Sie sind auf Seite 1von 126

568

Transformer Energization in Power Systems:


A Study Guide

Working Group
C4.307

February 2014
TRANSFORMER
ENERGIZATION IN
POWER SYSTEMS: A
STUDY GUIDE
WG C4.307

Members

Zia Emin, Convenor (GB)


Manuel Martinez-Duro, Task Force Leader (FR), Marta Val Escudero, Task Force Leader (IE)
Robert Adams (AU), Herivelto S. Bronzeado (BR), Bruno Caillault (FR), Nicola Chiesa (NO), David Jacobson (CA),
Lubomir Kocis (CZ), Terrence Martinich (CA), Stephan Pack (AT), Juergen Plesch (AT), Michel Rioual (FR), Juan A
Martinez-Velasco (ES), Yannick Vernay (FR), Francois Xavier Zgainski (FR)

Copyright © 2014

“Ownership of a CIGRE publication, whether in paper form or on electronic support only infers right of
use for personal purposes. Are prohibited, except if explicitly agreed by CIGRE, total or partial
reproduction of the publication for use other than personal and transfer to a third party; hence
circulation on any intranet or other company network is forbidden”.

Disclaimer notice

“CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any
responsibility, as to the accuracy or exhaustiveness of the information. All implied warranties and
conditions are excluded to the maximum extent permitted by law”.

ISBN : 978-2-85873-263-0
Transformer Energization in Power Systems: A Study Guide

Acknowledgements
The convenor wishes to express his thanks and gratitude to Nicola Chiesa, Manuel Martinez-Duro, Marta Val
Escudero, and Terrence Martinich for their enduring hard work during the preparation of this Technical Brochure.

Page 2
Transformer Energization in Power Systems: A Study Guide

Transformer Energization in Power


Systems: A Study Guide
EXECUTIVE SUMMARY ........................................................................................................................................5

ACRONYMS AND ABBREVIATIONS ....................................................................................................................6

CHAPTER 1 INTRODUCTION ..............................................................................................................................7

CHAPTER 2 INRUSH CURRENTS AND TRANSFORMER ENERGIZATION .......................................................9


2.1 Transformer Saturation and Inrush Currents ...............................................................................................9
2.2 Harmonic Content of the Inrush Current....................................................................................................10
2.3 Sympathetic Interaction ............................................................................................................................12
2.3.1 Parallel Sympathetic Interaction .......................................................................................................... 12
2.3.2 Series Sympathetic Interaction ............................................................................................................16
2.4 Pseudo-inrush due to Re-energization ......................................................................................................19

CHAPTER 3 TRANSIENT VOLTAGE DISTORTION DUE TO TRANSFORMER ENERGIZATION ..................... 22


3.1 RMS-Voltage Drop ...................................................................................................................................22
3.2 Resonant Overvoltages due to Network Parallel Resonances ...................................................................23

CHAPTER 4 MITIGATION TECHNIQUES AND BEST PRACTICES................................................................... 27


4.1 Mitigation of the Inrush Currents ...............................................................................................................27
4.2 Mitigation of the Temporary Overvoltages.................................................................................................29

CHAPTER 5 SIMULATING TRANSFORMER ENERGIZATION TRANSIENTS...................................................31


5.1 Modelling of the Network for EMT Studies ................................................................................................31
5.1.1 Study Zone..........................................................................................................................................31
5.1.2 Transformers....................................................................................................................................... 32
5.1.3 Other Components of the System ........................................................................................................45
5.2 Quantification of the Overvoltage Stress in Transformers and Surge Arresters.......................................... 60
5.3 Parameter Uncertainty Assessment and Stochastic Simulation ................................................................. 63
5.3.1 Important Uncertain Parameters ..........................................................................................................64
5.3.2 Uncertainty Quantification and Modelling .............................................................................................64
5.3.3 Stochastic Simulation: the Monte Carlo Method ...................................................................................66
5.3.4 Example of Application to Transformer Energization Studies ...............................................................68
CHAPTER 6 CONCLUSIONS .............................................................................................................................70

BIBLIOGRAPHY/REFERENCES .........................................................................................................................73

ANNEX A ANALYTICAL CALCULATION OF INRUSH CURRENT ................................................................ 82

ANNEX B TRANSFORMER MODELLING: CALCULATION OF THE SATURATION CURVE


FROM THE NO-LOAD TEST .........................................................................................................85

ANNEX C CASE STUDY EXAMPLES & SIMULATION RESULTS VS. FIELD MEASUREMENTS ................. 88
C.1 RMS-Voltage Drop ...................................................................................................................................88
C.2 RMS-Voltage Drop and Inrush Currents....................................................................................................92

Page 3
Transformer Energization in Power Systems: A Study Guide

C.2.1 Background ......................................................................................................................................... 92


C.2.2 Model Validation..................................................................................................................................92
C.2.3 Comparison Between Simulation and Field Test for the Third Energization .......................................... 94
C.2.4 Conclusions ........................................................................................................................................95
C.3 Sympathetic Interaction: On-Site Tests Measurements and Simulations ...................................................98
C.3.1 Background ......................................................................................................................................... 98
C.3.2 System Data .......................................................................................................................................98
C.3.3 116 MVA Transformers Data .............................................................................................................100
C.3.4 Transformer and Network Modelling ..................................................................................................101
C.3.5 Simulation of Transformer Energizations Producing High Inrush Current ........................................... 103
C.3.6 Conclusions ......................................................................................................................................112
C.4 Sympathetic Interaction: On-Site Tests Measurements ........................................................................... 114
C.4.1 Introduction .......................................................................................................................................114
C.4.2 Hydro Power-plant Brief Description ..................................................................................................114
C.4.3 Energizing a Transformer, the Second One Being Connected and Unloaded ..................................... 114
C.4.4 Energizing a Transformer, the Second One Being Connected and Loaded ........................................ 117
C.4.5 Conclusions ......................................................................................................................................119
C.5 Controlled Switching Eliminates Inrush Current.......................................................................................120
C.5.1 Introduction .......................................................................................................................................120
C.5.2 Commissioning of the POW Controller...............................................................................................120
C.5.3 Conclusions ......................................................................................................................................125

Page 4
Transformer Energization in Power Systems: A Study Guide

EXECUTIVE SUMMARY
This Technical Brochure provides a generic guidance on transformer energization studies in power systems by
suggesting tools and methodologies for practising engineers to study and analyse voltage-related problems.

Transformer energization is a regular operation in transmission and/or distribution networks and the issues
associated with it are not new. However, it is becoming increasingly relevant due to the changes that are shaping
power networks. Frequent modifications in the network topology due to liberalisation and the expected increase in
the penetration of offshore renewable energy are typical examples of such drivers for this change. Hence, there is
a need for power systems engineers, either at the planning or operational stage, to be aware of potential problems
that may arise due to transformer energization and, if necessary, study their probability of occurrence, likely effects,
and possibly evaluate various mitigation techniques that may be required in order to alleviate identified issues.

The Technical Brochure covers comprehensive description of inrush currents in transformer energization and the
various scenarios where they may be observed. Two distinct voltage issues associated with transformer
energization are explained in detail: a drop in RMS voltage and overvoltages due to network parallel resonances.

An overview is provided of various mitigation measures to control inrush current, including a proven and effective
technique to control the closing times of the energizing circuit breaker.

A large section of the brochure is dedicated to modelling and it provides extensive guidelines and explanations to
simulate both RMS voltage drop and overvoltages due to network parallel resonances and to assess their
probability of occurrence. Specific sections cover the assessment of the potential damaging effects of TOVs and
the treatment of uncertainty in energization studies.

Page 5
Transformer Energization in Power Systems: A Study Guide

ACRONYMS AND ABBREVIATIONS


AC Alternating current
ACSR Aluminium conductor steel-reinforced
ATD Ampere-turn diagram
AVR Automatic voltage regulator
BCTRAN British Columbia transformer model
CB Circuit breaker
CIGRE International Council on Large Electric Systems
CT Current transformer
CVT Capacitive voltage transformer
DC Direct current
e.m.f. Electromotive force
EHV Extra high voltage
EMT Electromagnetic transient
FDNE Frequency dependent network equivalent
FRT Fault ride through
GIC Geomagnetic induced currents
GSR Global stress rate
GSU Generator step-up
HV High voltage
HVAC High-voltage alternating current
HVDC High-voltage direct current
IEC International Electrotechnical Commission
IEEE Institute of Electrical and Electronics Engineers
IPP Independent power producer
LV Low voltage
MCOV Maximum continuous operating voltage
OHL Overhead line
p.u. Per unit
POI Point of interconnection
POW Point-on-wave
RDDS Rate of decrease of dielectric strength
RMS Root mean square
STC Saturable transformer component
SVC Static VAR compensator
TOV Temporary overvoltage
UMEC Unified magnetic equivalent circuit
WT Wind turbine generators
XFMR Hybrid transformer model

Page 6
Transformer Energization in Power Systems: A Study Guide

CHAPTER 1 INTRODUCTION
In electric power networks, energization of transformers, even on a daily basis, is a commonplace operational
action and it is generally performed without any adverse consequences. However, as it is the case with all
operations involving switching, such actions generate currents and voltages that are transient in nature. In the vast
majority of cases these transient currents and voltages generated by such actions are safely damped within the
power network. However, in certain situations, the transient currents and voltages generated during the
energization can lead to power quality issues in the supply system and, in extreme cases, produce transformer
insulation degradation or surge arrester failure.

Potential problems related to the reduction of power quality due to transformer energization can be categorized
under RMS-voltage drop caused by high transient currents, temporary overvoltages (TOV) generated when a
parallel resonance within the supply network is excited by energization currents and the misoperation of relays
causing protection tripping where the inrush current is interpreted as fault current.

In severe cases, transformer insulation damage may occur if either a slow-front overvoltage or a temporary
overvoltage is generated during the energization. The occurrence of slow-front overvoltages depends on the
capacitive and inductive characteristics of both the transformer and the neighbouring supply network it is connected
to. The occurrence of temporary overvoltages depends on the existence of low-frequency parallel resonances in
the supply network, which may be excited by the transformer inrush current with its rich harmonic content. The
latter may be of particular interest during network restoration following a system-wide blackout. This is mainly
because during such times the network tends to be weak, i.e. characterized by a relatively low short circuit level
and hence higher system inductance with the possibility of resonance at a much lower frequency accompanied by
ineffective damping of overvoltages due to light loading conditions.

As mentioned above, in the vast majority of cases transformer energization is performed without any adverse
effects or consequences. Issues associated with transformer energization are not new in transmission and
distribution networks but nowadays are becoming more relevant due to the changes that are affecting the
development of power networks. Liberalisation of electricity markets has led to an increased number of participants
with frequent changes in the network topology, and thus the possibility of an increased number of switching
operations. The expected increase in the penetration of offshore renewable energy resources will increase the
utilisation of cable circuits while integrating them into onshore grids as well as their own internal networks.
Considering that the capacitance per unit length of a cable circuit is much higher than that of an overhead line, this
will normally bring system resonant frequencies to the lower end of the frequency spectrum, closer to low-order
harmonics. The result is an increase of the risk of temporary overvoltages due to the possibility of exciting these
parallel resonances. Furthermore, energization of offshore renewable energy systems, where each wind turbine
generator is accompanied by its transformer, may generate complex sympathetic interactions. Wind turbine
generator transformers have a relatively small power rating (several MVA, according to the wind turbine generator
rating), but relatively higher inrush currents. Therefore, power systems engineers, either at the planning or at the
operational stage, need to be aware of potential problems that may arise from transformer energization and, if
necessary, study their probability of occurrence, their likely effects, and possibly evaluate various mitigation
techniques that may be required in order to alleviate identified issues.

In existing networks, where some components may be approaching their end of life, one of the main considerations
is to prolong the transformers’ life. In addition to condition monitoring and ageing control, this implies the use of
solutions for the mitigation of transients, such as inrush currents, to reduce mechanical and electrical stresses on
the equipment. This is a new challenge, not considered when building networks in the 1970’s and 1980’s when the
main focus was on insulation coordination issues.

The following chapters provide information and suggest tools and methodologies for practising power systems
engineers in order to study and analyse the voltage-related problems (RMS voltage drop and temporary
overvoltages) created by inrush currents during transformer energization. For slow-front overvoltages generated
during transformer energization the reader is referred to IEC 60071-4 (Computational guide to insulation co-
ordination and modelling of electrical networks).

Page 7
Transformer Energization in Power Systems: A Study Guide

CHAPTER 2 concentrates on the description of the inrush current phenomenon during transformer energization,
the characteristics of these currents and the special case of sympathetic interaction between the transformer that is
being energized and a one that is already energized. CHAPTER 3 describes the mechanisms by which the RMS-
voltage drop and the temporary overvoltages are generated and the system conditions under which these
phenomena may appear. CHAPTER 4 provides information on some usual techniques that can be implemented to
mitigate RMS-voltage drop and temporary overvoltages. CHAPTER 5 is dedicated to the modelling of power
system components and provides exhaustive guidelines to simulate both phenomena and to assess their
probability of occurrence.

ANNEX A presents a review of some formulae for the analytical calculation of the inrush current. ANNEX B gives
an example on how to calculate the saturation curve of a transformer from the no-load test report and ANNEX C
presents several study cases/examples, most with comparative simulation and field measurements.

Page 8
Transformer Energization in Power Systems: A Study Guide

CHAPTER 2 INRUSH CURRENTS AND TRANSFORMER ENERGIZATION


2.1 Transformer Saturation and Inrush Currents
The transformer core may become saturated due to an abrupt change in the voltage applied to it. This may be
caused by switching transients, out-of-phase synchronization of a generator, external faults and fault clearance.
When saturated, a transformer absorbs a magnetization current, also known as inrush current, which can reach
several times the nominal current of the transformer. The energization of a transformer normally yields the most
severe case of inrush current as the flux in the core can reach a maximum theoretical value of 2 to 3 times the
rated peak flux [1]. A qualitative and simplified representation of the inrush current phenomenon is illustrated in
Figure 2-1 for energization at voltage zero crossing. The flux-linkage is calculated as the time-integral of the
voltage applied to the transformer (upper-left part of Figure 2-1). The initial value of the flux-linkage is determined
by the residual flux in the transformer core prior to energization. The flux-linkage/current relation is nonlinear and is
determined by the saturation curve of the transformer (upper-right part of Figure 2-1). Therefore, the magnetization
current of a transformer contains harmonics. When a transformer is energized, the initial value of the flux may differ
from the prospective flux1. This causes a DC offset of the flux-linkage and a higher-than-rated peak value. The
result is an inrush current that may be several times the value of the nominal current. Due to the low relative
permeability of the ferromagnetic material in saturation, a marginal increase in the peak of the flux linkage results in
a magnification of the inrush current (lower-right corner of Figure 2-1). The inrush current transient is characterized
by asymmetrical current waveforms that are damped in some tens of cycles primarily by the series resistances of
the systems (transformer winding resistance, transmission line and cable series resistance, generator winding
resistance, etc.). A qualitative representation of the inrush current transient is shown in the lower-left corner of
Figure 2-1.

The main factors affecting the inrush current magnitudes can be divided into: transformer design, initial conditions,
and network factors.

The design of a transformer can affect the magnitude of the inrush current as it can shift the steady state operating
point on the saturation curve. A transformer with an operation point closer to the knee area of the saturation curve
is easily brought into saturation. Initial conditions affecting the magnitude of inrush current are residual flux and the
point-on-wave (POW) energization. These influence the magnitude of inrush currents and affect the DC offset of
the flux-linkage and the saturation of the transformer. The residual flux is the flux that remains trapped in the core
due to a previous de-energization of the transformer and defines the initial DC offset of the flux in the core.
Energization at a voltage zero crossing results in the most severe inrush current for a transformer as it induces a
flux-linkage of theoretically up to 2 p.u. (with 1 p.u. DC offset); the residual flux adds on top of that giving a
maximum possible flux-linkage of almost 3 p.u. Energizing a transformer at voltage peak, results in no DC offset
other than that caused by the initial residual flux. High network impedance acts as a limiting factor for inrush
current. The high current causes a voltage drop at the transformer terminals that limits the saturation of the
transformer.

1
The prospective flux is the flux that would circulate in the transformer core under steady state conditions.There is no
induced flux before energization, but the source voltage has the prospect to create an induced flux.

Page 9
Transformer Energization in Power Systems: A Study Guide

Figure 2-1 Qualitative representation of the inrush current phenomenon and the effect of the residual flux.

2.2 Harmonic Content of the Inrush Current


Transformer saturation is a highly nonlinear phenomenon and hence the inrush current contains harmonic and DC
components besides the fundamental component. The variation of the harmonic content of the transformer inrush
current with time has been analysed in several references [2, 3]. To obtain the magnitude and phase shift of each
harmonic component, [3] proposed the application of a Fourier analysis for each cycle of the inrush current
separately.

Figure 2-2 shows the peak values as a function of time of the main harmonic components of a recorded inrush
current [4]. Note that the peak value of any harmonic component during one cycle is generally different from its
peak during another cycle. For the case shown in the figure, the 2nd harmonic is by far the dominant one [4]. Note
also that the peak values of some harmonics decrease to zero and then increase again to a value higher than the
initial value at the instant of energization. A phase shift inversion is also common as the magnitude of a harmonic
passes through zero [3].

Another example of recorded inrush current waveforms along with the variation of the main harmonic components
with time during the energization of a 735 kV, 510 MVA autotransformer consisting of three single-phase units is
given in [2]. Individual harmonics vary with time and some attain their maximum value a few cycles after
energization.

The evolution of harmonic currents cannot be generalized as it depends highly on the transformer, feeding network
and initial energization conditions. For example, the cases given in Figure 2-2 and in [2] show different
developments of 2nd and 3rd harmonics. This implies the need for a case-by-case study when a specific network
configuration is identified to be susceptible to overvoltages.

The fact that the individual harmonic currents vary as the inrush current decays, explains one of the most typical
characteristics of the overvoltages generated as a consequence of transformer energization. The maximum
overvoltages often occur during the decay of the inrush current and not immediately after energization (when the
individual harmonics attain their maximum values).

Page 10
Transformer Energization in Power Systems: A Study Guide

Figure 2-2 Harmonic components of inrush current [4]

Page 11
Transformer Energization in Power Systems: A Study Guide

2.3 Sympathetic Interaction


Sympathetic interaction can occur when a transformer or shunt reactor is energized onto a system with long
transmission lines in the presence of other electrically close and energized transformers or shunt reactors. In the
case of shunt reactors this is less likely to occur since the knee of the saturation characteristic is relatively high
compared to power transformers. Sympathetic interaction significantly changes the duration and the magnitude of
the transient magnetizing currents in the transformers involved [4, 5, 6].

In practice, transformers are energized in series or in parallel with other transformers already in service. On
systems with appreciable series resistance, this inrush transient may trigger a transient interaction between the
transformer being energized and those that are already in operation. This occurs because the transformers already
in service go into saturation. This saturation is produced by the asymmetrical voltage waveforms at the busbar due
to the asymmetrical voltage drop across the series resistance of the system caused by the inrush current.

The occurrence of saturation of transformers already in operation during the inrush transient of a transformer being
energized (sympathetic interaction) was first reported by Hayward [7] in 1941 following field tests trying to establish
the reason for misoperation of transformer differential relays. It is reported that transient magnetic currents of
higher magnitude could flow “not only in the transformer being switched but also in other parallel transformers”, and
that these transient currents last longer, with the currents decaying at a much slower rate than would occur had the
transformers been switched onto a system having no other connected transformers. In the 1980’s, this
phenomenon appeared in Brazil in an SVC transformer connected to a 230kV busbar. This transformer invariably
tripped off by operation of transformer protection due to the high DC component in the magnetizing current when
one of the 100 MVA, 230/69kV transformers of the same substation was switched onto the 230 kV busbar [5].
Similar occurrences were also noticed during the energization of large shunt reactors near SVCs and HVDC
converters [3, 8, 9].

