LH . 61 . 61 = V L . Fig.25Voltage ratio low to high V LH .2 .4 V H .3 .
6 . 2 V L . Fig.Compared to the results obtained for TX2.This was observed during the measurements of the admittance for TX1 where more elements have to be replaced using symmetry of the 62 . Reasonably good agreement between calculated and simulated results are obtained for the voltage ratios VLH as it is shown in the plots of Fig.27 Voltage ratio high to low V HL.6.1 VL.6.3 V L .30 to Fig.10 10 Voltage Ratio 10 10 10 10 3 10 3 2 calculated measured Voltage Ratio 10 2 calculated measured 1 10 1 0 10 0 1 10 1 2 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 10 2 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 Fig.6.36 = V H .29 Voltage ratio high to low V HL . Transformer TX1 Similar plots have been made for transformer TX1 except the admittance matrix and voltage ratios were measured for 1601 number of samples and the simulated results are for the model obtained with an order of approximation of 60 for the measured admittance.it is seen that a relatively large deviation between the measured and calculated (simulated) is observed for TX1. This is due to the higher error on the measured admittance for TX1 than TX2.14 = 4 V H . 10 10 Voltage Ratio 2 calculated measured 10 0 10 2 10 4 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 Fig.32.4 .5 .28 Voltage ratio high to low V HL . 25 = VH .6.6.
matrix.6. 10 0 calculated measured simulated 10 0 calculated measured simulated Voltage Ratio Voltage Ratio 10 1 10 1 10 2 10 2 10 3 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 10 3 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 Fig.521 = V L . Fig.6.30 Voltage ratio low to high V LH .3 .8 and Fig.32 Voltage ratio low to high V LH .1 . 63 = VL. 4 V H .6. This can also be observed by comparing the admittance matrix plots for TX2 and TX1 shown in Fig. 63 .5 VH . 41 = VL.6.31 Voltage ratio low to high V LH . 10 0 Voltage Ratio 10 1 calculated measured time simulation 10 2 10 3 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 Fig.9. 2 .6.6 V H .
64 . 10 10 V oltage R atio 10 10 10 3 2 calculated measured 1 0 1 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 Fig. 2 V L . Fig.5 .6.36 = V H .34 Voltage ratio high to low V HL .35 Voltage ratio high to low V HL .37 compare the measured and calculated (using (5. Single phase Transformer Fig.10 10 V oltage R a tio 10 10 10 10 3 10 calculated measured 4 2 calculated measured V oltage R atio 10 2 1 0 10 0 1 2 10 1 2 10 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 10 1 10 2 10 10 Frequency[Hz] 3 4 10 5 10 6 Fig.6. 25 = VH .11) and (5.12)) values of voltage ratios.36 and Fig.4 .6. The calculated ratios are seen to closely match the measured ratios with some discrepancy.33 Voltage ratio high to low V HL.14 = V H .6 .6.6.3 V L . According to [23] the condition with one or more terminals of the transformer open is the worst case scenario because a 1% measurement error in the admittance matrix could cause as high as 100% error in the calculated voltage ratio.1 VL. hence the errors obtained are reasonable.
39.27 to Fig.2 V L .6. a transformer model may. 65 . This is especially observed for TX2 in plots of Fig. 21 = VH . YHL . YLH ) becomes very small because the current is too small to be accurately determined as the only connection to ground is via capacitors. 2 .6 Accuracy of the Model The models developed for TX2 and TX1 are not accurate at lower frequencies.5. YHL . YLH ) and hence the zerosequence component of diagonal and off diagonal blocks associated with ungrounded windings( YHH .5.5.5. Fig.This error is due to the ungrounded windings of the delta connected high voltage winding for both transformers. This can be seen from the comparison between measured and calculated results of the preceding sections. produce accurate results in one situation and inaccurate results in a different situation.1 .6. thereby causing a nearsingular condition for the blocks in the admittance matrix Y associated to ungrounded windings( YHH . The inaccurate result at low frequencies for transformers with ungrounded windings is due to the inaccuracy in the zero sequence component of the admittance matrix at low frequency corresponding to the ungrounded winding as the capacitive coupling to ground of an ungrounded winding approaches zero with decreasing frequency. Thus.36 Voltage ratio high to low V LH .27 to Fig.37 Voltage ratio high to low V HL .1 VH .12 = V L . see Fig.10 1 10 calculated measured 0 calculated measured 10 1 10 V oltage Ratio 0 10 1 V oltage Ratio 10 2 10 2 10 3 10 1 10 2 10 3 10 Frequency[Hz] 4 10 5 10 6 10 7 10 3 10 1 10 2 10 3 10 10 Frequency[Hz] 4 5 10 6 10 7 Fig. The sensitivity of the model to errors in measurements and the rational approximation was also found to be very high when one or more terminals were opencircuited on one or both sides of the transformer. 6. in general.33.