2.3.1 Parallel Sympathetic Interaction


The transient sympathetic interaction between paralleled transformers may be explained by using the circuit shown
in Figure 2-3. Parallel here means that only the transformer primary windings are connected in parallel to each
other. The secondary windings are not. So, when the switch S is closed, the transformer T2 is energized and its
primary winding is connected in parallel with that of the unloaded transformer T1 already connected to the system.

Figure 2-3 Circuit used to explain the sympathetic interaction phenomenon.

Page 12
Transformer Energization in Power Systems: A Study Guide

Figure 2-4 Transient currents calculated during a sympathetic interaction between two identical transformers of
180MVA, 275/66 kV connected in parallel.

Page 13
Transformer Energization in Power Systems: A Study Guide

Figure 2-4 shows a representative simulation of the transient inrush currents when T 2 is energized in parallel with
the already energized unloaded transformer T1. As can be seen in this figure, the sympathetic interaction
phenomenon takes place between the two transformers, with the peaks of the offset magnetizing current i 1
(sympathetic magnetizing current) and the inrush magnetizing current i 2 occurring in the opposite polarity to each
other on alternate half cycles. Note that the supply current i sys (lower part of Figure 2-4) is the sum of the currents i1
and i2. The sympathetic interaction may be explained as follows. Before the switch S is closed only the magnetizing
current of the unloaded transformer T1 flows through the system impedance from source to transformer. When
transformer T2 is energized, a transient inrush current is drained from the generator (e.m.f.), which flows through
the system. Due to the almost entirely unidirectional characteristic of this inrush current, the voltage drop across the
series resistance Rs makes the voltage at the transformer's busbar (point of common coupling) asymmetrical. As
the flux in a transformer is strictly proportional to the integral of the voltage waveform at the transformer terminals,
the flux generated in the transformer T1 begins to be asymmetrical by an amount which may be given by:
+

1 = [( + 1 ) 1 + 2] Eq. 2-1

where 1 is the flux change per cycle in the transformer T1 and r1 is the winding resistance of T1. It should be
noted that reactances (both system and the transformer) have not been taken into account in the above equation
as they don’t affect the decay of DC flux.

As the transformer T1 is initially in service (at steady state), the offset flux in it is zero. Thus, the flux change per
cycle 1 will produce an increasingly offset flux in the transformer T1 driving it into saturation. As a consequence,
a sympathetic magnetizing current i1 is produced in the transformer, increasing gradually from the steady state
value to a considerable magnitude when the transformer saturates. It should be noted that the polarity of the
transformer saturation is determined by the sign of 1.

At the same time, the flux change per cycle 2 produced in the transformer T2, may be given by:
+

2 = [( + 2 ) 2 + 1] Eq. 2-2

where r2 is the winding resistance of transformer T2.

Since 2 is opposite in polarity to the initial flux offset in transformer T 2 caused by its energization, the effect of 2
is to reduce this initial flux offset, producing the well-known phenomenon of inrush current decay.

From Eq. 2-1 and Eq. 2-2, it can be seen that at the beginning of the inrush transient both flux changes per cycle
1 and 2 will depend mainly on the voltage drop caused by the inrush current i 2. Initially, as the transformer T1 is
not saturated, its magnetizing current i1 is the very small steady state magnetizing current and essentially
symmetrical. Hence, it does not cause any appreciable flux change per cycle. When the transformer T1 becomes
saturated, with the opposite polarity to transformer T2, the peaks of the sympathetic magnetizing current i1 will
occur with opposite polarity to the peaks of the inrush current i 2 (on alternate half cycles). As a consequence, the
voltage asymmetry on the busbar caused by the inrush current i 2 during one half-cycle is gradually reduced by the
voltage drop produced by the sympathetic magnetizing current i 1 during the subsequent half-cycle. This will
decrease both flux changes per cycle 1 and 2 reducing the rate of change of the magnitude of both the
increasing current i1 and the decaying current i2. A few cycles later the flux change per cycle 1 reaches zero and
hence i1 stops increasing. Within this cycle:

1 = 2 Eq. 2-3
+ 1

Thereafter, the polarity of the flux change per cycle 1 inverts, reducing the flux in the transformer T1. As a result,
the sympathetic magnetizing current i1 begins to decay as does the inrush current i2. Under this condition, the

Page 14
Transformer Energization in Power Systems: A Study Guide

voltage on the busbar (transformers terminals) presents a wave shape nearly symmetrical with the flux change per
cycle in each transformer depending basically on the winding resistance of each transformer, i.e.:
+

1 = ( 1 1) Eq. 2-4

and
+

2 = ( 2 2) Eq. 2-5

It is interesting to note that, in this case, the system resistance Rs plays the paradoxical role of keeping both
transformers T1 and T2 saturated (on alternated half cycles), with the currents i 1 and i2 being concomitantly cause
and effect of the saturation of the transformers. That is to say: the voltage drop across the system resistance R s
produced by the current i1 during one half cycle reduces the flux offset in transformer T 1 and, at the same time,
increases the flux offset in the transformer T2. In the subsequent half cycle it is the current i2 that produces the
voltage drop across Rs, which increases the flux offset in the transformer T 1 and, at the same time, reduces the flux
offset in T2. This sequence repeats itself, developing the phenomenon of sympathetic interaction between the
transformers, keeping them saturated for a long period of time. The currents i 1 and i2 keeps flowing for a prolonged
period of time until the transformers reach their steady state magnetizing conditions. This may take several
seconds or perhaps minutes, depending essentially on the transformer winding resistances.

This aspect is also observed quite commonly with transformers for wind farms [10] when several transformers may
be energized within a minute. A typical case with a wind farm showing the simulation of the sequential energisation
of five step-up transformers is shown in Figure 2-5. The figure shows the flux in the iron core and the inrush current
(phase A) of the first energized transformer. Due to sympathetic interactions, the stress on this first transformer is
repeated five times during the total energization process.

Figure 2-5 Evolution of the magnetic flux in the iron core and the inrush current of the first transformer.

Page 15
Transformer Energization in Power Systems: A Study Guide

2.3.2 Series Sympathetic Interaction


This section discusses sympathetic interaction when the transformers are connected in series. Series here means
that the primary winding of one of the transformers is connected to the supply system with its secondary winding
feeding the transformer that is being energized (Figure 2-6).

system
R isys=i1+i2’
i2
i1 S

RS rp1 rp2 rs2


rs1
generator

TS1 TS2

Figure 2-6 Circuit used to explain the series sympathetic interaction phenomenon for series-connected transformers.

Figure 2-7 shows the transient currents measured during laboratory tests with two small transformers, where T s2
was energized in series with the unloaded transformer Ts1. As can be seen in this figure, the sympathetic
interaction phenomenon takes place between the two transformers, with the current i sys in the transformer Ts1 being
equal to the inrush current of Ts2 reflected on the primary side of Ts1 plus the sympathetic magnetising current in
the transformer Ts1 itself, i.e. isys = i1 + i’2.

Page 16
Transformer Energization in Power Systems: A Study Guide

Figure 2-7 Series sympathetic interaction measured during laboratory tests


a) Current in the transformer Ts1 (isys = i1 + i’2); b) Current in the transformer Ts2 (i2).

This phenomenon is also shown by simulating two identical three-phase transformers of 180 MVA, 275/66 kV, in
series (Figure 2-8). The results (Figure 2-9) suggest that the interaction between series transformers is more than
somewhat similar to the interaction occurring between the transformers in parallel.

T1 T2
isys S

i2
275/66 kV 66/275 kV

Figure 2-8 Electrical system used in the simulation to investigate the sympathetic interaction between series
transformers.

Page 17
Transformer Energization in Power Systems: A Study Guide

Figure 2-9 Transient currents calculated during a sympathetic interaction between two identical transformers of
180MVA, 275/66 kV connected in series.

Note that when T2 is energized its transient inrush current i2 flows through the secondary of transformer T1
producing a corresponding “inrush current” i’2 in the primary of T1. This current (i’2) is added to the current i1 to
yield the total current isys which flows through the circuit formed by the supply system and the primary winding of T1.
Thus:

= 1 + 2 Eq. 2-6

It should be noted that transformer T1 “sees” the inrush current i2 as a load current. The flux change per cycle s1
in transformer T1 can be given by
+

1 = + 1 Eq. 2-7

where Rsys is the system resistance and rp1 is the resistance of the primary winding of T1, or

Page 18
Transformer Energization in Power Systems: A Study Guide

1 = + 1 1 + + 1 2 Eq. 2-8

The flux change per cycle s1 will drive transformer T1 that was initially in the steady state into saturation, with the
offset magnetizing current i1 increasing gradually until the flux change per cycle s1 becomes zero. In this
condition,
+ +

1 = 2 Eq. 2-9

From this point onwards, the flux change per cycle s1 inverts the polarity so that the offset magnetizing current i 1
starts to decay, developing the sympathetic interaction between the series transformers T1 and T2. The rate of
decay of the inrush current i2 and, consequently, the rate of decay of the "primary inrush current" i’2 are essentially
determined by the flux change s2 in transformer T2, which can be described by
+

2 = + 1 + 1 + 2 2 Eq. 2-10

or
+

2 = + 1 1 + + 1 2 +( 1 + 2) 2 Eq. 2-11

where rs1 is the resistance of the secondary winding of T1 and rp2 is the resistance of the primary winding of T2
connected to the secondary of T1. It should be observed that when Eq. 2-9 is satisfied, the first two terms of Eq.
2-11 add to zero. In this condition, the flux change s2 will depend basically on the voltage drop across the total
resistance in the circuit formed by the secondary winding of T1 and the primary winding of T2. This indicates,
paradoxically, that the total resistance in the primary side of transformer T 1, i.e. Rsys + rp1, does not contribute
effectively to the decay of the inrush current in transformer T2 during the sympathetic interaction.

This interaction is greatly reduced if the series resistance of the supply system is small and also if the resistance of
the circuit between the transformers is relatively large. The latter could represent the case where transformers are
separated by an appreciable length of transmission line.

2.4 Pseudo-inrush due to Re-energization


“Pseudo-inrush” is a phenomenon by which transformers already in operation are driven into saturation after a fault
has been cleared [11, 12, 13, 14] and the normal system voltage is restored at the transformer terminals. This
process is also interchangeably referred to as “re-energization”.

Voltage sags are RMS-voltage drops that last a short period of time. Momentary reduction in RMS voltage may be
caused by large motor starting, transformer energization, etc., but here only voltage sags due to short-circuit faults
are considered. The “point-on-wave of sag initiation” is the point on the fundamental voltage wave at which the
voltage sag initiates, for example, the instant when the fault occurs. The “point-on-wave of voltage recovery” is the
point on the fundamental voltage at which the fault is cleared and the voltage recovery takes place. Figure 2-10
shows an example of a voltage sag, indicating both point-on-wave of sag initiation and recovery [12].

The “pseudo inrush” is caused by transformer saturation that occurs due to the voltage recovery following a voltage
sag. This phenomenon is illustrated in Figure 2-11, where the flux after the voltage recovery exceeds the steady
state flux before the voltage sag [14].

The “pseudo inrush” phenomenon is investigated by simulation below. A single-phase electric system, composed
by a 50 Hz voltage-source supplying a transformer through a series impedance was used (Figure 2-12). The
voltage sag was generated by switching in a small resistor (Rsc).

Page 19
Transformer Energization in Power Systems: A Study Guide

Figure 2-10 Measured voltage sag.

Figure 2-11 Behaviour of the magnetic flux within a transformer before, during and after voltage sag. ( pr =
prospective flux)

Figure 2-12 Electric circuit used to simulate the transformer “pseudo inrush” phenomenon.

Page 20
Transformer Energization in Power Systems: A Study Guide

Figure 2-13 Simulation results of transformer saturation due to voltage sag recovery.

Figure 2-13 shows the result of a simulation where the voltage sag starts at the instant the voltage crosses zero in
the positive direction (t=0.04s) and finishes when the short-circuit current (i sc) crosses zero (t 0,07s). The
waveforms appearing in the upper traces of Figure 2-13, show the voltage (vs2) on the bus-bar to which the
transformer is connected and the voltage (vs) generated by the source. Note that the depressed voltage exhibits an
abrupt phase-angle jump.

The waveforms at the bottom of Figure 2-13 are the magnetic flux within the transformer and the pseudo
magnetizing inrush current. Notice how the flux behaves as the transformer saturates after the voltage recovery. It
is interesting to observe that the peaks of the magnetic flux waveform flatten during saturation when the pseudo
magnetizing current is increasing. The distortion on the recovered voltage is due to the voltage drop across the
system impedance caused by the pseudo magnetizing current.

When this phenomenon occurs, the “pseudo inrush” magnetizing currents in all transformers connected to the
system will add together causing voltage drops across the system impedance (see Section 3.1), slowing the
voltage recovery and distorting the system voltage waveforms. Similar transformer behaviour may be observed
during geomagnetically induced current (GIC) events.

This problem, which is generally more severe in weak systems, may lead to tripping by under-voltage and
overcurrent relays. Furthermore, if the harmonics of the magnetizing current coincide with resonances in the
system, harmonic voltages and currents of high magnitude will build up, causing operational problems such as
surge arrester failures and capacitor bank over-current relay misoperation.

In general, the “pseudo inrush” phenomenon is not often observed as the “starting” and "ending" points of the
voltage sag are more likely to occur near maximum voltage. This is because most faults are associated with a
flashover, which is more likely to occur near voltage maximum. Also, as fault clearing takes place at current zero
and the power system is essentially inductive, the “ending” point of voltage sag occurs near to voltage maximum as
well. In this case, there is no DC component of the flux after the voltage recovery and thus no “pseudo inrush”
current is generated.

Page 21
Transformer Energization in Power Systems: A Study Guide

CHAPTER 3 TRANSIENT VOLTAGE DISTORTION DUE TO TRANSFORMER


ENERGIZATION
3.1 RMS-Voltage Drop
Transformers may be switched on and off for various reasons, such as to conduct equipment maintenance, to
temporarily reconfigure the station bus by local or remote supervisory switching, or protective relay operation, and
sometimes due to protection misoperation or operator error. While network transformers are seldom switched,
generator transformers may be switched more frequently depending on dispatch requirements. Hence, transformer
energization or re-energization is a normal planned operation in an electric power system. Sometimes, energizing a
transformer results in the transformer drawing a relatively large initial inrush current which decays over time to a
much smaller steady state magnetizing current. The transient magnetizing current that occurs during transformer
energization (the inrush current) is produced by transformer core saturation following switch-on. This current, which
must be supplied from the system sources such as generators and large motors, flows through the network
impedance to the transformer being energized. Consequently, there is a voltage drop across the network
impedance and a drop in the line voltages where the effect increases in the direction towards the transformer. The
time required for the inrush current to decay depends on losses in the circuit (including the resistive component of
the system impedance and the transformer winding resistance), the transformer leakage and magnetizing
impedances and the proximity to other significant energized transformers [4]. Since the magnetizing impedance of
the transformer is high, the decay of the inrush current can take seconds to minutes.

Short duration abrupt voltage drops that are caused by faults, by large inrush currents produced by transformer
energization or large starting currents during motor starting, are called voltage dips or sags, dip being the IEC
terminology and sag, though not formally defined, being a common North American term. Typically, a voltage sag
or dip is a decrease in system voltage to between 0.1 p.u. and 0.9 p.u. at power frequency, lasting from one-half
cycle to one minute. Some high power electronic devices, such as adjustable speed drives, programmable logic-
based process controls in the mining and pulp and paper industry, and electronic chip manufacturing plants are
voltage sensitive. For such equipment, power quality problems associated with voltage sags are an important
design issue. Voltage sags as short as 2 or 3 cycles can affect critical equipment, such as those used in the
electronic chip manufacturing, and adversely impact the production process. Outages due to poor power quality
can have as detrimental impact as sustained power interruptions.

It is important to recognize that, for a power system that is operated prudently, voltage sag caused by transformer
energization does not fall into the category of voltage fluctuation and should not be characterized by the flicker
curve. Flicker is the impression of fluctuating luminance occurring when the supply to an electrically powered
lighting source is subjected to voltage fluctuation [15]. The flicker curve is applicable to a frequency of voltage
change events ranging from a few events per hour to 20 or more per second.

Unlike faults, transformer energization is a planned operational event for which the associated voltage sags must
be limited to the withstand level of voltage-sensitive industrial loads. Unfortunately, there is a lack of standards
which quantify voltage sag withstand capability for high voltage industrial loads. The ITIC (Information Technology
Industries Council) curve, formerly called CBEMA (Computer Business Equipment Manufacturers Association)
curve [16] defines the over-voltage and under-voltage susceptibility for a very limited segment of voltage sensitive
loads, namely, information technology equipment operating at 120 V nominal voltage. Utilities normally restrict the
voltage sag in the network caused by the operation of their customers and, in the absence of recognized
standards, have developed their own internal standards. Some examples of the limitations on rapid voltage change
in use in a few countries are given below.

In Great Britain, if flicker-producing voltage fluctuation is not relevant, Engineering Recommendation P28 states
that a maximum 3 percent voltage change limit applies, with no distinctions made as to the voltage level or the
number of events in a given period. The grid code in France addresses limitation on rapid voltage change and
states that the RMS voltage change must be less than 5% for the 63kV to 225kV network and less than 3% for the
400kV network. IEC 61000-3-7, Table 6, provides the planning levels for rapid voltage changes on public supply

Page 22
Transformer Energization in Power Systems: A Study Guide

systems as a function of frequency of occurrence, taking into account all installations which may cause rapid
voltage change. For HV/EHV, maximum voltage changes of up to 3% to 5% cannot exceed 4 events per 24 hours
and voltage changes of up to 3% cannot exceed 2 events per hour. For MV systems, the permissible voltage
changes are basically 1% higher than the HV/EHV levels. For events occurring between 2 per hour and 10 per
hour, the voltage change for HV/EHV and for MV is stated as 2.5% and 3%, respectively. In Canada, limitations on
rapid voltage change vary from province to province. For example, a Canadian utility allows voltage sags up to 3%
of the nominal voltage at the point of common coupling with a frequency of occurrence not to exceed once per
hour. Voltage sags from 3% to 6% up to once per eight-hour shift are allowed. With prior approval, this utility may
allow a voltage sag of up to 9%. The voltage sag is calculated on a per phase basis using a sliding one-cycle RMS
window. The most affected phase is used as the basis for quantifying the voltage sag event. Over the past fifteen
years, a reasonably common issue of voltage sag due to transformer energization in this Canadian utility system
has occurred with Transmission Customers and Independent Power Producers (IPPs), usually run-of-river hydro
projects, who propose to connect relatively large projects in a weak part of the network. Reference [17] describes
such a case where post-test EMT simulations of transformer energization were able to reproduce, with some
success, the magnetizing inrush currents recorded during commissioning tests.

The case C.1 in ANNEX C provides an example of the good agreement that can be achieved between EMT
simulation and actual field recordings when the network and the transformer are modelled in appropriate detail. The
example shows the RMS voltage sags caused by the energization of a pair of 400 kV non-identical generator step-
up transformers. The EMT simulation of the maximum voltage sag in each phase agreed very closely with the field
test. In order to achieve this agreement, a rather detailed system model was required. This model included
dynamic voltage support (SVCs) and mechanically switched shunt capacitor banks. The simulated rates of
recovery of the voltages were higher than the measurements, indicating that some further refinement in the
resistive losses in the network impedance and transformers could be considered. Having verified the adequacy of
the EMT model of the system and transformer for the purposes required, additional simulation studies enabled the
selection of a functional and cost effective solution to reduce the worst-case voltage sag, considering a variety of
network conditions, to acceptable levels.