From the admittance matrix measurements taken at two different numbers of samples. Network realization of the approximating rational function to its equivalent electrical network generated given in a Matlab generated file was used later to implement the model in a PSCAD script for time domain simulation. Comparison of results between measured.6. calculated and simulated results show good agreement. the measurement setup was shown to give reproducible results.7 Conclusion This chapter demonstrated how a linear wide band frequency dependent model of a power transformer has been developed. The model was however found to be sensitive to measurement errors especially for a transformer with ungrounded windings. 66 . The errors were magnified at lower frequencies and during open circuit conditions. Vector fitting was adapted to subject the measured admittance matrix to rational function approximation. A dedicated measurement setup to measure the admittance matrix over a wider frequency range has been designed.
Chapter 7 CONCLUSIONS AND FUTURE WORK 7. and quenching capability of a VCB requires that a stochastic model be developed for it. Although it has been found that both simulations and measurements show an increase in damping of the EM transients with increasing frequency of oscillation as it should be. The following conclusions are inferred: o The statistical nature of the arcing time. VCBs and transformers as in a wind farm. It has been studied that the following conclusions can be drawn: o With certain adjustments that take into account the semiconducting layers and core and sheath stranding. its ability to interrupt high frequency current zeros causes escalation of steep front 67 . EM propagation. Small disagreements in these parameters are. The first part of this report dealt with cable modeling and experimental validations. reflections. o Although the VCB has good current interruption characteristics. quicker damping is observed in reality than in the simulations. very severe high frequency overvoltages may occur due to the switching action of the VCB. dielectric strength. the Frequency Dependent (Phase) Model provided by PSCAD/EMTDC gives a fairly very good model of a real cable. chopping current.1 Conclusions In power systems that comprise an interconnection of cables. VCB modeling and characterization of phenomena as current chopping. noticed and the authors believe they can be due to unaccounted capacitance of the measuring cables and additional stray capacitances and inductances in the real system. dielectric with stand and high frequency quenching capability makes the second part of this thesis. This model replicates high frequency behaviors like skin effect. Characterization of these overvoltages and the ultimate protection from them requires accurate modeling of the above components. however. o The results of the experiment show that good agreements have been achieved between measured and simulated values in the frequency of oscillation and propagation delay.
o The transformer model is a linear passive network equivalent of RLC elements. and high frequency quenching capability. these effects are small and can be neglected. simulated voltage ratios and those directly measured as function of frequency. 68 . Therefore. it can only be used in transient studies involving overvoltages at the terminals. for high frequency transients. This has been demonstrated by using a stochastic VCB model in a generic test circuit. These characteristics are very pronounced for low frequency overvoltages. The following points have been noted: o Since the model is a linear black box equivalent based on measurements on the terminals as opposed to a detailed model of the internal parts of a transformer. rate of rise of dielectric strength. there are discrepancies at lower frequencies. surge transfer and how it interacts with other external components during these events. the model can work well for voltages significantly lower than the transformer rated voltage. Although good agreements were obtained between voltage ratios calculated from the measured admittance matrix. Finally. 7. These errors are due to the ungrounded windings of the delta side which produces very small zero sequence components of elements associated with this winding at low frequencies. In addition at relatively high frequencies where the core excitation flux varies very rapidly. o It has been tried to validate the model developed by studying voltage transfer from the delta side to the wye side and vise versa.2 Future Work A lot could be done to get an improved and more accurate model of the components especially the transformer. Therefore. o The generic test circuit was able to replicate actual VCB characteristics and expected results were obtained by looking into the effects of arcing time. the model can be a sufficient representation.overvoltages on the load side due to reignitions or prestrikes specially when inductive currents are interrupted. non linearity and core saturation of the transformers have not been accounted for. a wide band frequency dependent model of a power transformer was developed. However. This model is suitable for the study of transients induced due to VCB switching or other types which contain a very wide range of frequencies.
Cable Model The cable model simulates most of the characteristics of a real cable and this was verified to measurements on real cable. Transformer Model It was seen that the transformer model developed in this work is not accurate at lower frequencies. How ever the damping of the real cable was found to be higher than the PSCAD/EMTDC model besides the model doesn’t take in to account the presence of the armour. For the three phase transformer with 36 admittance matrix and 9 voltage ratio elements for example 45 different connections was made. Reconnection has to be made every time a single element of the admittance matrix or voltage ratio was measured. Hence more work could be done to get much more accurate model in this regard. One problem faced in this work was the measurement set up was so simple that it doesn’t allow easily reconnection. This was time consuming not mentioning its effect on the accuracy of the data set due to bad connection and inconsistency in the length of the connection cable used. The following is suggested: Improving the measurement set up The accuracy of this model depends on the accuracy of the measured admittance matrix and hence accurate measurement is required as the model is very sensitive to error in measurement. One way to improve the accuracy of the measured data is by designing Special measurement setup dedicated to measure the admittance matrix and voltage ratio which allow easily reconnection and accurate measurement. This was also seen during measurements where some elements of the admittance matrix and especially voltage ratio were very noisy. This was reflected on the measurements obtained on the admittance and voltage ratio. VCB Model The VCB model in this project simulates most of the phenomena of a real breaker very accurately and hence the authors feel it is good enough for its purpose. 69 . To deal with this problem the cables to be used in future work should be shielded. Besides the simple connection cables used has greatly contributed to the measurement error. However a more accurate model can still be obtained by considering more aspects that were not dealt in this work. hence more work should be done to get a refined and robust model by taking in to account a lot of aspects that were not taken in to consideration on this work.
are modified by removal of their zero sequence components as the zero sequence component of the admittance matrix is inaccurate due to small zero sequence current. A new zero sequence system has to be obtained by dedicated measurements and the two functions can then be approximated with rational functions which can be combined to produce the final model.Measuring the Zero sequence Admittance In accuracy of the model at lower frequencies for a transformer with ungrounded windings like the transformers we dealt with can be overcome by explicit measurement and modelling of the zero sequence system. YHL . In this work we have tried to measure the zero sequence admittance though we didn’t succeed to get accurate measurement. 70 . Blocks of Y. which correspond to ungrounded windings ( YHH . YLH ).
Völund “Experiences from Middelgrunden 40 MW Offshore Wind Farm. Chapman. January 1995. Sweet. November 2001.REFERENCES [1] D.Issue 11.” Power Quality and Reliability Center. Elektron magazine. Section 2. Winter Meeting.” IEEE Transactions on power delivery. Venthanar Ilango. Naef. [8] J.1. Vol. “Failures of encapsulated transformers for converter winders at Oryx Mine”. Vic Gosbell. [9] B. 2004. [7] W.Soerensen.2427. University of Wollongong. A.Larssen.J. “The Cost of Poor Power Quality”. No. J. [11] PSCAD manual. Proc. H. 2000. CDROM (6P). Gustavsen. Copper Development Association of UK. IEEE. 1. pp. 3. [6] Van Craenenbroeck T. Gustavsen. Nov.” IEEE Transactions on Power Delivery. [3] A. 3034. [4] Pritindra chowdhuri. [12] Janko KosmaE and Peter Zunko."Experimental and Numerical Analysis of Fast Transient Phenomena in Distribution Transformers". “Parameter Determination for System Transients.”Electrical Transients in Power Systems”. Sarath Perera. 71 . Brouwers B. 10. 2628 Oct. no.” in Proc. July 2005. 2nd Edition. IEEE/PES . Marly J P. De Ceuster J. [13] “Over voltages – Measurement and Statistical Simulation”. [5] D. Singapore.Clare. Greenwood. causes and propagation. S. TAVRIDA ELECTRIC. Van Dommelen D. “Electromagnetic Transients in Power systems”._ Spectrum. _Danish wind turbines take unfortunate turn. Duane Robinson. vol. 41. IEEE Power Engineering Soc.Christiansen. pp. Power Quality Application Guide. “Transient Overvoltages On The Electricity Supply Network – Classification.Winter Meeting. “Panel session on data for modeling system transients: Insulated Cables. [10] B. 2005. Industrial Group. vol. 2nd Edition. and D.Part II: Insulated Cables. “A Statistical Vacuum Circuit Breaker model for simulation of transient over voltages. Durbak.March 1991. 2001.” Copenhagen Offshore Wind. 20. De Herdt H. Martinez. WileyInterscience. E. P. [2] Vic Smith.