3.2 Resonant Overvoltages due to Network Parallel Resonances


Description of the phenomenon
Transformer inrush currents can have a high magnitude with a significant harmonic content. The inrush currents
interact with the power system, whose resonant frequencies are a function of the series inductance (associated
with the short circuit strength of the system) and the shunt capacitances of lines and cables. This may result in
long-duration resonant temporary overvoltages. Higher inductances (relatively weak systems) and higher
capacitances (long lines and cables) yield lower resonant frequencies and a higher chance of TOV, which can have
hundreds of peaks of about the same magnitude if the TOV duration is several seconds [18].

A transmission system will generally be weak during the first steps of a system restoration following a black out.
The equivalent system inductances are then relatively high because relatively few generators are on line and the
grid tends to be sparse. Therefore, the first system resonant frequency can be much lower than for normal system
operating conditions. Large capacitances also contribute to the low resonant frequencies. One of the major
concerns during the early stages of a power system restoration is the occurrence of overvoltages as a result of
switching procedures [19]. Energizing equipment during black start conditions may result in higher overvoltages
than that during times of normal operation, particularly when system loading is light so that damping of
overvoltages is also light.

Harmonic Resonance Voltages


Harmonic resonance voltages are oscillatory undamped or weakly damped TOVs that originate from switching
operations and equipment non-linearities [19], and are a result of network resonant frequencies at multiples of the
fundamental frequency. The resonances can be excited by harmonic sources, such as saturated transformers and

Page 23
Transformer Energization in Power Systems: A Study Guide

lead to long duration overvoltages which can produce arrester failures and system faults which prolong system
restoration [19].

During a restoration phase, the voltage rise in the system due to capacitive line and cable charging currents can be
sufficient to overexcite transformers and generate significant harmonics. If the combination of the system
impedance and the line capacitance is adverse, then a harmonic resonance will result. Harmonic components in
the transformer magnetizing current during saturation will excite these resonances, which can result in damaging
overvoltages. Even if transformers are not continuously overexcited, harmonics generated by magnetizing inrush
on energization can be sufficient to excite the resonance. Several factors that are characteristic of networks during
restoration [19] contribute to harmonic overvoltages:

1) the natural frequency of the series circuit formed by the source inductance and line charging capacitance may,
under normal operating conditions, be a low multiple of the power frequency;
2) magnetizing inrush caused by energizing a transformer contains many harmonic components;
3) during early stages of restoration, lines are lightly loaded and transients therefore are lightly damped.

Harmonic resonant overvoltages may also develop when transformers are switched in high voltage cable systems
and HVDC stations [9]. Unlike overhead-line systems, cable systems generally have a pronounced resonance
point, which occurs at a relatively low frequency because of the high cable capacitance; if this resonance happens
to be near one of the harmonics produced during transformer energization, TOVs will build up. The AC filter circuits
connected at the HVDC stations introduce several parallel resonances into the impedance-frequency characteristic
of the system [20], so high TOVs may occur during transformer energization if the system also has a low degree of
damping. This is the case where generators feed the HVDC stations directly (i.e., without local AC loads being
connected), although field measurements show that not every system configuration of this kind leads to high TOVs.

Resonant Overvoltages during Transformer Energization


A harmonic analysis can be carried out by representing an inrush current as a harmonic current source I(h)
connected to the transformer bus. The relation between nodal voltages, network impedance matrix and current
injections can be then analysed by means of impedance equations [21]:

( )= ( )( ) Eq. 3-1

where h represents the harmonic frequencies (multiples of the fundamental frequency), and Z(h) is a symmetrical
matrix with as many rows and columns as the harmonic currents. V(h) and I(h) are, respectively, the vectors of
harmonic voltages and currents.

The harmonic current components of the same frequency as the system resonant frequencies are amplified in the
case of parallel resonance, thereby creating high voltages at the transformer terminals, as discussed in the
previous section. This leads to a higher level of saturation and, consequently, higher harmonic components of the
inrush current which again results in increased voltages. This can happen particularly in lightly damped systems,
common at the beginning of a restoration procedure when a path from a black start source to a large power plant is
being established and only a few loads have been restored [19, 22].

The diagram shown in Figure 3-1 is used to illustrate a condition that can lead to harmonic resonant overvoltages.
It is a very simplified representation of a power system at the early stages of a restoration procedure in which the
analysis is concentrated on the energization of a transformer that is assumed to be unloaded. The plots of Figure
3-2 show some simulation results for this case. A conclusion from an impedance versus frequency scan from the
point of connection of the transformer is that the impedance seen from this bus shows a parallel resonance at the
second harmonic. When the transformer is energized, this resonance condition results in the overvoltage depicted
in Figure 3-2c.

Page 24
Transformer Energization in Power Systems: A Study Guide

Figure 3-1 Diagram of the test system.

Impedance

0 100 200 300 400 500


Frequency ( Hz)
a)

b)

c)
Figure 3-2 Energization of an unloaded transformer. a) Impedance at transformer bus.
b) Transformer current during energization. c) Transformer terminal voltage.

The effects of the system resonance, network damping and inrush current on the harmonic overvoltages that can
occur in high-voltage cable and HVDC systems were examined in [9]. Both systems are likely to have a low
resonant frequency. For high-voltage cable systems, this results from the high charging capacitance and the fact
that the short-circuit power is still low in the initial stage of reconstruction. The frequency response of such a
system is of the single-resonance-frequency type with a quality factor of about 15. In HVDC systems, the large
capacitance of the harmonic filter circuits produces parallel resonance at a low frequency. In addition, parallel
resonances may also occur between the filter frequencies. Under conditions of resonance, the magnitude and
duration of the overvoltage is a function of the system damping. Therefore, losses in transformers, generators, filter
circuits and other system loads must be determined and specified as accurately as possible for the frequency
range under consideration.

Page 25
Transformer Energization in Power Systems: A Study Guide

The key factors for analysis of harmonic overvoltages include the resonant frequency of the network, the system
damping (including the network losses and the load connected to the network), the voltage level at the transformer
terminals, the saturation characteristic and air core inductance, the residual flux in the core of the transformer, and
the closing time of the circuit breaker pole. Factors that contribute to a higher level of resonance overvoltage are
[23]:

1) higher rating of the transformer to be energized;


2) lower value of source fault level;
3) longer circuit length;
4) smaller amount of load in the system;
5) higher than normal system operating voltage;
6) higher designed working flux density of the transformer (lower knee point);
7) position of the tap resulting in lower turns ratio.

Page 26
Transformer Energization in Power Systems: A Study Guide

CHAPTER 4 MITIGATION TECHNIQUES AND BEST PRACTICES


4.1 Mitigation of the Inrush Currents
Energization of a transformer from the network by uncontrolled and random closing of a switching device produces
inrush currents that can have undesirable or unintended consequences. As mentioned previously, these include
excessive RMS voltage dips (possibly coincident with excessive TOVs) that can be disruptive to power quality
sensitive loads, protective relay misoperation and transient and temporary overvoltages that can impose excessive
stresses on equipment. Besides, transformer inrush currents can create unusual problems such as pulsating low
frequency electromagnetic torques in generators or large motors that are remote from the energizing bus, causing
mechanical vibrations, shifting of windings and slippage of shaft couplings. The following methods have either been
proposed or are already in use to control or eliminate transformer inrush currents and prevent the above
undesirable effects, including harmonic resonant overvoltages [2, 19, 21, 23, 24, 25].

Controlling the Switching Times of the Energizing Circuit Breaker


This class of mitigation, which is probably more complex to implement in practice than some of the other
techniques, is given a somewhat longer discussion than for the others. Controlling the closing times of the
energizing device, as opposed to random three-phase closing, is an important method to significantly reduce the
inrush current. Transformer energization by segregated-pole breakers equipped with a special purpose point-on-
wave (POW) control can minimize the inrush currents and has been demonstrated to be very effective, particularly
in cases where voltage dip due to random closing can be high. The success of the POW technique requires that
the closing characteristics of the breaker be stable, repeatable, and have a relatively small pole scatter. This is
particularly true when the CB operates several times a day or hundreds of times per year, which is the case of
pumped storage power plants. Reference [26] describes the case of a 400 kV pumped storage power plant, where
such a device has been installed, reducing the peak inrush currents by a factor of 30, from 1500 A to a value of
50A, this being obtained by reducing the value of the flux circulating in the iron core during the transformer
energization.

To achieve almost complete elimination of the inrush current, the residual flux in the core must be known. The
POW control system will account for this as well as the closing characteristics of the breaker. If the residual flux is
not known or is equal to zero, it will close the first pole at its voltage peak. If the flux is known, it will close each pole
at the optimal closing point on its voltage wave. With the modern generation of POW controllers, the transformer
does not need be switched off in a particular phase sequence to establish a consistent disposition of the residual
fluxes in the core. This switch-off strategy, obviously, would not be valid when the transformer is tripped by
protection.

The first commercial installations of POW controlled breakers appeared in the 1990’s, with a few installations in
service by 1995. A wider acceptance of the technology occurred at the end of 1990’s thanks to the introduction of
effective compensation algorithms to account for the variation in the operation timing. Controlled switching
technology was first developed for the energization of capacitor banks but has since been adapted for the
energization of transformers, reactors, and for EHV transmission lines [27]. Three different synchronized switching
strategies (rapid closing, delayed closing and simultaneous closing), are extensively examined and discussed in
[28, 29]. This has been followed up with experimental work, presented in [30, 31]. Synchronized switching with
delayed closing and rapid closing outperform the proposed simultaneous closing strategy. The most common
controlled closing strategy is to close the first phase at its voltage peak and to delay the energization of the
remaining two phases by a quarter of a cycle. The rapid closing technique, with its minimal delay in the closing of
all three poles of the CB, is said to avoid possible resonance issues. Closing time scatter and residual flux
measurement uncertainties reduce the field performance of controlled switching as reported in [32]. The
incorporation of the measurement of residual fluxes in the closing strategy is reported in several publications in the
last 10 years [33, 34, 35, 36, 37] and there are now special purpose built commercially available POW switching
packages which include the calculation of residual flux based on voltage measurements from the previous de-
energization of the transformer. The state of the art of controlled switching with and without the consideration of
residual fluxes is given in [33]. This publication also discusses the practical calculation of the residual fluxes from

Page 27
Transformer Energization in Power Systems: A Study Guide

integration of the voltage waveforms and the possible influence of network disturbances. There are recent field
tests which demonstrate that a properly designed and commissioned POW controlled closing system together with
suitable segregated-pole breakers can repeatedly and consistently virtually eliminate the inrush current (see the
example in ANNEX C.5), at least with transformers having three-leg cores.

It must be noted that the commissioning of any practical and effective POW system for transformer switching
involves a number of tests, including some on-line uncontrolled or pseudo-random closings in order to determine
the parameters required for input to the controller. In the interests of obtaining the long term benefits of POW
switching, the owner/operator must therefore be prepared to accept that some tests at commissioning time may
produce high inrush current, that the effects may require mitigation as much as practical and the consequences, if
any, will have to be accepted. These commissioning tests include:

1) Off-line breaker timing tests to establish the mechanical operating speed of the circuit breaker and the
timing relationships of the main contacts with the 52a contacts,
2) Energization (uncontrolled closing) tests, when the residual flux is unknown, to calibrate the sensors at the
transformer HV capacitive bushing taps (if used),
3) Energization tests to determine the Rate of Decrease of Dielectric Strength (RDDS) based on mechanical
operating time and current conduction time of the arc,
4) (Optional) Energization tests at maximum voltage without taking into account residual flux, and,
5) Optimal Point-on-Wave Closing tests to verify performance of the complete POW system, including the
calculation of the residual flux.
When both RDDS and the residual flux are unknown, the breaker manufacturer can usually estimate the RDDS so
that the controller can be set to close the first phase near its maximum voltage. Once the RDDS parameter has
been determined, only the residual flux is to be accounted for.

For a more comprehensive description of the effectiveness of a POW switching scheme showing the energization
at voltage peak as well as Point-on-Wave closing versus uncontrolled closing, the reader is referred to ANNEX C.5.
This report summarizes the commissioning report of an actual POW scheme now in service to switch a 16 MVA
13.8/138 kV three-legged core generator transformer. Sensors were installed at the HV capacitive bushing taps
and then calibrated against the voltage signals derived from 138 kV capacitive voltage transformers (CVTs) on the
system side of the breaker. These voltages were used in the calculation of the residual flux upon switch-off. The
RDDS was measured during a series of seven pseudo-random energizations. The field measurements
demonstrated that closing the first phase at its voltage peak and delaying the closing of the other two phases by
several half-cycles resulted in up to a 50% reduction in the worst-case inrush current due to random closing. When
residual flux was accounted for, the inrush current for optimal closing was repeatedly less than about 1% of the
worst case, and virtually eliminated the inrush effect. If stress due to overvoltages is a concern then a controller
strategy based on the measurement of residual fluxes is required.

Installing Pre-insertion Resistors in Series to the Circuit Breaker Energizing the Transformer
Pre-insertion resistors require relatively large resistors to be installed in series to the main circuit breaker
interrupters and an effective reduction of inrush current is achieved only by an optimal choice of the resistance
value and the pre-insertion time. However, the addition of switching devices equipped with pre-insertion resistors in
series with the energizing breaker is probably not practical because such devices would have to be specially
ordered from the manufacturer at a high cost. Circuit breakers equipped with pre-insertion resistors are also no
longer available off-the-shelf for voltages less than 500 kV since modern EHV breakers are designed for use with
POW closing.

Page 28
Transformer Energization in Power Systems: A Study Guide

Adjusting the On-load Tap Before Energizing the Transformer


The inrush current can be reduced by selecting an on-load tap on the transformer to be energized which equals or
exceeds the power frequency voltage applied before energizing. A higher number of excited turns produce a lower
flux density in the core. The influence of the transformer tap changer on an inrush current is investigated in [38].

Reducing the System Voltage Before Energizing the Transformer


Lowering generators’ scheduled voltage leads to a proportional decrease in the pre-switching steady state voltages
and consequently reduces the voltage on the system side of the open energizing breaker. Minimizing the number
of unloaded and energized lines and setting the sending-end transformers at the lowest tap position will also assist
in reducing system voltages. While effective when voltage dip is only somewhat higher than acceptable limits, this
method will probably not be sufficient, by itself, to correct severe voltage dip problems or resonant overvoltages.

Energizing the Transformer using Air-break Disconnect Switches


Energizing the transformer using disconnect switches instead of a breaker is a technique that has been used by
some utilities to reduce the inrush currents during transformer energization. One Canadian utility used this method
for many years to energize HVDC converter transformers with, apparently, no negative consequences. A
disconnect switch, being a relatively slow device, will result in pre-striking or flash over close to the voltage peak.
However, with the arc in air, there could be multiple extinctions and re-striking of the current. High frequency
transients could be imposed on the transformer windings and this could stress the transformer insulation. Also, the
disconnect switch contact has to be rated for this duty, to prevent the contact from becoming welded closed.

De-fluxing the Transformer Core Before Energization


De-fluxing the transformer core prior to energization will reduce the magnitude of the worst-case inrush currents but
this technique, for most applications, is neither convenient nor practical.

Other Methods
Alternative methods for inrush current mitigation are frequently being proposed. A method for controlling the POW
closing instant and defining a value of residual flux using a dc coil is proposed in [39].

An inrush mitigation strategy based on a pre-insertion neutral resistor is presented in [40]. This technique is
however restricted to star grounded transformers.

Reference [41] presents a synchronized switching method for three-pole spring-driven circuit breakers with a fixed
delay between poles (mechanically staggered). Optimum energization is obtained by controlling the residual flux
with controlled de-energization.

Reference [42] proposes a method based on simultaneous-closing strategy. The method is valid for almost all
common transformer winding configurations. The optimum energization instant is calculated based on the instant of
disconnection of the circuit breaker.

Another method that is being employed by utilities especially during network restoration, involves the slow increase
of generator voltage to avoid the saturation of the transformers.

4.2 Mitigation of the Temporary Overvoltages


As the harmonic resonant overvoltages are generated by the inrush currents, the methods mitigating the latter will,
conveniently, also mitigate the former. In addition, other specific methods have been proposed in [2, 19, 21, 23, 24,
25].

Page 29
Transformer Energization in Power Systems: A Study Guide

De-tuning the Parallel Resonance in the Path from the Network to the Transformer
EMT type studies can identify the network configurations that produce a parallel resonance (as seen from the
energizing bus) at or close to undesirable harmonic frequencies that correspond to one or more of the low order
harmonic components in the inrush current. Second, third, fourth and fifth harmonics are of most concern while
seventh and higher harmonics are generally of decreasing concern. If only a few configurations of the external
network are problematic, it might be feasible to de-tune the parallel resonance prior to energization by temporarily
changing the system configuration by connecting or disconnecting elements such as unloaded lines, cables, shunt
capacitor banks or shunt reactors.

Adding as much Load as Possible Before Energizing the Transformer


This leads to a decrease in the magnitude of the impedance and, consequently, to a reduced amplification of the
injected harmonic currents. To assure that resonance is damped, sufficient load should be connected to the
underlying system at both ends of the line to damp the resonance [19].

Selecting a Low Impedance Path for Energization of the Transformer


A high source impedance can be reduced by bringing additional generators online since a higher number of
generators results in a lower overall inductance and, consequently, in a higher resonant frequency. This means
that if generators are added, the resonance peak is shifted to higher frequencies and if generators are omitted, it is
shifted to lower frequencies.

Applying Surge Arresters to Limit Harmonic Overvoltages


Surge arresters are designed to protect transformers, switchgear and other expensive pieces of equipment
connected to the network from damage due to lightning strikes and switching surges. They have only a limited
capability to absorb energy during temporary overvoltages (see Section 5.2). Their maximum continuous operating
voltage (MCOV) must be safely above the maximum steady state operating voltage (including during
contingencies) at their bus. This means that they will be ineffective for TOVs up to about 1.8 pu, which is quite high.
For more severe harmonic overvoltages, surge arrester failure due to thermal runaway becomes possible. Hence,
surge arresters are generally not a good option to control harmonic overvoltages excited by transformer
energization.

Page 30
Transformer Energization in Power Systems: A Study Guide

CHAPTER 5 SIMULATING TRANSFORMER ENERGIZATION TRANSIENTS


Transformer energization simulation studies provide a significant insight into the phenomenon and are quite
desirable due to several interrelated reasons. As energization transients are heavily dependent upon initial
conditions, field measurements may not always provide enough information to assess the potential consequences
of an energization transient. Due to operational restrictions, it may not always be possible to perform
measurements for various system operating configurations and system contingencies. In extreme cases,
measurement conditions may actually be such that the resultant transient exceeds the equipment dielectric
withstand capability and thereby leading to equipment damage under field test conditions. Therefore, in order to
acquire as much information as possible for a particular set of initial conditions, system configurations as well as
system contingencies without the risk of damaging any equipment, it is desirable to perform simulation studies.
Furthermore, at the network planning stage, it is not possible to perform field tests. Simulation studies are the only
available tool to predict the likely effect of transients. They are equally applicable when choices about several
mitigation measures are available for implementation. The best and most cost effective solution can be identified by
way of simulation studies, accounting for the effect of various system conditions and parameters.

However, not every possible transformer energization configuration needs to be studied as this is not practical in
most cases and most of the system configurations are such that the energization of the transformer will not cause
any problem. Hence, the practicing engineer needs to decide which system and transformer configurations are the
most critical and require studying. It is difficult to provide a general and reliable rule of thumb for this, and the
decision largely relies on the experience of the engineer and his/her knowledge of the system.

In general, two cases need to be considered: RMS-voltage drop and resonant overvoltages. For the RMS-voltage
drop, a rough estimate of the drop can be computed from the short-circuit impedance of the supply system and the
inrush current analytically calculated (see ANNEX A). The higher the short-circuit impedance and the higher the
rated power of the transformer (for a given rated voltage), the higher the RMS-voltage drop. It should be noted that,
if other similar transformers are connected nearby, the user might be able to assume that the voltage drop will be
similar based on operational experience. The prediction of temporary resonant overvoltages is more complex (the
reader is referred to the key factors in Section 3.2). As a general approach, the frequency response of the feeding
network and especially the frequencies and amplitudes of the parallel resonances (i. e., presenting maximum
impedance values) should be considered. If any of the resonant frequencies are below 600-800 Hz, then resonant
overvoltages may be generated, with a higher probability for large resonance impedances. For instance, the
system in Section 5.3.4 is potentially dangerous, as its first resonant frequency is close to the 4th harmonic. As a
general rule, for a given voltage rating, the higher the rated power of the transformer, the higher the inrush currents
and the higher the risk of resonant overvoltages. Note that the short-circuit power of the feeding system at a
particular location does not provide sufficient information to characterize the frequency response of the system at
that location. Indeed, the resonant frequencies and amplitudes of the feeding network depend on the short-circuit
power, the degree of meshing and the amount of connected load [43]. This means that, in general, knowing the
short-circuit power of the feeding network is not enough to assess the risk of TOV when energizing a power
transformer.

5.1 Modelling of the Network for EMT Studies


General guidelines to be used for representing system components are those recommended for analysis of low-
frequency or slow-front transients [44, 45, 46]. The frequency of interest for transformer energization studies
ranges from DC up to 1kHz [44].

5.1.1 Study Zone


A common practice in electromagnetic transient studies when dealing with large systems is to divide the system
into a study zone, where transient phenomena occur, and an external system that encompasses the rest of the
system.

Page 31
Transformer Energization in Power Systems: A Study Guide

In the study zone the system must be modelled in detail. Besides the transformer to be energized and the circuit
breaker energizing it, the study zone includes the equipment in the same or neighbouring substations that may
interact with the inrush harmonic currents and associated voltages. This equipment includes other transformers
(sympathetic interaction, see Section 2.3), reactors and capacitors, synchronous compensators, generators and
voltage regulators, loads, and the lines or cables between them [46].

For the rest of the system, a network equivalent may be used in order to reduce the time needed to build the model
and the computational burden. The network equivalent must reproduce the frequency response of the network in
the range of the inrush currents frequencies [46, 44], especially when simulating resonant overvoltages (it is less
critical when simulating RMS-voltage drop). For network equivalent modelling, see Section 5.1.3.4.

5.1.2 Transformers

5.1.2.1 Physical Aspects to be Represented


The transformer being energized has to be represented in sufficient detail so that it generates the inrush currents
that are the primary source of current harmonics injected into the system. Other transformers in the study zone
have to be modeled accurately but do not require the same level of detail unless sympathetic inrush is foreseen. An
exact representation of transformers is more complex than any other network component required for energization
studies. According to [44, 45, 46, 47], when studying transformer energization and resonance phenomena, the
following aspects need to be represented (in order of importance):

leakage impedance and winding resistance


nonlinear saturation and core losses
air-core inductance
magnetic phase coupling
residual flux
frequency dependent winding losses
zero sequence impedance
hysteresis and frequency dependent iron losses
capacitances.

Leakage Impedance and Winding Resistance


The leakage inductance (from the short-circuit test), and the winding resistance define the series impedance of the
transformer. While the winding resistance is specific for a winding, a particular winding does not have a leakage
inductance; the leakage inductance is defined from one winding to another.

Nonlinear Saturation and Core Losses


Transformer saturation is an important component of many low frequency electromagnetic transient phenomena,
including ferroresonance, temporary overvoltages, and transformer inrush. For most phenomena, the critical
transformer saturation parameters are the final linear slope of the saturation curve (air-core inductance) and the
value of the saturation flux (the point where the final slope intersects the zero current axis). Other details of the
transformer saturation curve (i.e. the shape in the knee area) are of secondary importance for the study of the
worst case scenarios. The location of the core representation in the transformer model topology is also important.
Core losses can also be critically important in some low-current phenomena involving saturation, such as
ferroresonance and temporary overvoltages.

The transformer core has an intrinsic nonlinear behaviour due to the saturation of the ferromagnetic material. This
nonlinear characteristic is the primary source of harmonic current injection from a transformer and needs to be
modelled accurately. Core losses are inherent in the hysteretic behaviour of ferromagnetic material.

Page 32
Transformer Energization in Power Systems: A Study Guide

Air-core Inductance
The air-core inductance of a coil is the self-inductance of the coil without the iron-core. It defines the inductance of
a winding when the core is completely saturated. It is important to extrapolate the nonlinear saturation curve for
studies that involve core saturation such as energization transients.

Magnetic Phase Coupling


In three-phase transformers with three-legged or five-legged stacked core and shell core, the coupling between
phases is provided through the magnetic core. Therefore, the representation of the core topology is important in
three-phase transformers for simulations where phase unbalance, zero sequence, and magnetic phase coupling
have to be taken into account.

Simple, single-phase equivalent models cannot represent the core topology. In order to respect to a certain degree
the topology of the core, it is suggested to connect the nonlinear inductance in delta for three-phase core-type
transformers, and in wye for three-phase shell transformers and banks of single-phase transformers. The coupling
of the core equivalent representation does not need to agree with the coupling of the winding it is attached to. Care
should be taken to ensure that the saturation characteristic is rescaled if the winding and core coupling are
different.

An advanced three-phase transformer modelling approach is presented in Section 5.1.2.4.

Residual Flux
The residual flux value is a fundamental parameter during the re-energization of a transformer since it affects the
first peak of the inrush current. Due to the flat nature of the saturation curve above the knee point, a small increase
of flux peak (residual flux) can drive the iron core of the transformer into heavy saturation. The residual flux is
created when a transformer is disconnected from the power grid. At the end of a de-energization transient both the
voltages and currents decrease to zero. However, the flux in the core retains a certain value defined as the residual
flux. A ringdown transient is a natural LC response that appears as the stored energy dissipates whenever a
transformer is de-energized [48].

Due to the ringdown transient, the residual flux is always somewhat lower than the theoretical maximum residual
flux equal to the intersection of the applicable hysteresis curve with the zero current axes. The decrease of the
maximum residual flux involves the resonant circuit formed by the main inductance of the transformer and the stray
capacitances of the circuit breaker which opens. The magnetic energy stored in the iron core is then partly
transformed into resistive losses in the circuit.

Frequency-dependent Winding Losses


A frequency-dependent winding resistance model should be used if the representation of the damping of transients
and resonance conditions is important. Series winding losses are frequency-dependent due to eddy currents
induced in the winding and in other transformer components.

Zero-sequence Impedance
The zero-sequence impedance of a transformer is determined by its delta windings and core configuration. While
the delta winding provides linear zero-sequence impedance, the core usually involves nonlinearities. In general, it
is not considered accurate to assume that the presence of a delta winding is sufficient to represent inter-winding
coupling in a transformer. Both winding coupling and core topology should be represented to obtain an accurate
representation of the zero-sequence and unbalanced operation behaviour.

Hysteresis and Frequency-dependent Iron Losses


The main limitations of a piecewise RL core representation (with linear R and piecewise-nonlinear L) are:

Page 33
Transformer Energization in Power Systems: A Study Guide

It cannot represent the nonlinear core losses due to the linear R. Core losses are accurate only at rated
excitation. Losses are overestimated at lower than rated excitation and underestimated at higher excitation.
It cannot retain residual flux after de-energization; residual flux is always zero.
It cannot represent the frequency-dependent and frequency-independent components of the core losses.

Hysteresis and frequency-dependent iron losses should be represented to avoid these limitations.

Stray Capacitances
Capacitances are only of marginal importance in the study of low-frequency phenomena which includes
transformer energization. However, for energization studies, their representation is important if de-energization
transients have to be simulated to calculate residual fluxes. In addition, they play an important role in the study of
ferroresonance . They also influence the measurement of the transformer excitation characteristic (no-load test).

5.1.2.2 Available Models


In its simplest form, a transformer is represented by means of a single-phase equivalent circuit. Such
representation is adequate for single-phase and a bank of single-phase transformers. However, since it lacks a
suitable representation of the magnetic coupling of the phases in the core, it can only give approximate results for
single-core three-phase transformers.

This section does not cover all transformer models, but only the two most frequently used models since they are
implemented in most electromagnetic transient analysis programs. The first model is an extension of the Steinmetz
model to multi-phase transformers. The second model uses either a branch impedance or admittance matrix. Both
types of models present important limitations for simulating some core designs.

Saturable Transformer Component (STC Model)


The Saturable Transformer Component (STC) [49, 50, 51] is a two- and three-winding single phase transformer
model. It is the nonlinear version of the classic Steinmetz model [52]. This model is based on the star-circuit
representation shown in Figure 5-1. It represents the short-circuit impedances between windings, the load and
magnetization losses, and the nonlinear inductive magnetization. The primary branch is treated as an uncoupled R-
L branch, each of the other windings being handled as a two-winding transformer. Saturation and hysteresis effects
are modelled by adding an extra nonlinear inductor at the star point. If the transformer has two windings, the core
representation can be connected to the terminal of the winding closest to the core (normally the LV winding). If the
transformer has three windings, attaching the core to the winding closest to the core is mandatory, as the voltage at
the star point has no relationship with the core flux [53]. Moreover, this also prevents numerical instability.

BCTRAN (see below) can also be used with single-phase transformers, with the advantage of no winding number
limitation. The simulation results should be the same as those provided by STC.

For two winding transformers, the winding resistance and leakage inductance are artificially split between primary
and secondary winding. A 50% splitting factor is often used. However, it should be more accurate to put 75%-90%
of the leakage inductance on the HV side and the rest on the LV side [54], as the leakage inductance between the
HV winding and the core is normally larger than the one between the LV winding and the core. In energization
studies this results in a larger inrush current in p.u. on the LV side than on the HV side as observed in practice.

Page 34
Transformer Energization in Power Systems: A Study Guide

Figure 5-1 Star-circuit representation of single-phase three-winding transformers. Alternative location of the core
representation in grey.

The STC model can be extended to three-phase units through the addition of a zero-sequence reluctance
parameter, but its benefit is limited. In addition, when it is used to model three-phase transformers, the inter-phase
magnetic coupling is not represented in this model therefore introducing great limitation in studying unbalanced
operation.

The input data consist of the R-L values of each star branch, the turns ratios, and the data for defining the
magnetizing branch. This model has some important limitations: it cannot be used for more than three windings,
since the star circuit is not valid for N>3, the magnetizing inductance Lm with resistance Rm, in parallel, is connected
to the star point, which is not always the correct topological connecting point, and numerical instability has been
reported for the three-winding case, although this problem has been identified as the use of a negative value for
one short-circuit reactance [55, 56].

Matrix Representation (BCTRAN Model)


The BCTRAN model [57] is an n-phase transformer model where inter-winding coupling can be taken into account.
The model is linear and assumes phase symmetry. It consists of a coupled RL or RL 1 matrix representing short-
circuit impedances between windings, load losses at rated frequency and optionally linear inductive magnetization.

Phase-to-phase couplings, as well as the terminal characteristics, are included with these approaches, but they are
linearised and do not consider differences in core or winding topology, since all core designs get the same
mathematical treatment.

This model is linear; however, for many transient studies it is necessary to include saturation and hysteresis effects.
Exciting current effects can be linearised and left in the matrix description, which can lead to simulation errors when
the core saturates. Alternately, excitation may be omitted from the matrix description and attached externally to the
terminals of the modelled transformer in the form of non-linear elements as shown in Figure 5-2. Such an externally
attached core is not always topologically correct, but good enough in many cases.

When the core equivalent representation is connected at the terminal of the innermost winding (normally the LV
winding), it has to be connected before the winding resistance, as shown in the figure below. This is required to
correctly take into account the damping of inrush current provided only by the winding resistance of the energized
winding.

Although these models are theoretically valid only for the frequency at which the nameplate data was obtained,
they are reasonably accurate for frequencies below 1 kHz.

Page 35
Transformer Energization in Power Systems: A Study Guide

BCTRAN
RLV
(RHV & L-1)

CORE
EQUIVALENT

Figure 5-2 BCTRAN-based model for two-winding transformers with an externally attached core.

5.1.2.3 Practical Modelling


Leakage Impedance and Winding Resistance
Leakage impedances and winding resistances can be derived from standard short-circuit tests. Information usually
available from short-circuit test reports are the short-circuit voltage in per-cent (v sc%) and the active power (Psc) at
rated current for each winding pair (short circuit test: one winding energized, another short-circuited and all other
windings left open). The short-circuit winding resistance and leakage inductance for a winding-pair based on phase
quantities are:
2
= 2
Eq. 5-1

2 2 2 2
%
= Eq. 5-2
2 2
100

with k = 1 for single-phase and three-phase wye-connected transformers, and k = 3 for three-phase delta-
connected transformers.

Alternatively, the leakage impedance can be analytically estimated from an ampere-turn diagram (ATD) based on
geometrical winding dimensions [58].

For a two winding transformer, the inductance representation is rather simple. Different modelling approaches are
available for the representation of the leakage inductance of multi-winding transformers, with the most common
being the star-equivalent representation (STC) and the fully coupled impedance or admittance formulation
(BCTRAN).

Winding losses are modelled with linear resistors to match the measured power frequency short circuit losses. For
two winding transformer a total equivalent resistance ( ) is calculated from the short circuit losses that has to be
divided between low- and high-voltage windings. It is common practice to equally divide the resistance between the
windings on a p.u. base. When the DC resistance value is known, it may be preferable to use it in order to achieve
higher accuracy based on the following allocation [59]:

= Eq. 5-3
+

Page 36
Transformer Energization in Power Systems: A Study Guide

= Eq. 5-4
+

where and are the p.u. AC resistances, the p.u. total short-circuit resistance and and
the p.u. DC resistances. The splitting of the AC resistance based on DC resistance measurements usually results
in < on a per unit base.

For three winding transformers the winding resistance can be obtained by solving the following system of
equations:

12 110 1
23 = 011 2 Eq. 5-5
13 101 3

where RSC12 is the short circuit resistance between windings 1 and 2 and so on and R1, R2 and R3 are the winding
resistances. For more than three-winding transformers the above system of equations become over defined as the
( )
number of short-circuit test is , larger than the number of windings . A subset of the short-circuit test can
be used for the calculation of the winding resistance and the remaining values used for verification.

An autotransformer is modelled as a normal transformer where the primary and secondary windings are connected
in series to form the common and series windings. A tertiary delta winding is normally present in three-phase
autotransformers. The model should represent the actual series, common, and delta coils as shown in Figure 5-3
and not the three-winding "black box" equivalent typically assumed. For this reason, autotransformer coil
reactances should be calculated with care from the short-circuit factory test report [51, 60].

Figure 5-3 Schematic representation of an autotransformer model with series and common connected in series.

With reference to Figure 5-3, defining the voltages as:

= +
= Eq. 5-6
=

the series-to-common impedance value in p.u. is given by:

+
= = Eq. 5-7

No modification is needed for the common-to-delta impedance and its value in p.u. is simply:

Page 37
Transformer Energization in Power Systems: A Study Guide

= Eq. 5-8

Finally, the series-to-delta impedance value in p.u. can be calculated by:

= + Eq. 5-9
( )
The autotransformer can therefore be treated as a transformer with three windings designated as S,C and D and
with short-circuit impedances ZSC, ZCD, and ZSD.

Nonlinear Saturation and Core Losses


One of the most critical steps in the creation of a transformer model for energization transient calculation is the
construction of the saturation curve accounting for the nonlinear behaviour of the core. The primary source of data
available for the calculation of nonlinear saturation and core losses is the standard no-load test. Information usually
available from a standard no-load test is the no-load current in per-cent and the no-load losses at different
excitation voltages. Typical values are limited to 100% and 110% of the rated voltage. It is becoming more common
to perform the no-load test for an extended voltage range (between 90% and 120%). This is desirable if higher
accuracy in the modelling and calculation of the saturation curve is required.

The simplest approach for modelling saturation and core losses is with a parallel R-L representation as shown in
Figure 5-4 where L is a piecewise non-linear inductance and R is a linear resistance.

L R

Figure 5-4 Parallel R-L circuit for modelling core saturation.

The value of the resistance can be calculated from the core losses at rated excitation as:
2
= Eq. 5-10

with k=1 for single-phase and three-phase wye-connected transformers, and k=3 for three-phase delta-connected
transformers.

The value of the inductance can be calculated from the magnetization current that is the no-load current from the
test report by subtracting its loss component as:

= 2 2 Eq. 5-11

Once the magnetization current is calculated for all available no-load test values, the piecewise nonlinear L can be
calculated. The process involves a conversion of the RMS values of the magnetization current and voltage to the
peak values of current and flux-linkage. Since L is nonlinear, the points of the saturation curve cannot be calculated
independently from the preceding ones, but sequentially using an analytical [61, 62] or numerical procedure. This
calculation is based on the assumption of a sinusoidal applied voltage. In order to enhance the accuracy of the
saturation curve calculation, one should take into account the triplen harmonic correction in transformers tested
from the delta winding [62] and capacitive correction for large transformers [54, 63].

Page 38
Transformer Energization in Power Systems: A Study Guide

It should be noted that instantaneous flux versus excitation current information can also be obtained from
manufacturers on request.

Air-core Inductance
When the value of the air-core inductance is known, an additional point in the nonlinear saturation curve should be
defined beyond the last calculated value from test report data to set the saturated core inductance according to the
air-core inductance. The saturated inductance seen from the low-voltage winding can be established by :

= , Eq. 5-12

where Lair core,HV is the air-core inductance calculated for the HV winding, LHL is the leakage inductance between HV
and LV windings and Lsat is the final differential magnetization inductance. The inrush current peak value is highly
sensitive to the air-core inductance value; therefore, special care should be taken to consider this parameter. The
value of air-core inductance can normally be provided by a transformer manufacturer or estimated from design data
[58]. When the core is connected to the star point, the saturated inductance is calculated by removing only a
fraction of the total leakage inductance LHL, corresponding to the fraction of leakage placed on the primary side.

A suggested value of the air-core reactance seen from the HV side is 0.3 p.u. [47]. For large, high BIL
transformers, the typical range is 0.3-0.8 p.u. on the self-cooled (ONAN) base and for distribution transformers [54]
it is around 0.05 to 0.15 p.u. Typical values given in [44] for air-core inductance of the HV winding (outer winding) is
2-2.5 LHL for a step-down transformer and 4-4.5.LHL for autotransformers.

Detailed formulas for the calculation of air-core reactance taking into account the exact geometry and the
construction details of the transformers are presented in [64, 65].

Residual Flux
For residual flux potential values, see Section 5.3.2.

Frequency-dependent Winding Losses


A simple approach for representing the frequency dependence of winding resistance is with foster networks. It is
however critical to obtain the characteristic of the winding resistance as a function of the frequency required as
input data for the modelling. The frequency dependency can be estimated analytically but requires a detailed
knowledge of the winding configuration and is quite complex [66, 67].

Taking into account the frequency dependency of the winding resistance will provide higher damping for higher
frequency components as well as reduced DC losses. For practical purposes when frequency dependency is
neglected it is safe to say that the duration if not the amplitude of the resulting voltage is overestimated.

Zero-sequence Impedance
In three-legged transformers, the zero-sequence impedance is given by the zero-sequence flux closing through the
air and the tank. The impedance representing this flux is approximately linear and should be added at the HV
terminals of the transformer model [51, 53]. The impedance is calculated from the zero sequence magnetizing
current, which should be measured during the transformer tests.