[26] http://www.. Phys. 3. Knobloch.[14] Mietek T. Vol. 12.sintef.”Wide Band Modeling of Power Transformers”.” [16] Helmer J. “Over voltages and Reignition behaviour of Vacuum Circuit Breaker”. switching capability of double and singlebreak vacuum interrupters – experiments on real high highvoltage demonstrationtubes. July 1993. Gustavsen and A. IEEE TRANSACTIONS ON POWER DELIVERY. NO. No. Giere. IEEE TRANSACTIONS ON POWER DELIVERY. Gerardus C. 10521061. 4.” IEEE Transactions on Power Delivery 2006. Lindmayer M. Popov. Gutierrez and Dieter Brauii “Voltage escalation and re ignition behaviour of vacuum generator Circuit Breakers during Load Shedding. Kärner. “Stability of lowcurrent vacuum arcs. Ottevangers” A High Frequency Transformer Model for the EMTP”.A. [22] A.” Rational approximation of frequency dependent admittance Matrices". 1996. L. H. 8. Gliiikowslti .W. Gustavsen. Vol. IEEE Trans. Mad (M). [23] Bjørn Gustavsen . hdoises R. [25] B. International Conference on Power System Transients – IPST 2003. Morched (SM). and Gopal Gajjar. “Mathematical Modeling of the High Frequency Behaviour of Vacuum Interrupters and Comparison with measured transients in power systems.no/produkt/VECTFIT/index. Power Delivery. Germany. L. vol. van der Sluis “Comparison of two vacuum circuit breaker arc models for small Inductive Current Switching. D: Appl. Kondala Rao. [15] M. no. [18] R P P Smeets. H. L. "Rational approximation of frequency domain responses by Vector Fitting". 1. 19. VOL. pp.1 9 (1986) 575587. Technical University of Braunschweig. [21] Marjan Popov. Oct.” J. Phys. PWRD. 14. 1. Semlyen. Wong. no. Berkley. 2002.and Hans De Herdt “Computation of Very Fast Transient Overvoltages in Transformer Windings”. IEEE Trans. vol. [19] S.M. [20] S. 4.”Development and Application of Vacuum Circuit Breaker Model in Electromagnetic Transient Simulation. NO. Lou van der Sluis. USA.” XVIIth International Symposium on Discharges and Electrical Insulation in vacuum. Snider and E. 3.” IEEE Transactions on Power Delivery. 17. Paap. 18.asp 72 . [24] B. C. IEEE Transactions on Power Delivery. July 1999. VOL. [17] B. Printed in Great Britain.energy. Lo. January 1997.C. No. JANUARY 2004. 10931098. J. OCTOBER 2003.
16. [30] Maialen Boyra. Vol. E transactions on circuits and systems—I: regular papers. [31] Daniel Mireanu.February 2001. No.”Enforcing Passivity for Admittance Matrices Approximated by Rational Functions”.3. vol. no. ” Transient Overvoltages in Cable Systems”. 2007. July 1999. Sweden. [29] Stefano GrivetTalocia. 73 . 2007.Gustavsen.vol. Göteborg. Göteborg.” Transient Overvoltages in Cable Systems”. No. Sweden. IEEE transactions on power systems. Gustavsen.[27] B. 9. Master of Science Thesis. 14. 51. IEEE Transactions on power delivery. september 2004. [28] B. “Rational approximation of frequency domain responses by vector fitting”. 1. “ Passivity enforcement via perturbation of Hamiltonian Matrices”. Master of Science Thesis.
Connect these blocks between corresponding nodes (connections) referred in 1 as shown in Fig. lumped R. and to which electrical connections they fall between. Implementation of 3phase Transformer model in PSCAD/EMTDC In this appendix the procedure for implementing the transformer model developed in this thesis in transient time simulation software PSCAD/EMTDC is demonstrated. ground to model the whole equivalent network as in Fig. 74 . L and/or C). The component definition is written as a script in the format $<TO> $<FROM> [<R>] [<L>] [<C>] where $<TO> and $<FROM> refer to the branch end nodes and [<R>] . A. [<C>] to the passive element values in Ohm.APPENDICES Appendix A. Double click on the newly created page module and create components corresponding to each one of the big branches. Henry and micro Farad respectively.e. The steps are summarized as follows: 1.2. A branch based interface was used to define the branches of the equivalent network. Create a new component page module in the circuit page with 3 connections on the left. Therefore each one of the big branches between two nodes or between a node and ground is modeled as sub modules which are then connected to form the whole network. The later can be used directly in PSCAD/EMTDC or other time simulation packages to build the transformer component. Branch design is accomplished by specifying the type and size of passive elements (i. The out put of the main Matlab source code is either a state space equivalent or a file of branch cards of an equivalent RLC network of the transformer.1. [<L>] . 2. 3 on the right and one at the bottom. A. Nested page modules are used to build the model as the number of branches is too big to implement as a single component.
3.1 Transformer component page module. 75 .3. Enter component definitions for each one of the components in 2 as script. A. A.Fig. A typical script page would look like as in Fig.
76 .Fig. A.2 Components corresponding to internode branches.
77 .Fig.3 Component definition as script page. A.
From (2) the admittance of the branch connected between terminal i and j may be written as y ij = −Yij (3) Similarly the branch between terminal i to ground can be computed from (1) y i = Yii − n j =1. Yij = − y ij (2) Where y i denotes the admittance of a branch connected between terminal i to ground and yij admittance connected between terminal i and j.B. j ≠ i y ij (4) If the value of yij from (2) is replaced in to (4) then yi = Yii + n j =1. including branches to ground. j ≠ i Yij (5) We may write (5) in compact form according to: yi = n j =1 Yij (6) 78 . Yii = y i + n j =1 .. Given the admittance matrix we can also compute the admittance of the branches connected between terminals and a terminal to ground.e. j ≠ i y ij (1) The offdiagonal elements are the negative of the admittances connecting buses i and j i.i.Appendix B. Network Realization Consider the multiterminal network shown in Fig.1 the admittance matrix for this network can be computed as A diagonal element Yii of the admittance matrix is obtained as the sum of admittances for all branches connected to terminal i..e.
1.B.1 is given by 79 . This gives a network of the form shown in Fig. The voltage and current at the terminals of these elements in time domain t and frequency domain s is given in Table B. capacitor. resistor.B. 1 y 12 2 y2n y1 n y1 n yn y2 Fig. and inductor.1 Multi. This network functions are either impedance or admittance. Network functions are expressed in terms of the complex frequency variable s. This rational function is called network function. Given the impedance or admittance function then it possible to find the network representation.Given the admittance matrix of a device in frequency domain the corresponding network can easily be obtained connecting branches between terminals and terminal to ground by realizing the admittances obtained from (3) and (6).terminal network equivalent of a device. resulting from the application of Laplace transformation to time domain network equations. Fundamentals of Network Synthesis A network is represented by a rational function in sdomain. As a passive net work is an inter connection of the basic passive elements. let’s start by driving the network function of this elements which is the basic for synthesis of a complex network including the network we are trying to find for the transformer model.
1 are zero then the impedance and admittance functions of the elements can easily be computed (7) and (8).B..Table B.1 Voltage and current relation ship of passive elements RESISTOR CAPACITOR INDUCTOR di ( t ) Relation ship v (t ) = L t dt between voltage v ( t ) = Ri ( t ) 1 v (t ) = i ( x ) dx + v ( 0 − ) t and current in time i ( t ) = Gv ( t ) 1 C 0 i (t ) = v ( t ) dt + i ( 0 − ) domain dv ( t ) L 0 i(t ) = C dt Relation ship V ( S ) = RI ( S ) between voltage I ( S ) = GV ( S ) and current in frequency domain V (s) = 1 v(0 − ) I (S ) + SC S I ( S ) = C [ SV ( S ) − v (0 − )] V ( S ) = L[ SI ( S ) − i (0 − )] I (s) = 1 i (0 − ) V (S ) + SL S The impedance and admittance functions are defined as the ratio of voltage to current transform and current to voltage transformer.2 Network function of the basic elements RESISTOR CAPACITOR Impedance Z(S) V (S ) V (S ) 1 Z (S ) = = R Z (S ) = = I (S ) I ( S ) CS Admittance Y(S) Y (S ) = I (S ) =G V (S ) Y (S ) = I (S ) = CS V (S ) INDUCTOR V (S ) Z (S ) = = LS I (S ) I (S ) 1 = V ( S ) LS Y (S ) = Suppose Y(s) represent the admittance function of one of the branches of nterminal network representation of the Fig. V (S ) (7) I (S ) I (S ) (8) Y (S ) = V (S ) If the initial conditions in Table.B. + d + s. Z (S ) = Table B.e S −a (9) (10) Y ( s) = r1 '− jr1 " r1 '+ jr1 " r2 '− jr2 " r2 '+ jr2 " r + + + + + ..1 and assume that it can be written in partial fraction expansion of the form Y (S ) = N 1 r + d + S .e s − a s − (a1 '− ja1 " ) s − (a1 '+ ja1 " ) s − (a 2 '− ja 2 " ) s − (a 2 + ja 2 " ) 80 .