However, if the transformer has a delta winding, the zero-sequence flux will circulate in the air between the
windings and, as a consequence, the zero-sequence impedance will be similar to the positive-sequence short-
circuit impedance. In this case, no specific handling of the zero-sequence impedance is necessary.

In four and five-legged transformers and shell-type transformers, the zero-sequence impedance is given by the
zero-sequence flux circulating in the unwound legs of the core. Because of the relative dimensions of the core
being designed for positive-sequence excitation, higher saturation of the unwound versus the wound legs is likely
to occur during zero-sequence excitation. However, since much of the zero-sequence flux will still circulate in the

Page 39
Transformer Energization in Power Systems: A Study Guide

core, one can assume that the zero-sequence impedance is similar to the positive-sequence. Therefore, no specific
handling of the zero-sequence impedance is necessary in this case [51].

In general, topological models are needed for a better account of the zero-sequence behaviour of the transformer
(see Section 5.1.2.4).

Hysteresis and Frequency-dependent Iron Losses


The core losses can be divided in three components: hysteresis, classic eddy current and excess (or anomalous)
eddy current:

0 = + + Eq. 5-13

To achieve a more accurate representation, each component should be modelled separately; however, this is not
straightforward. The first difficulty is how to split the total core losses measured at power frequency into the three
components. For example, at 1.7 T each loss component is between 30% and 40% of the total core losses.

Classical and excessive eddy currents are approximately proportional to the square of the flux and frequency and
can be modelled by a linear resistor based on the following equations:
2 2
Eq. 5-14

2
Eq. 5-15

The representation of hysteresis losses is more complex as they are frequency independent and nonlinear. The
hysteresis loop of the magnetic material needs to be represented to account for this loss component. Nonlinear
resistors could be used to more accurately match the nonlinear loss behaviour; however, their transient
performance is somewhat uncertain. A non-linear resistance can be used to represent the excitation-dependent
hysteresis core losses, however this representation estimates the average loss per cycle, but fails to accurately
reproduce the exact waveforms and hysteresis loops as shown in Figure 5-5. In addition, this model cannot
represent the rate-independent property of the hysteresis losses.

Figure 5-5 Flux-current trajectories with a nonlinear resistor model for core-loss representation. [54]

Page 40
Transformer Energization in Power Systems: A Study Guide

Advanced models for the representation of hysteresis are Preisach and Jiles-Atherton models. While these models
may be accurate, they are complex and it is difficult to obtain model parameters. Their use is justified only in
special situations.

A commonly used model that can provide a simple representation of the hysteresis loop is a piecewise nonlinear
hysteretic inductor (also known as “type 96” hysteretic inductor). While this model received some criticism because
it cannot represent minor loops and sub-loops accurately, it has the advantage that it can be built from standard
test report data. One part of the losses is attributed to eddy current losses and is included in a parallel resistor. The
other part is assigned to hysteresis losses and is included in the type 96 component. The hysteresis losses PH are
further scaled to the maximum excitation with a Steinmetz coefficient of 2 as shown in Eq. 5-16 which seems
reasonable according to [68].
2
= Eq. 5-16

where Physt is the hysteresis part of the total core losses at rated excitation (at rated).and max is the final point of the
hysteresis loop as shown in the figure below.

The width of the hysteresis loop is assumed to be constant (but goes to zero at the second highest flux-linkage
point) and calculated as [69]:

= [ ] Eq. 5-17
( 1 + 2)

where f is the power frequency and the flux-linkages np-1 and np-2 are given in the piece-wise nonlinear curve in
the figure below and their sum multiplied by the width W becomes the area of the hysteresis loop.

Figure 5-6 Construction of the hysteresis curve in the type-96 component. [69]

In this approach the maximum residual flux is determined by the intersection of the hysteresis loop with the zero-
current axis. One should be careful to verify that the resulting maximum residual flux has a reasonable value in the
range of 0.6-0.8 p.u. and that the measured saturation, when plotted on the hysteresis characteristic lies within the
loop.

Capacitances
Source of capacitances in a transformer are parasitic winding and bushing capacitances and includes:

Capacitance between a winding to the core or ground including the bushing capacitance: CLG and CHG

Page 41
Transformer Energization in Power Systems: A Study Guide

Capacitance between two concentric windings of the same phase: CHL


Capacitance between two phases (only for the HV or outermost windings): CPh

For low-frequency transients, they can be modelled as lumped parameters with two capacitive networks as the one
shown in Figure 5-7 each connected to one end of the windings.

Figure 5-7 Capacitive network connected to one end of the windings. [59]

Data for the modelling of capacitance are normally included in the test report or can be obtained from a system of
simple measurements at the terminals of the transformer. However, CPh is not normally available and can be
neglected without any significant impact on the overall results.

5.1.2.4 Advanced Modelling: Topology-based Models


Topology-based models address the limitations of standard models and many approaches have been proposed
[50]. Their derivation is performed from the core topology and can represent very accurately any type of core
design in low- and mid-frequency transients if parameters are properly determined.

The main characteristics are:

Topologically correct core model (triplex, three-legged, five-legged, shell).


Effects of saturation in each individual leg of the core accounted for.
Interphase magnetic coupling accounted for.

Typical drawbacks of advanced models are the limited availability of input data and the difficulty in estimating the
model parameters. Many models have been proposed but almost none have become widely adopted. Only two
model are known today to have been partially implemented in commercially available tools (UMEC [70, 71] and
Hybrid Transformer [69, 72, 73, 74, 75] models).

The Hybrid Transformer model is an engineering transformer model based on limited input data. The modelling of
the transformer is based on the magnetic circuit transformed to its electric dual. The leakage and main fluxes are
then separated into a core model for the main flux and an inverse inductance matrix for the leakage flux. The
copper losses and coil capacitances are added at the terminals of the transformer. In the current implementation
the model can represent two- and three-winding three-phase transformers. Three-limb, ve-limb, shell and three-
phase bank (triplex) transformer core constructions are possible con gurations in the model where only relative
dimensions are required. Input data can be specified as typical values, from test report and design information.

Page 42
Transformer Energization in Power Systems: A Study Guide

Figure 5-8 The Hybrid transformer model: electric-equivalent network representation. [69]

The UMEC transformer model is based on the concept of a uni ed magnetic equivalent circuit. A normalized core is
used in order to remove the requirement of design data; only relative dimensions are required. The magnetic
network is derived from the transformer core topology. Three-limb, ve-limb, and three-phase bank (triplex)
transformer core constructions are possible con gurations in the model. The magnetic network representing the
core and leakage inductances is described with a matrix formulation using a permeance matrix (in the current
implementation the number of windings are limited to two for three-phase transformers). Both the magnetic
coupling between windings of different phases and the coupling between windings of the same phase are taken
into account. Winding and core losses are not included in the magnetic circuit and are represented by an
equivalent linear resistance at the winding terminals. Load losses are equally divided on a p.u. base and
represented by linear series resistances connected at one terminal of each winding. Core losses are assumed
linear and equally divided on a p.u. base between primary and secondary windings. The use of a magnetic network
does not allow a simple representation of topological core losses, therefore they are represented by linear shunt
resistances connected at the terminals of each winding.

Page 43
Transformer Energization in Power Systems: A Study Guide

Figure 5-9 The UMEC model: magnetic network [71].


These two topology-based models are compared with a standard model for the study of a black start energization
[76]. The study concludes that a topologically correct core model produces higher accuracy results when simulating
highly nonlinear and unbalanced electromagnetic transients.

Table 5-1 provides a concise summary of the discussed transformer models.

Page 44
Transformer Energization in Power Systems: A Study Guide

Table 5-1 Representation of transformers

STC BCTRAN UMEC Hybrid Transformer


(Star-equivalent or T (Pi-equivalent or (Topological model) (Topological model)
model) matrix
representation)

Simplified Figure 5-1 Figure 5-2 Figure 5-9 Figure 5-8


Representation

Short circuit Requires the artificial Matrix formulation Matrix formulation Based on the
splitting of leakage based on inverse using a permeance BCTRAN approach.
inductance inductance matrix
formulation

# windings 3 (stable with 2, No limitation No limitation, limited No limitation, limited


possible instability to two windings in to three windings in
with 3 due to current current
negative inductance). implementation implementation

Core At the star point or At the terminals of N/A: No separation Fictitious winding at
connection terminals of the the innermost between core and the core surface is
innermost winding winding leakage introduced to create a
topological
connection point for
the core

Core modelling Equivalent single- Equivalent single- A magnetic network Topological model
phase representation. phase representation. represents both the representing the core
No magnetic coupling No magnetic coupling core and leakage construction and
inductances. Doubtful magnetic coupling
handling of open- between phases
circuit data

5.1.3 Other Components of the System

5.1.3.1 Shunt Reactors


Reactors can be represented similar to transformers but without the leakage impedance. In a shunt reactor model
there is only a single inductance instead of leakage and magnetizing inductances of multi-winding transformers.

Air-gaps are normally used in shunt reactors to define the total inductance. The equivalent per-phase effect of the
air-gaps should be represented by a linear inductance in series to the magnetizing inductance as shown in Figure
5-10. The resistance for representing the core losses should be placed in parallel to only the magnetizing
inductance.

The rules for modelling saturation effects and frequency dependent losses for transformers apply also to shunt
reactors. Consideration should be given to the fact that for three-phase reactors the zero-sequence impedance is
generally lower than the positive-sequence impedance.

Page 45
Transformer Energization in Power Systems: A Study Guide

L air-gap

Lm Rm

Figure 5-10 Simplified shunt reactor representation.

5.1.3.2 Overhead Lines


According to the IEC 71-4 [45] and CIGRE TB 39 [44] recommendations for the modelling of temporary
overvoltages and the CIGRE WG 33.10 [47] recommendations for the modelling of transformer energization, the
important phenomena to be represented in overhead-line (OHL) models are:

The line asymmetry (especially in untransposed lines) and the coupling with other circuits.
The frequency dependency of the ground mode.

Line Asymmetry and Coupling with other Circuits


The line asymmetry is given by the different distances between the conductors and by the different distances
between the conductors and the ground. The coupling with other circuits is present when several circuits are in the
same right of way.

The modelling of the line asymmetry is important to accurately represent the resonant frequencies and impedances
for each phase, which may be excited by the harmonic components of the inrush currents. This is especially
important for untransposed lines where both resonant frequencies and impedances may be quite different for
different phases. Transposition tends to reduce the asymmetry effect. However, if no resonant harmonic
overvoltage is expected (for instance, if the total line length is short), the line asymmetry may be neglected.

It should be noted that if the line asymmetry must be represented, the sequence impedances of the lines used in
steady state or power-flow studies are not adequate, as they assume the line is symmetric. Instead, the line model
parameters must be calculated from the actual geometrical coordinates of the conductors.

Frequency Dependency
According to [44, 45, 47, 77], the range of variation of the line parameters in the aerial mode (i.e., positive-
sequence if line symmetry is assumed) is negligible in the TOV frequency range, whereas it may be important in
the ground mode (zero-sequence). Indeed, positive sequence inductance and resistance are fairly constant up to
1 kHz, but zero-sequence values are very dependent on frequency because of the skin effects in the earth return
mode [51, 77].

The variation of the line parameters with frequency is important because it will affect the damping of the harmonic
overvoltage components. However, if no harmonic overvoltage is expected, frequency dependency may be
neglected (reference [78] shows no difference in the inrush current damping whether the frequency dependency of
the line parameters is represented or not).

To illustrate the effect of the line asymmetry and the frequency dependency of the line parameters, consider the
harmonic impedance of the network shown in Figure 5-11, where an unloaded 3x550 MVA 400/20 kV step-up
transformer is energized at the end of an 84 km double-circuit untransposed overhead line fed by the 400 kV
network.

Page 46
Transformer Energization in Power Systems: A Study Guide

a a a a ?i 1 2
RL1 b
+ b b + FD b +
c c c c
405/20

+
+

L1
25Ohm

+
+
+

Network equivalent, 400 kV OHL, double-circuit, untransposed 3x550 MVA (triplex)

Figure 5-11 Test circuit #1: Transmission network, OHL and transformer.

Figure 5-12 shows the network impedance for a positive (left) and a zero-sequence (right) current injection at the
transformer location. In the upper figures, the line parameters are frequency dependent while in the bottom ones,
they are constant and calculated at power frequency (50 Hz).

These curves illustrate the effect of the line asymmetry. Irrespective of the figure considered, the resonant
frequencies and impedances are different for the three phases. The effect of the frequency dependency is also
clearly shown. In this frequency range (below 1 kHz), the impact of the frequency dependency is moderate in the
positive-sequence (figures on the left) but it is very important in the zero sequence (figures on the right).

Phase A
Phase B
Phase C

Figure 5-12 Network impedance: Exact-PI (top) compared to cascaded-PI (bottom) for positive (left) and zero-
sequence (right) excitations.

Other Aspects: Corona Effect, Surge Arresters


The corona effect involves the ionization of the air surrounding the line conductors. Its consequences are an
increase of the losses and capacitance of the line [45, 79, 80]. The increase in capacitance is important for
travelling wave phenomena (switching or lightning studies), as it reduces the steepness of the surges. For
temporary overvoltages, the increase of the losses will result in additional damping. The corona effect occurs if the

Page 47
Transformer Energization in Power Systems: A Study Guide

corona inception voltage is reached, which is usually about 30% to 40% higher than the line rated voltage [80] (yet,
in systems with high rated voltage, the difference with the corona inception voltage may be smaller). However, as it
is very difficult to model, the corona effect is usually neglected (therefore the modelling will tend to produce
pessimistic results) [47, 81].

If the overvoltages are high enough, the surge arresters will absorb part of the energy, thus contributing to
damping. Temporary overvoltages release large energy, however, the surge arresters energy absorption capability
is relatively small and hence the principal concern is whether or not they fail during the energization transient [47]
(see Section 5.2).

Practical Modelling
In the frequency domain, a line is modelled by a lumped RLC multiphase representation, i.e., a PI-circuit, and it is
easy to represent the line asymmetry, the coupling with other circuits and the frequency dependency of the
parameters. The line asymmetry and the coupling with other circuits are taken into account if the model parameters
are calculated from the geometrical coordinates of the conductors. The frequency dependency is modelled by
recalculating the PI-circuit parameters at each frequency –the model is then called “Exact-PI” (or “PI-exact”).

For the time domain, Table 5-2 shows the standard modelling possibilities depending on the model and the input
data. Three standard models are considered in this brochure: cascaded-PI, Bergeron and J. Marti’s (other
frequency dependent models are available). Available input data may be just sequence characteristics of the line,
i.e., series impedance and shunt capacitance in positive and zero-sequence, or the geometrical coordinates of the
conductors and their resistivity and diameter.

Table 5-2 Standard OHL representations

Line asymmetry
No Yes
PI/sequence PI/conductor
Frequency No
Bergeron/sequence Bergeron/conductor
dependency
Yes LMarti/conductor

If no frequency dependency is to be modelled, a cascaded-PI representation may be used, where several PI-
circuits are connected in series to represent the total line length. These PI-circuits are calculated at a fixed
frequency (see below, model frequency)2. The maximum line length (lmax) to be modelled by a single PI-circuit
depends on the maximum frequency (fmax) of the simulated phenomena. Reference [44] recommends this
expression relating both parameters:

= Eq. 5-18
5

where is the propagation speed of the electromagnetic wave ( 300 km/ms). Considering that inrush harmonics
are negligible beyond 1 kHz, each PI-circuit can account for up to 60 km of line length.

2
As already mentioned, in the frequency domain EMT programs recalculate the parameters of the PI-circuits at each
frequency, so the frequency dependency is taken into account. In the time domain (where the signals contain more than one
frequency), however, the parameters of the PI-circuits are kept constant and thus no frequency dependency is modeled. In
order to distinguish between these two usages of the same equivalent circuit, one calls “exact-PI” or “PI-exact” the PI-circuits
used in the frequency domain and “nominal-PI” the PI-circuits used in the time domain [48].

Page 48
Transformer Energization in Power Systems: A Study Guide

The “Bergeron model” (also known as “CP model”) takes into account the distributed nature of the line parameters
[82]. Therefore it is more suitable than the PI model for the calculation of switching overvoltages due to travelling-
wave reflections. The parameters of the model are constant and thus may be used when no frequency dependency
is to be represented. However, this model may be quite sensitive to the model frequency specified by the user (see
below).

In the widely known J. Marti’s model [83] (also known as “FD model”), parameters are both distributed and
frequency dependent. Therefore, this is the only standard model that is able to represent the frequency
dependency of the line parameters amongst considered models.

If line symmetry is assumed (for instance, if the line is transposed), the cascaded-PI model and the Bergeron model
parameters can be calculated from the sequence characteristics of the line. If the line asymmetry is modelled or if
the JMarti model is used, the model parameters are calculated from the geometrical coordinates of the conductors
and their physical characteristics. When computing parameters from conductor’s geometrical coordinates, it is very
important to include the shield wires (if they exist) and to ground them or to set them to zero voltage.

The three models take ground resistivity as an input parameter. The ground resistivity value along the line is
uncertain and varies with climatic conditions, but [77] shows that the line impedance sensitivity to this parameter is
very low below 5 kHz.

Furthermore, in all the three line models, a unique model frequency must be specified. In the cascaded-PI model,
this is the frequency at which the parameters are calculated. In the Bergeron and JMarti models, it is the frequency
at which the modal transformation matrix Q is calculated. As the main harmonic component of the energization
voltages and currents is the power frequency [2], it seems that the most suitable value to be used in the line
models is the power frequency itself.

Bergeron and JMarti are distributed-parameter models and as such they are suitable to model both harmonic TOV
and possible travelling-wave switching surges at the circuit breaker closing. However, these models limit the
maximum simulation time-step to half the value of the propagation time in the shortest line. Notice also that for low
frequencies (below 1 kHz), Bergeron and JMarti models perform well for single-circuit lines, but much worse for
double or triple-circuit lines [84].

5.1.3.3 Cables
The subject of cable modelling has been extensively covered by CIGRE C4 WG502 (Power System Technical
Performance Issues Related to the Application of Long HVAC Cables). A full review of modelling issues related to
underground and submarine HVAC cables is included in Chapter 3 of [85]. Furthermore, recommendations are
made for the usage of models for various types of studies. The frequency of interest for transformer energization
studies ranges from DC up to 1kHz [44]. In fact, the phenomena of highest concern in the presence of long HVAC
cable circuits or large concentration of short HVAC cables is normally harmonic resonance excited by inrush
currents, hence the frequency of interest is normally restricted to the lower harmonic range.

The major difficulty in modelling underground cables is the fact that they are highly nonlinear in nature; due mainly
to the frequency dependency of the conductors (skin effect), as well as the ground or earth return path. This
frequency dependency is more pronounced than in overhead lines.

Because of the complex geometry, the most accurate methods to calculate cable parameters are those based on
finite element methods. However, finite element methods are not suitable for the types of EMT program cable
models presently available for use in commercial or free-ware simulation tools.

Unless the cable length is very short, a distributed parameter model is recommended for transient studies.
Furthermore, for accurate results at various frequencies, the selected model must also be capable of reproducing
the frequency-dependent behaviour of the cable.

The exact pi model is only recommended for frequency scans. It is not adequate for transient studies, such as
transformer energization, where more than one frequency component needs to be considered.

Page 49
Transformer Energization in Power Systems: A Study Guide

The distributed parameter model at constant frequency (e.g. Bergeron) is only recommended when the studies
involve a single frequency, such as power frequency load flow or very low harmonics.

Frequency dependant modal domain models (e.g. LMARTI) are very accurate in most simple cable systems
because the transformation matrix can be assumed to be constant, with very little error in the time domain solution.
When this approximation can be done, these types of models lead to a very efficient time domain simulation.

When complex cable configurations, or a large range of frequencies need to be considered, frequency-dependent
phase domain models must be used (e.g. ARMA, zcable model and Universal Line model). These models
accurately represent the frequency dependency of the cable parameters and are, therefore, recommended for
studies covering various frequencies, such as transients and harmonic resonance studies.

In terms of the required accuracy in the input data, the results of a sensitivity analysis carried out by C4 -502 WG
indicate that the conductor radius and the permittivity and thickness of the main insulation layer of the cable are the
most critical parameters in the model. Other parameters such as conductor resistivity, sheath thickness and
resistivity, inner semi-conductive layer or earth resistivity have a lesser impact on the accuracy of the model for the
low frequency range of interest in transformer energization studies.

5.1.3.4 Network Equivalents


In order to reduce computational burden, a reduced three-phase (network equivalent) model of the power supply
(external zone) may be required in some transient studies. The aspects to be considered for obtaining this model in
transformer energization studies include the type of transient that can be produced when energizing the
transformer, the size of the external system that will be represented by the equivalent power supply model and the
information to be used for developing the supply model.

For RMS-voltage drop studies in networks where no resonant overvoltages are generated because there are no
low frequency resonances, only the short-circuit impedance of the supply network (the network as seen from the
terminals of the transformer being energized) needs to be represented. If switching surges due to line energization
or re-energization are to be simulated, it is generally sufficient to represent the lines, at least in meshed systems,
up to two busbars away from the energizing bus, and the remaining part of the system may be represented by a
power frequency network model. The Power Frequency Network Equivalents are described further below.

However, if the supplying network impedance has low frequency resonances in the impedance, the inrush currents
and the temporary overvoltages observed during transformer saturation can be very sensitive to the frequency
response of the network. Therefore the modelling of the network has to represent resonances from DC to few kHz
(less than 1 or 2 kHz), at least at the nodes of the study zone where transformer saturation is being simulated. This
chapter will discuss the size of the network and the network equivalent models which are of primary importance
regarding the correct evaluation of possible resonances during transformer saturation.

It is difficult to give general recommendations in terms of minimum number of nodes or distance to be accurately
modelled for transformer energization studies. Reference [86] provides some recommendations for the size of the
study zone to calculate impedances on the HV and EHV grids. For EHV networks, this reference suggests that the
complete primary transmission network has to be modelled. However, a complete modelling of the network leads to
several problems: very time-consuming simulations, high risk of human modelling errors and difficulties to update
the data to take into account different network configurations (for example peak and off-peak loads).

If the frequency response of the network is known (from measurements or from a complete network model used for
a frequency scan), a network equivalent can be built that reproduces its frequency response. With this technique,
the external zone is assumed to be a linear portion of the system that can be reduced to gain calculation speed.
Since the behaviour of the external system is assumed to be linear and the generators in the external system are
not close to the study zone, they are represented by power-frequency voltage sources, which avoid electro-
mechanical-type low-frequency transients in the model. Two network equivalent techniques are described below:
an approximate Single-Resonance Network Equivalent and a full Frequency Dependent Network Equivalent
(FDNE) reproducing the exact response of the network.

Page 50
Transformer Energization in Power Systems: A Study Guide

If the frequency response of the network is unknown or if the user wants an accurate result without using a
sophisticated frequency dependent equivalent, another method consists in modelling the real network (the lines or
cables and the substation equipment) up to a certain distance from the target transformer and representing the rest
of the system by power frequency equivalents. This technique is described below as detailed Modelling of the
Network in the Neighbourhood of the Transformer of Interest.

Power Frequency Network Equivalent


A power frequency network equivalent consists of a voltage source in series with the sequence impedances of the
network. Sequence impedances are easily obtained by performing short-circuit calculations with the model of the
whole network usually available for steady state power flow analysis. Outside the study zone, mutual coupling
between multiple equivalent points could also be implemented using network reduction techniques. Figure 5-13
shows some of these power frequency network equivalents [87].

The first, type a), represents the short circuit impedance (Thevenin equivalent) of the connected system; the
X/R ratio is selected to represent the damping (the damping angle is usually in the range 75º-85º).
The second, type b), represents the surge impedance of connected lines. This equivalent may be used to
reduce the connected lines to simple equivalent surge impedance where the lines are long enough so that
reflections are not of concern in the system under study.
If the connected system consists of a known Thevenin equivalent and additional transmission lines, the two
impedances may be combined in parallel as in Figure 5-13c. It should be noted however, that this approach
may yield an incorrect steady state solution if the equivalent impedance of the parallel connected lines is of
comparable magnitude to the source impedance. In such a case it may not be possible to lump the source and
the lines into a single equivalent impedance.

R X
a) Short-circuit impedance

Z
b) Surge impedance

c) Short-circuit impedance
+ Surge impedance
R X

Figure 5-13 Conventional power frequency network equivalents [87].

Single Resonance Network Equivalent


The model proposed in [45] is shown in Figure 5-14 and can be obtained from information about the short-circuit
currents and the frequency response of the power supply system (external zone). This model simplifies the network
frequency response to a single resonant frequency. To compute Z0 and Z1, the zero and positive-sequence
impedance of the source, one can assume that they are mostly inductive, i.e. 2 , and 2 , f being
the power frequency. L0 and L1 are calculated from the values of single-phase and three-phase short-circuit
currents and by the formula:

0 3
=3 2 Eq. 5-19
1 1

Page 51
Transformer Energization in Power Systems: A Study Guide

Resonance capacitors are determined from the zero- and positive-sequence resonant frequencies by:

1
1 = 2
1 1
1
0 = 2
0 0
1 Eq. 5-20
= 2
1 1
1
= 2 2
0 0 1 1

The damping resistors are given by:

1
1 =
1

0
0 = Eq. 5-21
0

= 1 1

= ( 0 0 1 1)
3
where the factor k defines the level of attenuation.

Figure 5-14 Equivalent circuit of the power supply for low-frequency transients [45].

Frequency-Dependent Network Equivalent (FDNE)


Complex equivalents which properly represent the network’s frequency response characteristic can be used to
reduce the system model and the computational time of the study.

The aim is to obtain a three-phase equivalent circuit which has the same frequency response as the detailed
network from the nodes connected to the portion of the system represented in detail (i.e., the study zone), as
shown in Figure 5-15.

Page 52
Transformer Energization in Power Systems: A Study Guide

Figure 5-15 Linear network equivalent [45].

The linear modelling of the external system is normally based on an admittance formulation which defines the
relation between voltages V and currents I on the terminals of the equivalent [88]:

= Eq. 5-22

The development of a frequency-dependent network equivalent (FDNE) usually involves the following procedure
[89]:

1) simulation of the system to obtain the frequency response (either impedance or admittance) to be modelled
by the equivalent;
2) fitting of model parameters (identification process);
3) implementation of the FDNE in the simulation tool.

The known frequency dependent admittance characteristic of the external system can be estimated by fitting it to a
function of the appropriate order:
2
+ + + +
( )= 0 1 2
2
Eq. 5-23
1+ 1
+ 2
+ +

or in the equivalent form:

( )= 0
+ Eq. 5-24
=1

Since the values of f(j p) are known at an arbitrarily large number of frequency points, Eq. 5-24 can be expressed
as an over-determined fitting problem in the 2n + 1 variables a1, a2, …, aN, and c0, c1, c2, …, cN. However, this is a
nonlinear problem that cannot be solved by linear regression methods.

Early work reported in the literature used frequency domain computed data to fit parameters to the model in Eq.
5-23 [90]. References [91, 92] overcome ill-conditioning problems of FDNE, by dividing the frequency response into
sections. Other techniques like column scaling, adaptive weighting, and iterative step adjustment are also utilized in
these references. A time domain approach to obtain the fitted function Eq. 5-24 using Prony analysis was
presented in [93]. Time domain approaches have also been applied to identify the external system as a digital filter
in [94, 95]. In [96, 97, 98] the external system is modelled using lumped parameters. Vector fitting is another option
that converts the problem in Eq. 5-23 into a linear problem as described below [99, 100, 101, 102, 103].

Many practical applications involve several busses. In such multi-port cases, the same modelling procedure is
applicable as vector fitting can be applied to several elements simultaneously. In practice, one stacks the elements
of Y into a single vector and subjects it to vector fitting which produces a rational model with a common pole set,
which, after rearrangement of fitting parameters, gives the following pole-residue model:

Page 53
Transformer Energization in Power Systems: A Study Guide

= + 0 Eq. 5-25
=1

A symmetrical model is obtained by fitting only the upper (or lower) triangle of Y. References [43, 104] present two
practical cases using this technique.

When the terminals of the equivalent include more than a single three-phase bus, the modelling becomes more
challenging as error magnification problems may arise. When applying a voltage source to one bus, the model is
required to produce large short-circuit currents with a short circuit applied to the other bus, and small currents if the
second bus is open (charging currents). This behaviour is reflected in the admittance matrix Y by large and small
eigenvalues, respectively. Direct fitting of the elements of Y may easily result in corruption of its small eigenvalues,
which may lead to error magnification with certain terminal conditions. Some approaches such as modal vector
fitting overcome this problem by assigning high weights to the small eigenvalues of Y in the least-squares fitting
process [105].

More details on the determination of frequency-dependent network equivalents can be found in [88, 106].

Derivation of such network equivalent requires a significant amount of expertise and experience and it is rare that
practicing power system planning engineers will be able to carry out the whole procedure. In such cases it is
advisable to use the model once it is clear the detail with which it must be represented.

Detailed Modelling of the Network in the Neighbourhood of the Transformer of Interest


As previously mentioned, the supply network can be represented by modelling the real network (the lines or cables
and the substation equipment) up to a certain distance from the transformer of interest and representing the rest of
the system by power frequency equivalents. In order to decide up to what distance from the transformer the system
is to be represented in detail, either at the same voltage level (horizontal distance) or at different voltage levels
(vertical distance), the following rule is suggested: progressively increase the distance (horizontal and vertical) of
the detailed modelling until the results do not change in a significant manner. As a general principle, the model
should be extended with priority given to the electrical nodes containing the higher capacitive components
(including cables), as these will most affect the frequency response. Figure 5-16 and Figure 5-17 illustrate this
technique. They provide the network impedance seen from the transformer to be energized, depending on the
horizontal and vertical distances modelled in detail (before the power frequency equivalents). Figure 5-16 shows
that, for this particular case, representing the 400 kV network up to three nodes away from the transformer of
interest is enough to represent approximately the resonance (the response when representing up to four nodes is
close). Note that in mesh networks a distance of three nodes away from the target transformer involves more than
three nodes. In this particular case, about ten 400 kV nodes were at a distance of three nodes from the transformer
of interest. Figure 5-17 shows that, for this particular case, representing 225 kV up to three nodes is enough
(representing up to four, five and six nodes provides a similar response).

Summary
For RMS-voltage drop studies in networks where no resonant overvoltages are generated, a simple power
frequency network model can be used to represent the network outside the study zone.

When the frequency response of the network is of concern, a more detailed model of the system is required. If the
user has the frequency response of the network, a network equivalent can be built. A single resonance network
equivalent is easy to build but it may oversimplify the frequency response. For better accuracy, a FDNE should be
used, but then this requires a higher level of expertise. Another option, if the user is not familiar with network
equivalents or if there is no information on the frequency response of the network, is to build a detailed model of
the network in the neighbourhood of the transformer to be energized.

Page 54
Transformer Energization in Power Systems: A Study Guide

Figure 5-16 Network modelling: frequency response as a function of the horizontal distance (same voltage level)
between the transformer of interest and the power frequency equivalents.

Figure 5-17 Network modelling: frequency response as a function of the vertical distance (lower voltage levels)
between the transformer of interest and the power frequency equivalents.

Page 55
Transformer Energization in Power Systems: A Study Guide

5.1.3.5 Generators
Two cases must be distinguished: synchronous and asynchronous generators. According to IEC TR 60071-4 [45]
and CIGRE TB 39 [44] guidelines for low-frequency transients modelling, synchronous machines must be modelled
by the generalized Park’s model based on the complete equations in the direct and quadrature axes, representing
saturation, excitation and mechanical torque, as well as voltage and speed controls. The effect of the capacitances
is negligible.

However, in transformer energization studies the speed control may be neglected as transformer energization does
not generate any significant active power variation. As for the voltage control, the modelling must be aware of the
behaviour of the regulator in the presence of voltage harmonics, a fact not taken into account in power-flow
models. In cases where temporary overvoltage is a concern, automatic voltage regulator (AVR) can be included in
the overall generator model as it could be responding to an RMS voltage dip while the network itself is experiencing
a TOV due to harmonic resonance.

If no significant voltage distortion is generated at the energization, the generator may be represented by an ideal
sinusoidal voltage source with appropriate phase angles (from load flow) in series with the subtransient direct
reactance and the armature resistance, thus neglecting generator dynamics and voltage control [47, 78].

As mentioned in Section 5.1.1, if the synchronous machines are far enough from the transformer being energized,
they can be included in a network equivalent.

The case of wind turbine generators (WT) using asynchronous machines is different. The behaviour of wind turbine
generators during a voltage change is complex and is mainly determined by the electronic converter controls
designed to comply with the Fault Ride Through (FRT) requirements of Distribution and Grid Codes. These
requirements are different from country to country, for instance, some codes require maximization of Mvar injection
during voltage dips and other codes require MW injection to be proportional to the retained voltage. Therefore it is
not possible to use a generic WT model. The user should ask for a detailed model from the WT manufacturer
applicable for the case under investigation.

5.1.3.6 Loads
Loads form a fundamental component of the real power system, no less important than the generating stations
which provide the electrical power and the transmission and distribution systems which convey the power to the
system loads. In simulations investigating the effects of transformer energization on the power system, the
appropriate representation of loads in the network model is important for the following reasons:

1) Loads must be included in the network model to achieve correct initial conditions for an EMT case, prior to any
time domain simulation. The EMT model and resulting three-phase load flow should be based on and agree
with a load flow calculation from a load flow program, as would be calculated by system development and/or
planning engineers.
2) Normal aggregate loads provide the main damping of transient and temporary overvoltages and may affect
resonance conditions, particularly at higher frequencies [86, 89, 107].
3) Loads can transiently and dynamically interact with the inrush current, particularly for loads that are of
significant size.
System loads can have a significant effect on the response of the network to transformer energization. Loads
provide important damping of transient overvoltages caused by excitation of low order harmonic parallel
resonances in the network from the harmonic components in the magnetizing inrush currents. Customer’s
synchronous motors contribute to the system fault levels and may help to mitigate voltage sag due to the flow of
inrush current through the network. On the other hand, the response of large induction motors when terminal
voltages sag, due to their voltage versus slip characteristic, will be to absorb increasing reactive power, which may
aggravate the voltage sag after a few cycles. For transformer energization studies it is particularly important to
include, in the system model, large loads that are located close to the transformer to be energized.

Page 56
Transformer Energization in Power Systems: A Study Guide

If the subject of the EMT simulation study is transient overvoltages, then the worst case would be a light load case
where the damping of overvoltages provided by the loads is minimum. The presence of shunt capacitor banks in
the network and in customer premises can contribute significantly to transient overvoltages. Industrial customers
having significant power factor correction shunt capacitor banks or distribution stations having switchable
distribution level shunt capacitor banks should be modelled at the minimum load for which the capacitor banks
would still be in service. On the other hand, if voltage sag is the concern, a worst case would be obtained by
modelling distribution system loads as being high but industrial customers should be modelled with a minimum
number of large rotating machines in service. This, obviously, requires some knowledge of the customer’s various
operating modes if a realistic case is to be created. If the rate of the recovery of the voltages is expected to be a
concern, then large induction motors and their dynamics should be explicitly included in the system model.

Loads can be categorized as industrial, commercial, and residential/distribution, as will be described below. The
modelling of different types of loads in EMT programs is described as follows.

Industrial Loads
These loads, which are generally relatively large, are supplied directly from the transmission system via isolating
breaker(s), disconnect switches and intertie transformer(s). Pulp and paper mills, mining loads where ore is
crushed and processed, compressor loads to liquefy natural gas, steel mills, ship loading and off-loading facilities,
and the petrochemical industry are examples of industrial loads. Figure 5-18 shows a simplified electrical single-line
diagram of a 30 MW copper-gold mine industrial load with synchronous and induction motors supplied from two
independent busses, as well as a power correction capacitor bank.

The supply voltage at the point of interconnection with the utility can range from 60 kV to 345 kV (and perhaps
even higher). The size of the load can range from a few MW to hundreds of MW, depending on the utility supply
voltage. This type of load typically consists of a mixture of large synchronous and induction motors supplied at 2 kV
to 25 kV according to power rating, a number of smaller induction motors supplied from lower voltage utilization
busses and, finally, a possibly large number of quite small induction motors supplied from utilization busses where
voltage can be as low as a few hundred volts. In general, there could also be motors that are supplied from motor
drives based on high power electronic converters, where the horsepower rating of the motors could range from a
few tens of kW to many thousands of kW. Usually, large harmonic-producing motor drives, such as LCI, cyclo-
converters and variable speed drives, will have some associated harmonic filter banks and/or shunt capacitor
banks for power factor correction. Industrial loads may also contain a relatively small component dedicated to
providing the power requirements for lighting, heating and ancillary services.

Large industrial loads that are located electrically close to the transformer to be energized or along the path of the
inrush current, should preferably be modelled in detail. This requires possession and knowledge of an up-to-date
electrical single-line diagram of the customer’s facilities. The single-line should show the ratings and locations of
the intertie and low voltage transformers, the disposition of the LV utilization busses, the larger motors, motor
drives, power factor correction capacitor banks and harmonic filter banks. Ideally, large rotating machines, such as
generators, synchronous motors and induction motors should be represented by electro-dynamic machine models.

The utility intertie and lower voltage transformers should be modelled according to Section 5.1.2. Saturation effects
should be included when the customer is located electrically close to the transformer being energized, due to
concerns of sympathetic interaction [4].

Synchronous motors (or generators if the customer has self-generation) should be modelled as per Section 5.1.3.5
by the generalized Park’s model, such as the Synchronous Machine Model available in EMT programs. Modelling
should include the excitation system and mechanical torque. Large induction motors should be modelled by using,
for example, the Universal Machine Model. Reference [108] lists typical data for induction motor models over a
large range of kW ratings. The behaviour of the mechanical load with change in speed should be accounted for by
the model.

Page 57
Transformer Energization in Power Systems: A Study Guide

Figure 5-18 Simplified single-line diagram of a typical industrial load

In many cases, however, this approach of detailed modelling of rotating machinery becomes impractical, due to
lack of data or because of overwhelming complexity. Some simplifications to the modelling can be made that
provide conservative yet usually reasonable results for transformer energization studies. Multiple synchronous
machines supplied from the same bus can be combined into a single equivalent machine and the model can then
be simplified as ideal sinusoidal voltage sources behind the d-axis subtransient reactance with a resistor in series
per phase to account for the armature winding resistance (see Section 5.1.3.5). The subtransient reactance is
normally around 0.2 p.u. based on machine ratings. In a similar way, induction motors can be simplified as ideal
voltage sources behind the operational impedance of the motor which, for low frequency transients, is the locked-
rotor impedance. The locked rotor impedance of squirrel-cage induction motors is generally around 0.2 p.u. based
on ratings.

Large converter drives, such as LCI or adjustable speed, should be modelled explicitly but, if this is not possible,
they could be approximated and treated as a DC converter terminal as seen from the system side. As a crude
approximation, the converter can be modelled as a constant impedance load. Any associated filtering and shunt
capacitor banks should be explicitly represented.