+ d + s.S + − + 2 r1 2 r1 ' 2 2 r1 ' 1 2 r1 ' 2 r ' a '− 2 r1 ' ' a1 ' ' ) S− 1 r1 ' ' 2 r ' '2 a ' ' 2 (1 + 2 ) a ' ' 2 (1 + 1 2 ) r1 ' r1 ' The above equation can be written as 81 ...e s − a s − 2a1 ' S + a1 ' 2 + a1 " 2 s 2 − 2a 2 ' S + a 2 ' 2 + a 2 " 2 (11) The network function in (11) can be realized in to its corresponding network.. where (12) Y1 ( s ) = d Y2 ( S ) = S .represents admittance of a resistor with a resistance value of R = 1 d Y2 ( S ) = e.1 Y1 ( s ) = d .Y ( s) = 2r ' S − 2r1 ' a1 '−2r1 " a1 " 2r2 ' S − 2r2 ' a 2 '−2r2 " a 2 " r + 21 + + .represents admittance of a capacitor with a capacitance value of C = e One can write Y3 ( S ) as 1 1 1 a Y3 ( S ) = = . We can further simplify (12) by Y ( S ) = Y1 ( S ) + Y2 ( S ) + Y3 ( S ) + Y4 ( S ) + Y5 ( S ) + .S . r (13) 2 r ' S − 2 r ' a '− 2 r " a " 1 Y4 ( S ) = 1 1 1 1 1 s 2 − 2 a1 ' S + a1 ' 2 + a1 " 2 = 1 a ' ' r ' ' a1 ' .. Each part represent admittance of an element or combination of elements and the final network can be obtained by connecting these branches in parallel as admittance in parallel sum up.. where Z 3 ( S ) = S − is an impedance function representing a Z 3 (S ) 1 r r S− r r 1 a series combination of an inductor with an inductance value of L = and a resistor with r −a a resistance value of R = as impedances in series sum up.e r Y3 ( S ) = S −a Now let’s realize each section starting from the simplest Y1 ( s ) From Table..B.
R = 2r1 2r1 ' 2 2r1 ' (15) L= Z 5 (S ) = 1 2r1 ' 2r ' a '−2r1 ' ' a1 ' ' ) S− 1 2 r '' r ' '2 a ' ' 2 (1 + 1 2 ) a ' ' 2 (1 + 1 2 ) r1 ' r1 ' (16) Similarly if (16) is rearranged as Z 5 (S ) = 1 CS + G (17) Then Z 5 ( S ) can be realized as a parallel combination of a capacitor and a resistor whose values can be computed by equating (16) and (17).Y4 ( S ) = 1 Z 4 (S ) + Z 5 (S ) Z 4 ( S ) and Z 5 ( S ) are in series and each can be realized as follows Z 4 ( S ) is a series combination of an inductor with a resistor whose values can be computed as follows Z 4 (S ) = (14) can be written as: 1 a' ' r ' ' a1 ' .S + − 2r1 2r1 ' 2 2r1 ' (14) Z 4 ( S ) = LS + R Equating (14) and (15) gives the values of L and R as 1 a' ' r ' ' a1 ' − . C= 2r1 ' 2r ' a '−2r1 ' ' a1 ' ' ) .G = − 1 2 r1 ' ' r1 ' ' 2 2 2 a' ' (1 + 2 ) a' ' (1 + 2 ) r1 ' r1 ' The final network will then look like: 82 .
.2 Network realization of Y(s). 83 .. Cc Lc Gc .B..e term Real Pole Complex conjugate pole Rr Rc R0 C0 ..Constant term s. Lr Fig.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.