Many small low voltage (typically 600 V and below) motors supplied from the same bus, where they are separated
from the utility supply by several layers of transformers, can be treated as an aggregate load and approximated by
a constant impedance equivalent R in parallel with j L. This should account for the real and reactive power
consumption of the load. Adjustable speed drives to small motors can be aggregated with the small motors that are
directly connected to the same low voltage AC bus and included in the constant impedance model.

Page 58
Transformer Energization in Power Systems: A Study Guide

Shunt capacitor banks used for power factor correction and harmonic filter banks, where the voltage rating is more
than a few hundred volts, should be explicitly modelled. These capacitor banks may affect the lower frequency
series and parallel resonances as seen from the transformer being energized and can influence the transient
response of the network.

Commercial Loads
These loads, generally smaller than typical industrial loads, are supplied from feeder circuits originating from a
distribution substation at voltages usually at or below about 35 kV. Examples of this type of load would be sawmills,
commercial refrigeration plants, and the induction motor driven pumps that are found in an oil and gas field for
extraction of fluids.

The induction motors will be smaller than the large ones found in most industrial customers, and ratings could be
from a few kW to possibly a few hundred kW. Some motors might be fed from adjustable speed drives.

The modelling of individual commercial loads, for transformer energization studies, should follow the guidelines
given for industrial loads where the motors are not expected to have a large impact on the simulation. The largest
motors can be aggregated, where appropriate, and modelled as ideal sinusoidal voltage sources behind locked-
rotor impedance. The remainder of the load, including all of the small motors, can be combined and modelled as
constant impedance equivalent to reflect the real and reactive power consumption. Shunt capacitor banks should
be included in the model, if the kVAr rating is significant.

Distribution Substation for Residential/Light Commercial Loads


These loads are generally supplied from a system of feeder circuits emanating from a distribution substation. While
the individual loads are relatively small (perhaps only a few kW), there could be many such loads involved.
Residential loads comprise the usual domestic small low voltage single-phase induction motors (for refrigerators,
washing machines, clothes dryers, pumps etc.), consumer electronics, resistive heating, and so on.

The modelling of a distribution system, where the total load is the aggregate of many small low voltage loads,
presents some difficulties as the system exhibits some frequency-dependence and the nature of the load varies
with time of day. Ideally, this complicated system would be represented by a frequency-dependent equivalent
circuit based on field measurements. However, in practice, such field measurements are not easily available for the
model developer. The literature provides ideas for modelling such loads for harmonic studies [86, 89, 107]. For
such studies, loads are often represented by simple models consisting of series and/or parallel resistances and
inductances derived from the active and reactive power flow parameters. These R and L parameters can be given
harmonic weighting factors to reflect frequency dependency. For transformer energization studies, it might
sometimes be impractical to develop a frequency-dependent model for a low voltage distribution-type load, so a
simple constant equivalent impedance comprised of R in parallel with L to ground should be sufficient. This
equivalent should account for the steady state real and reactive power consumption of the load being represented.

It is important to model shunt capacitor banks (switchable or fixed) within or electrically close to the distribution
substation. Also, if there are distribution cables emanating from the station that are of significant length (few
hundred metres) then the effective shunt capacitance of these should be calculated and combined into an
equivalent shunt capacitance at the station LV bus.

Modelling of a Loaded Transformer


Sometimes the transformer to be energized, for simulation purposes, cannot be considered to be unloaded. Firstly,
there might be some load connected to the transformer LV bus so that, when the transformer is picked up, so is the
load. An example would be a distribution station transformer where the feeder circuit breaker or disconnect switch
remains closed even though the transformer and distribution system are isolated from the grid and are de-
energized. In this case the inrush current, when the transformer is energized, includes the effects of a “cold-load
pickup”. Secondly, there could be one or more secondary transformers and some low voltage auxiliary load. A
typical example would be a generator transformer with an LV station service transformer to auxiliary loads

Page 59
Transformer Energization in Power Systems: A Study Guide

comprised of small motors, heating, lighting, and so on. The ratings of these secondary transformers will normally
be significantly less than the rating of the main transformer. The presence of a secondary transformer and/or
connected load can affect the overall inrush currents and the response of the external network compared to the
results obtained with only the main transformer.

If the MVA rating is significant compared to the main transformer, a station service transformer (or other secondary
transformer) should be included in the system model. The transformer will have a nonlinear magnetizing
characteristic and residual flux in the core as a result of being switched off. The model developer should be aware
that a secondary transformer might be comprised of single-phase units rather than being a three-phase unit.
Ideally, it should be modelled in accordance with the guidelines discussed elsewhere in this document for
modelling of the main transformer. Three-phase motors will probably not be a consideration since the motor
contactors will have been opened by motor protection and they should remain disconnected for the duration of the
transformer energization. Small low voltage single-phase motors, on the other hand, will still likely to be connected
to their low voltage system and should be included in the modelling. They can be aggregated where appropriate
and represented by a locked-rotor equivalent impedance. Heating, lighting and small electronic loads can be
modelled as an equivalent constant impedance load, as discussed in the previous subsection.

5.1.3.7 Circuit breakers


For the closing operation, a simple ideal switch can be used [44] since the arc condition, and particularly the
prestrike phenomena, can be neglected. It is very important to represent the mechanical pole spread (see Section
5.3.2) and if the circuit breaker is equipped with closing resistors, they have to be modelled too.

If the opening operation is being simulated in order to determine the residual flux (this implies that the transformer
model is capable of representing the de-energization transient and can retain the residual flux), then a more
sophisticated CB model capable of representing the arc condition has to be used. The model should at least
represent the high current interruption and current chopping capability of the CB; restrike characteristics and high
frequency current chopping can be neglected [44]. Black box arc models can be used for this application [109,
110]. The grading capacitors (if any) of the CB are also to be included in the model, as the capacitively coupled
voltage on the transformer side of the breaker may interact with the initial residual flux.

5.2 Quantification of the Overvoltage Stress in Transformers and Surge


Arresters
The previous sections of this document have shown how to simulate the transient voltages and currents generated
during the energization of transformers. Due to resonance phenomena already described, harmonic resonant
overvoltages may appear during the energization that could damage the equipment insulation. This section deals
with the assessment of these potentially damaging effects of the overvoltages.

In the IEC documentation, resonant overvoltages belong to the class of temporary overvoltages (TOV) [111]. In
order to prevent any damage to the substation equipment, the amplitude and duration of the calculated
overvoltages must be compared to the TOV withstand capability of the most vulnerable equipment. For that, phase-
to-ground and phase-to-phase voltages must be distinguished.

Most Vulnerable Equipment Phase-to-ground and Phase-to-phase Overvoltages


The TOV withstand capability depends on the equipment considered. In general, for phase-to-ground TOV lasting a
few seconds or less, the surge arresters connected between phase and ground to protect the transformers from
switching and lightning surges are the most vulnerable equipment. For longer duration overvoltages, the most
vulnerable equipment to phase-to-ground TOV, generally, are the power transformers [19, 47, 81]. As arresters
between phases are seldom used, three-phase power transformers are then the most vulnerable equipment to
phase-to-phase TOV.

Note that per unit phase-to-phase withstand of three-phase power transformers is usually smaller than the per unit
phase-to-ground withstand capability of the surge arresters (see figure below); this means that a given overvoltage

Page 60
Transformer Energization in Power Systems: A Study Guide

may be below arresters’ (phase-to-ground) withstand capability but above transformers’ phase-to-phase withstand
capability.

It follows from these considerations that the calculated phase-to-ground voltages must be below the withstand
capability of the surge arresters and transformers, and that the phase-to-phase voltages must be below the three-
phase transformers phase-to-phase withstand capability.

Evaluation of the Surge Arresters Stress


According to the standards [112, 113], the manufacturers provide information on the surge arrester TOV withstand
capability. This information is given in the form of a voltage/duration curve. These curves provide the maximum
allowable duration of a power frequency voltage of given constant amplitude, as illustrated in Figure 5-19.

Figure 5-19 Typical V-t withstand capability curves for A surge arrester and A transformer.

Usually, two curves are provided with and without prior energy absorption (taking into account a possible prior
surge). To be conservative, it is recommended to use prior energy curves. In the IEC standard, the rated voltage of
the arrester corresponds to the TOV withstand for 10 s with prior energy. Then, the V/t curve takes this general
form [112, 114]:
1
10
, = Eq. 5-26
,

where B1 defines the arrester V-t curve slope.

Calculated phase-to-ground TOV must be compared to the TOV capability provided by the surge arrester
manufacturer [113, 114, 115]. Thus, if the calculated overvoltage amplitude and duration are V and t, it must be
verified that

< , () Eq. 5-27

or, alternatively,

< , ( ) Eq. 5-28

Page 61
Transformer Energization in Power Systems: A Study Guide

However, transient overvoltages due to transformer energization have time-varying amplitude. In order to take this
into account, [81] suggests dividing the voltage signal in several constant-amplitude regions as shown in Figure
5-20 and accumulating the effects:

1 2
+ + <1 Eq. 5-29
, ( 1) , ( 2)

Figure 5-20 Variable amplitude overvoltage (single frequency) (source: [81])

Moreover, if the voltage contains harmonics and therefore it is not sinusoidal, it should be treated as purely
sinusoidal at power frequency having amplitude equal to the maximum instantaneous voltage. This is equivalent to
assuming a power frequency envelope of the voltage.

An alternative method to assess the arrester stress is to calculate its energy duty due to the overvoltage and
compare it to its energy absorption capability (given in kJ/kVrated or kJ/kVmcov). This method has the advantage that
it takes harmonics into account (whereas TOV withstand capability provided by manufacturers is for power-
frequency voltages) [113]. However, the energy absorption capability depends on the form of the overvoltage [116],
and the manufacturer kJ/kV values are in general for switching surges [114]. Furthermore, the energy absorption
capability is not covered by the standards.

Evaluation of the Transformers Stress


As already mentioned, for TOVs with short duration, the transformers’ phase-to-ground withstand capability is
generally larger than that of the surge arresters. Therefore, it can be concluded that if the surge arresters are
capable of withstanding the simulated overvoltages, the transformer phase-to-ground insulation should also be
capable. However, it must be emphasized that this will not be true for longer duration TOVs. The case of phase-to-
phase overvoltages is different. They do not affect the surge arresters (surge arresters between phases are
uncommon) but they do stress the inter-phase insulation of three-phase power transformers.

Page 62
Transformer Energization in Power Systems: A Study Guide

a) Using the short-duration power frequency test level

The overvoltage withstand information usually available for transformers are the standard test levels for power
frequency (short and long duration), switching impulse and lightning impulse3. According to IEC 71 on insulation
coordination [111], the standard withstand voltage test for temporary overvoltages is the short-duration power
frequency test. Thus, the calculated overvoltages should be compared to the voltage applied during the short-
duration power frequency test.

Moreover, when comparing expected overvoltages to equipment standard test levels, IEC 71 suggests using a
safety factor in order to “compensate for the differences in the equipment assembly, the dispersion in the product
quality, the quality of installation, the ageing of the insulation during the expected lifetime, other unknown
influences” [117]. For internal insulation, IEC suggests using a security factor of 1.15, which means a 13%
reduction of the initial insulation strength. As for the ageing effect alone, [118] suggests that the reduction in
transformer strength is between 5 to 10%.

In conclusion, following the IEC 71, the simulated overvoltages should be compared to the voltage applied during
the short-duration power frequency test reduced by a safety margin varying from 5% to 13%. Notice that this
method of evaluating the stress pays no attention to the duration of the overvoltage: there is only a voltage
threshold beyond which the transformer withstand capability is exceeded.

The user must also consider the overfluxing capabilities of the transformer [19]. For instance, for transformers
connected to generators, IEC 76 asks for 1.4 times the rated voltage during 5 s.

b) Building V-t curves

It should be recognised that the threshold method may be too conservative and lead to unrealistic and too
restrictive results. This is due to the fact that the transformer overvoltage withstand is a function of the duration, the
same as for surge arresters (see Figure 5-19), and the duration of the short-duration power frequency test is much
longer than that of the harmonic TOV generated by transformer energization. The power frequency test duration is
usually between 30 and 60 seconds, whereas the duration of the energization TOV is seldom longer than several
seconds. Thus, the transformer’s actual withstand capability to energization TOV is much larger than the short-
duration power frequency test level.

A more realistic and less conservative evaluation of the transformer stress is obtained by using voltage/duration
curves such as those presented for surge arresters [19, 47, 119], both phase-to-ground and phase-to-phase.
These V/t curves, however, aren’t covered by the standards and must be individually requested from the
manufacturers [120]. If no curve can be obtained from the manufacturer, a tentative V/t curve may be built by
interpolating between voltage levels induced during the standard tests, as shown in [118], to which a safety factor
should also be applied. This is illustrated in Figure 5-19 above, where phase-to-ground and phase-to-phase TOV
capability curves have been built for a transformer by semi-logarithmic interpolation between the phase-to-ground
and phase-to-phase voltages induced during the switching and power frequency tests, reduced by a safety factor of
1.3 suggested by the manufacturer [121].

5.3 Parameter Uncertainty Assessment and Stochastic Simulation


The previous sections have shown how to model the system in order to simulate energization transients, i. e., the
inrush currents, the RMS-voltage step change and the possible resonant overvoltages. Yet, when modelling the
system, the value of a number of parameters is not precisely known, that is, they are affected by some uncertainty.

3
Not all the tests are always performed, but standards provide test conversion factors in order to calculate equivalent tests
levels.

Page 63
Transformer Energization in Power Systems: A Study Guide

The variation of many of these parameters in their uncertainty range does not affect the results (i. e., the computed
currents and voltages); thus, these uncertainties can be neglected. However, this will not be the case for a small
set of parameters: those for which the results are highly dependent on the particular parameter value within its
uncertainty range. Therefore, the uncertainty on these parameters cannot be neglected but rather has to be taken
into account in the study.

Due to their dependency on uncertain parameters, the currents and voltages are not deterministic but rather
stochastic variables, their values depending on the particular range taken by the uncertain parameters. Therefore,
the output of the simulation study is not a single figure but a probability distribution. Often, however, the user is not
interested in the whole distribution of the output but in the risk of exceeding a particular threshold limit, for instance
a given RMS-voltage drop limit or the equipment overvoltage withstand limit. This is achieved by computing the
threshold-exceeding probability, pthr.

In order to compute the probability distributions of the stochastic variables (currents, voltages, stresses…) and the
threshold-exceeding probability, a number of simulations must be performed to assess the effect of the
uncertainties associated with each input parameter. For this, the important uncertain parameters must be identified
and characterized, and a suitable stochastic simulation technique must be used. The following sections treat each
of these aspects. For a general framework on the treatment of uncertainty the reader is referred to [122].

5.3.1 Important Uncertain Parameters


Many input parameters may be affected by uncertainties; however, only those whose variation may have an
important influence on the output variable need to be considered. It is possible to identify potentially important
parameters thanks to the knowledge of the physical phenomenon under study. Regarding the inrush currents at
transformer energization, it is well established that they are highly dependent on two random parameters: the
residual flux in the core before the energization and the circuit-breaker point-on-wave switching times. Indeed, it is
widely accepted that the residual flux within the transformer core does not change over time unless there are
external influences4. As for the circuit-breaker switching times, they can occur any time in the sine wave and in
general they are not simultaneous because of the mechanical pole spread. A third uncertain parameter affecting
the inrush currents may be taken into account: the air-core inductance of the transformer, as its value cannot be
measured and is analytically calculated by the manufacturer with some estimated accuracy. Notice that the residual
flux and switching-times values change at each energization due to the intrinsic random behaviour of the physical
phenomena, whereas the air-core inductance has a constant value that is not precisely known due to a lack of
knowledge. These two kinds of uncertainties are called, respectively, aleatory and epistemic uncertainty [122].

If the system resonant frequencies are low (say below 800 Hz), then the (epistemic) uncertainties on the capacitive
and inductive components of the system must be taken into account, as very small variations of these parameters
can tune the system resonances to the inrush current harmonic frequencies and thus lead to resonant overvoltages
[43, 121]. As an example, in the power restoration configuration presented in Section 5.3.4, two parameters
influencing the network frequency response are considered as affected by some uncertainty: the generator
reactance and the phase-to-ground capacitance of the overhead lines.

5.3.2 Uncertainty Quantification and Modelling


The uncertainty of each relevant parameter must be quantified and modelled by a probability distribution. For most
of the parameters, a uniform distribution can be used as there is no reason to favour one or another value in their

4
Grading capacitor together with the stray capacitance of the bus bars and transformer induce a power frequency voltage on
the transformer after the opening of the CB. [30] suggest that the presence of grading capacitor has the tendency to
decrease the residual flux in the transformer after de-energization.

Page 64
Transformer Energization in Power Systems: A Study Guide

uncertainty range, i.e., all the possible values are equally probable. (A triangular distribution could also be used if
the central value is considered the most probable.)

Circuit breaker closing times


The uncertainty in the circuit breaker closing times can be modelled by four random parameters: a common order
time, the same for the three poles of the CB, and the random offset time of each pole due to the mechanical pole
spread: tclose,i=torder+toffset,i i=A, B, C. The common order time may take any value in the sine wave, thus following a
uniform probability distribution over the power frequency period T (equal to 20 ms if f=50 Hz). The offset time of
each pole is considered to follow a Gaussian law whose mean is zero (the three poles tend to close
simultaneously) and whose standard deviation ( ) is calculated from the maximum pole span, MPS (i. e., the
maximum delay between poles). The three toffset,i follow the same Gaussian law N(0, ). As the interval ±3. has
99.7% probability in a Gaussian distribution, the standard deviation may be calculated from this interval width as
=MPS/6. A 5 ms maximum delay between poles can be considered if no specific information on the CB is
available [44, 45]. However, delays up to 20-40 ms have been measured in special circumstances. For MPS=5 ms,
torder=U[100, 120] ms, toffset,i=N(0, 0.83) ms.

Residual Flux
The residual flux in each leg of the transformer core may depend on a number of factors: the core type, the winding
coupling, the neutral grounding, the circuit breaker, the load condition at the de-energization, and the time elapsed
since the previous de-energization. For discussion purposes, only the flux in the wound legs is considered and it is
also assumed that, at the de-energization, the transformer was unloaded and fed by balanced symmetrical
voltages.

As the steady state transformer magnetizing current is small, it can be chopped by the circuit breaker (CB).
Therefore, the opening of the three phases can occur at any phase angle. This means that, just after the opening,
the fluxes in the three wound legs will form a 120° phase-shifted balanced symmetrical set. Depending on the
capacitances linking the transformer to the network and/or to the ground (stray and CB capacitances), these fluxes
will decay and then stabilize at a certain value. The final values depend on the network characteristics. However,
they will range between 0 and 80% of the flux at the rated transformer voltage [9, 123, 124]. Moreover, assuming
that the system after the de-energization is symmetric, the three fluxes will remain balanced and 120° phase-
shifted. This assumption is supported by the experimental results in [48, 59].

Let r be the flux at the rated voltage and 0 and two auxiliary variables representing the amplitude and the phase
angle of the balanced symmetrical set of the residual fluxes. According to the preceding considerations, 0 can take
any value between 0 and 0.8. r and can take any value between 0 and 360°. In statistical terms,

0 = 0.8 [0,1] Eq. 5-30

= [0,2 ] Eq. 5-31

where U[a,b] represents a uniform distribution between a and b.

The potential residual flux values in the three wound legs of the transformer are then obtained from the values of 0
and as:

2
= 0 cos +( 1) , = 1. .3 Eq. 5-32
3

By generating random samples of 0 and , one can obtain random samples of the three residual fluxes k.

Page 65
Transformer Energization in Power Systems: A Study Guide

If the user does not want to deal with generating random samples, a simpler but approximate method to account for
the potential residual flux values consists in using a fixed number of meaningful and equally probable flux patterns.
As an example, the following can be used as potential residual flux values:

{ 1 , 2 , 3 } = {0,0,0} Eq. 5-33

with relative probability of 1/3

{1, 0.5, 0.5} 0.8


{ 1 , 2 , 3 } = { 0.5,1, 0.5} 0.8 Eq. 5-34
{ 0.5, 0.5,1} 0.8

each of which with relative probability of 1/9

{0,0.76, 0.76} 0.8


{ 1 , 2 , 3 } = {0.76,0, 0.76} 0.8 Eq. 5-35
{0.76, 0.76,0} 0.8

each of which with relative probability of 1/9.

Air-core Inductance
The air-core reactance can be modelled by a uniform distribution centred at the manufacturer’s value and covering
the accuracy range (±20% accuracy is often suggested by the manufacturers).

Supplying Network Parameters


The uncertainty on the inductive and capacitive parameters of the system must be considered when the resonant
frequencies of the system are low (generator inductances, overhead line inductances and capacitances, etc.). The
level of uncertainty depends on the degree of confidence in the data used to build the system model. For instance,
overhead-line sequence parameters may have been calculated for an average tower and conductor coordinates
and not the actual ones. In the example below, Section 5.3.4, some other uncertainties are considered in a power
restoration configuration. Reference [43] presents an analysis of some uncertainties to be considered when
modelling large networks.

5.3.3 Stochastic Simulation: the Monte Carlo Method


Stochastic simulation techniques provide a way to compute the output variables probability distributions given the
probability distributions of the input parameters. The best known technique is the Monte Carlo simulation, the only
one that is treated here. As Monte Carlo simulation can be very slow, a number of alternative techniques have
been suggested to accelerate the process [125, 126]. Some of them have been compared for transformer
energization studies in [121].

The Monte Carlo method


The Monte Carlo method [125, 126, 127] estimates the output variable probability distribution by generating a
sufficient number of random samples of this variable according to the random probability distributions of the
uncertain parameters. For example, if the output variable is the RMS-voltage drop and the uncertain parameters
are the residual flux and the CB closing times, then the RMS-voltage drop probability distribution is estimated by
simulation. The residual flux and CB closing times values used in each simulation are randomly generated
according to their respective probability distributions. The estimate of the cumulative distribution function (CDF) of
the output variable (in this case the RMS-voltage drop), Y, F(y) = Prob(Y<y), is then obtained as:

Page 66
Transformer Energization in Power Systems: A Study Guide

1
( )= 1 < Eq. 5-36

where N is the number of simulations performed, yi (i=1,2..N) are the output variable outcomes and 1yi<y equals 1 if
yi<y and 0 otherwise.

Usually, the user is mainly interested in the probability of exceeding a particular threshold limit of the output
variable, y0, pthr = Prob(Y>y0), for instance the probability of exceeding a given RMS-voltage drop limit. Then, only
one point of F(y) needs to be calculated, as the threshold-exceeding probability corresponds to pthr=1-F(Y<y0). This
is given by:

1
= 1 > 0 Eq. 5-37

i.e. the threshold-exceeding probability equals the number of the output variable simulation outcomes that exceed
the threshold divided by the total number of simulations.

For the estimate to be meaningful (unbiased) in the Monte-Carlo method, the only condition that must be met is
that the outcomes of the output variable are randomly generated according to the probability distributions of the
uncertain parameters.

The second essential point of the Monte-Carlo method is that the statistical error of the output variable estimate
can be assessed by means of a (asymptotic) confidence interval of the estimate. In the case of a threshold-
exceeding probability, the 95% confidence interval is easily calculated as:

1
± 1.96 Eq. 5-38

If the empirical threshold-exceeding probability is zero, i. e., if no simulation exceeds the threshold limit, then the
so-called rule of 3 provides the following confidence interval for pthr [128]:

3
=0 0, Eq. 5-39

Explicit Monte Carlo Algorithm


Based on the foregoing discussion, the list below presents the explicit steps for a Monte Carlo study to calculate a
threshold-exceeding probability, pthr :

1) Choose a stop criterion; for instance, the desired maximal confidence interval width pthr,MAX
2) Generate a random sample xi of the uncertain parameters
3) Simulate the output yi for xi
4) Compute the confidence interval width pthr (Eq. 5-38 or Eq. 5-39)
5) Plot the convergence graph (i.e., the confidence interval as a function of the number of simulations: see
Figure 5-23)
6) Test convergence of the algorithm: if pthr pthr,MAX continue to step 5; else go to step 1

Plotting the convergence graph (step 4) is not a necessary step but it is highly recommended in order to check the
whole process.

Page 67
Transformer Energization in Power Systems: A Study Guide

Minimal Number of Simulations


The confidence interval is said asymptotic because it is valid only if the error is Gaussian distributed, which,
according to the Central Limit theorem, is true only as N tends to infinity. In practice, this means that N must be
large enough. How large N needs to be depends on the problem. 30 is usually suggested as a minimal N. Another
rule of thumb states that a binomial distribution approximates a Gaussian distribution when p.N>5 and (1-p).N>5,
where p is pthr or F(y) (estimated by their empirical estimates).

Computer Implementation
EMT simulation packages rarely implement stochastic simulation capabilities (usually they can only generate
random switching times for circuit breakers). Thus, the random parameter generation and confidence interval
calculations need some form of programming either in the simulation package itself (if it has scripting capability), or
through coupling with a numerical computational software.

5.3.4 Example of Application to Transformer Energization Studies


As an example, consider the computation of the stress due to the overvoltages generated when energizing the
transformer of the following network [121] (for an application to the RMS-voltage drop assessment, see [129]):

Figure 5-21 Study case.

Figure 5-22 presents the feeding system frequency response to positive-sequence excitation, showing two
resonance points, the first close to 200 Hz and the second close to 700 Hz.

Figure 5-22 Feeding system frequency response to positive sequence excitation (all three phases shown).

A stress rate function has been built according to the indications in Section 5.2, so each simulation is associated
with a single figure, called the global stress rate, GSR. If the GSR exceeds 1, then the energization is considered
as leading to failure. The purpose of the study is to compute the threshold-exceeding probability GSR>1, i.e., the
risk of failure at the energization, pf.

All the uncertain parameters mentioned in Section 5.3.1 are included in the study: As for the parameters influencing
the inrush currents, the uncertainty of the switching times of the circuit-breaker poles (t 0, toffset,A, toffset,B, toffset,C), the
residual flux at the energized transformer ( 0, , k) and the air-core inductance of the transformer (Lair-core) are taken

Page 68
Transformer Energization in Power Systems: A Study Guide

into account. As the first system resonant frequency is low, 200 Hz, the uncertainty of two parameters influencing
the system frequency response is considered: the generator reactance (L’’d,gen), whose value is provided by the
manufacturer with a given accuracy, and the conductor heights of the overhead lines (HOHL), affected by some
uncertainty due to the averaging of the tower heights, the varying sag of the conductors with temperature and the
terrain topography. Reference [121] suggests using 15% accuracy for both parameters. Notice that 15%
uncertainty on the conductors’ height is equivalent to approximately 3% uncertainty on the phase-to-ground
capacitance. Both uncertainties are modelled by uniform distributions centred at their nominal values.

Figure 5-23 shows the confidence interval for the risk of failure, pf, provided by the application of the Monte Carlo
algorithm depending on the number of simulations performed. This figure shows that the confidence interval
becomes smaller as the number of simulations increases (although the reduction slows down as the number of
simulations increases, a known drawback of the Monte Carlo method).

Figure 5-23 Figure 5 7: 95% Monte Carlo confidence interval for risk of failure against the number of simulations,
compared to the confidence interval for N=10 000.

After a sufficient number of simulations, the confidence interval will be narrow enough and the user will have a fairly
good idea of the risk of failure, pf, which depends only on the system being modelled and the uncertainties
considered for the parameters.

If the final value of pf is too high to be accepted, the user can try to reduce epistemic uncertainties, for instance, by
performing specific measurements of selected uncertain parameters, in order to compute more realistic risk
estimation [130]. For that, the user should focus on the uncertain parameters that affect the result most. For the
above example, Figure 5-23 shows that the final calculated risk is around 3-3.5%. This risk is associated with the
considered aleatory (residual flux, CB switching times) and epistemic (the rest of the uncertain parameters)
uncertainties. If this risk is unacceptable (i. e., the user cannot accept 3-3.5% chance of the transformer being
damaged at the energization), then the user may consider reducing the epistemic uncertainties to perform a more
realistic risk calculation. For that, the generator reactance will be one of the target parameters for uncertainty
reduction, as it has a high influence on the results: for instance, if the generator reactance is taken at its lowest
value (L’’d,gen=0.35±15%), 0.3 p.u., the risk of failure (for 1000 simulations) is pf=0%, whereas if the reactance is
taken at its highest value, 0.4 p.u., the risk of failure is pf=56%.

Page 69
Transformer Energization in Power Systems: A Study Guide

CHAPTER 6 CONCLUSIONS
This study guide provides information and suggests tools and/or methodologies for practising power systems
engineers in order to study and analyse voltage-related problems created by the inrush currents during transformer
energization. The problems of interest are the transient RMS-voltage drops and the temporary overvoltages.

A comprehensive description of the inrush current phenomenon and the characteristics of these currents is
provided. It is shown that the inrush currents depend on the residual flux in the transformer core and the closing
instants of the circuit breaker poles. Due to transformer saturation, these currents are not sinusoidal but have a rich
harmonic content, especially in the lower-order harmonics. The inrush currents progressively decay to zero with a
time-constant that depends on the damping of the system. However, the harmonic spectrum changes with time and
the maximum of each harmonic component is reached at a different time.

Furthermore, the phenomenon of sympathetic interaction between transformers is explained. Whenever a


transformer is energized, the asymmetry of the generated inrush currents can drive the nearby already energized
transformers into saturation. This can happen if the other transformers are connected at the same voltage-level
(parallel sympathetic interaction) or if they are connected at a different voltage-level (series sympathetic
interaction). As a consequence, the transient can last longer than without sympathetic interaction and the total
inrush current from all of the transformers tends to be symmetrical.

The phenomenon of “pseudo inrush” can occur during the recovery process following a voltage sag event, such as
after the clearance of a fault. The “pseudo inrush” magnetizing currents in all transformers connected to the system
will add together causing voltage drops across the system impedance, slowing the voltage recovery and distorting
the system voltage waveform. This problem, which is generally more severe in weak systems, may lead to tripping
by under-voltage and over-current relays.

The mechanisms by which the RMS-voltage drop and the temporary overvoltages are generated and the system
conditions under which these phenomena may appear are described. The inrush current is supplied from the
system sources such as generators and large motors and flows through the network impedance to the transformer
being energized. Consequently, there is a voltage drop across the network impedance and a drop in the line
voltages where the effect increases in the direction towards the transformer. Unlike faults, transformer energization
is a planned operational event for which the associated voltage sags must be limited to such levels that can be
accepted by customers with voltage-sensitive equipment or processes.

The temporary overvoltages (TOV) during transformer energization are generated by the interaction of the
harmonic components of the inrush currents with the resonances in the system. They are a function of the short
circuit strength of the system, its topology, the shunt capacitances of lines and cables and the connected load. In
the case where the system has low frequency parallel resonances (high impedance amplitude), harmonic voltages
will be generated and superimposed on the power frequency voltage, thus leading to voltage distortion and a
temporary overvoltage that will decay according to the decay of the harmonic currents involved. The fact that the
individual harmonic currents vary over time as the inrush current decays explains one of the most typical
characteristics of the overvoltages generated as a consequence of transformer energization. The maximum
overvoltages often occur during the decay of the inrush current and not at the moment of energization. The
occurrence of TOV is of particular concern during network restoration following a system wide blackout because,
during such times, the network tends to have much lower resonant frequencies and higher impedances as well as
less damping due to light loading conditions.

Extensive information on some useful techniques that can be implemented to mitigate the RMS-voltage drop and/or
the temporary overvoltages is also provided. Controlling the closing times of the energizing circuit breaker (point-
on-wave closing) is the most effective technique because the whole transient can be virtually eliminated: the inrush
current, the RMS-voltage drop and the TOV. However, to achieve this degree of success, the technique requires
that the energizing breaker has independent pole breakers, has stable and repeatable closing characteristics and
that a special purpose controller be used which calculates the residual flux in the core.

Page 70
Transformer Energization in Power Systems: A Study Guide

Among other techniques, reducing the system voltage and/or adjusting the on-load tap before energizing the
transformer are good ways to reduce the inrush current and its consequences. Selecting a low impedance path
from the network to the transformer, de-tuning the parallel resonance in that path, and adding as much load as
possible before energizing the transformer are other effective ways to reduce the TOV.

A whole chapter is dedicated to modelling and it provides exhaustive guidelines to simulate both RMS voltage drop
and TOV and to assess their probability of exceeding pre-defined levels. It is important to first define the study
zone, where the system must be modelled in detail. Besides the transformer of interest and the circuit breaker
energizing it, the study zone includes the equipment in the same or neighbouring substations that may interact with
the harmonic components in the inrush currents, thereby producing distortion voltages. Such equipment includes
other transformers, reactors and capacitors, synchronous compensators, generators and their AVR/exciters, loads
(including significant motors), and the lines or cables between them. For the rest of the system, a network
equivalent may be used in order to reduce the modelling time and the computational burden.

The most important component when simulating a transformer energization transient is the transformer model.
While the short-circuit characteristic is important, the most important aspect to be represented is the behaviour
when the core is saturated. This highly depends on the transformer core type whether single-phase or three-phase
core, and, for three-phase cores, whether the construction is 3-leg core or 4/5 leg/shell cores. Precise instructions
are provided to use the standard models, depending on the core type and the available data.

Guidance is also provided on how to model the rest of the components of the study zone such as the circuit
breakers, overhead lines, cables, shunt reactors, generators and loads.

As for the network equivalent for the external zone, the model to be used depends on the transient phenomena to
be studied and the available data on the system. For RMS-voltage drop studies in networks where no resonant
overvoltages are generated, a simple power frequency network model can be used. When the frequency response
of the network is of concern, a more detailed model of the system is required. If the frequency response of the
network is known then, a network equivalent can be constructed. A single resonance network equivalent is easy to
build but it may oversimplify the frequency response. For better accuracy, a frequency dependent network
equivalent (FDNE) should be used, but this requires a higher level of expertise. Another option particularly
important if the user is not familiar with network equivalents or if no information on the frequency response of the
network is available is to build a detailed model of the network in the neighbourhood of the transformer to be
energized.

A specific section has been dedicated to the assessment of the potential damaging effects of the temporary
overvoltages. It is shown that, for phase-to-ground TOV lasting a few seconds or less, the surge arresters
connected between phase and ground to protect the transformers from switching and lightning surges are generally
the most vulnerable equipment. For longer duration TOVs, the most vulnerable equipment to phase-to-ground TOV
generally is power transformers. Detailed information has been provided on how to evaluate the TOV withstand of
both the surge arresters and the power transformers.

A large section is devoted to the treatment of uncertainty in energization studies. Indeed the uncertainty of those
parameters whose variation may have an important influence on the outcome needs to be considered. Regarding
the inrush currents, they are highly dependent on two random parameters: the residual flux in the core before the
energization and the circuit-breaker point-on-wave closing times. The residual flux depends mainly on the last de-
energization transient (but may also be affected by CB if equipped with grading capacitors), which in general is
unknown, whereas the circuit-breaker switching times can occur at any time in the sine wave and in general the
three poles do not close simultaneously because of the mechanical pole spread. Another parameter which may
also be taken into account with uncertainty affecting the inrush currents is the air-core inductance of the
transformer. It is generally calculated by manufacturers with a given accuracy. If the system resonant frequencies
are low, then the uncertainties in the capacitive and inductive components of the system must be taken into
account. Very small variations of these parameters can tune the system resonances to the inrush current harmonic
frequencies and thus lead to higher resonant overvoltages. Due to their dependency on the uncertain parameters,
the currents and voltages are not deterministic but rather stochastic variables. Therefore, the output of the

Page 71
Transformer Energization in Power Systems: A Study Guide

simulation study is not a single figure but a probability distribution. Often, however, the user is not interested in the
whole distribution of the output but in the risk of exceeding a particular threshold limit, for instance a given RMS-
voltage drop limit or the equipment overvoltage withstand limit. Practical techniques have been described in order
for the practising engineers to be able to compute this risk, which in general is the final result of the engineering
study.

Page 72
Transformer Energization in Power Systems: A Study Guide

Bibliography/References

[1] L. F. Blume, A. Boyajian, G. Camilli, T. C. Lennox, S. Minneci and V. M. Montsinger, Transformer engineering
: a treatise on the theory, operation, and application of transformers, 2nd red., New York, N.Y.: John Wiley
and Sons, Inc., 1951.

[2] G. Sybille, M. Gavrilovic, J. Belanger and V. Q. Do, «Transformer saturation effects on EHV system
overvoltages,» IEEE Trans. on Power Apparatus and Systems, vol. 104, nr. 3, pp. 671-680, March 1985.

[3] R. Yacamini and A. Abu-Nasser, «Transformer inrush currents and their associated overvoltages in HVDC
schemes,» IEE Proc., vol. 133 pt. C, nr. 6, pp. 353-358, Sept. 1986.

[4] H. S. Bronzeado, P. B. Brogan and R. Yacamini, «Harmonic analysis of transient currents during sympathetic
interactions,» IEEE Trans. on Power Systems, vol. 11, nr. 4, pp. 2051-2056, November 1996.

[5] H. S. Bronzeado, «Transformer Interaction Caused by Inrush Current,» University of Aberdeen (Scotland),
MSc Thesis, 1993.

[6] H. S. Bronzeado and R. Yacamini, «Sympathetic Interaction Between Power Transformers,» Proc. of 29th
Universities Power Engineering Conference -UPEC, vol. 1, pp. 236-239, Sept. 1994.

[7] C. D. Hayward, «Prolonged Inrush Currents with Parallel Transformer Affect Differential Relaying,» AIEE
Trans., vol. 60, pp. 1096-1101, Jan. 1941.

[8] H. R. Puent, M. L. Burges, E. V. Larsen and H. Elai-IL, «Energization of Large Shunt Reactors Near Static
Var Compensators and HVDC Converters,» IEEE Trans. on Power Delivery, vol. 4, nr. 1, pp. 629-635, Jan.
1989.

[9] D. Povh and W. Schultz, «Analysis of Overvoltages Caused by Transformer Magnetizing Inrush Current,»
IEE Trans. on Power Apparatus and System, Vol. %1 av %2PAS-97, nr. 4, pp. 1355-1365, Jul./Ago. 1978.

[10] M. Rioual and M. Sow, «Study of the sympathetic interactions when energizing transformers for wind-farms:
Description of the phenomena involved and determination of the stresses during their energization,» IEEE
Workshop on Wind Integration in Power Systems, Madrid, Spain, 2008.

[11] E. Styvaktakis, M. Bollen and I. Gu, «Transformer Saturation after a Voltage Dip,» Power Engineering
Letters, IEEE Power Engineering Review, vol. 20, nr. 4, pp. 62-64, Apr. 2000.

[12] IEEE Project Group 1159.2, «Test data set up by Morgan, R.L.».

[13] R. Yacamini and H. Bronzeado, «Transformer inrush calculations using a coupled electromagnetic model,»
IEE Proc. - Science, Measurements and Technologies, vol. 141, nr. 6, pp. 491-498, Nov. 1994.

[14] H. Bronzeado, J. C. de Oliveira and A. L. A. Vilaça, «Transformer Saturation after a Voltage Sag Recovery,» i
International Conference on Power System Transients (IPST), Rio de Janeiro, 2001.

[15] «Planning Limits for Voltage Fluctuations Caused by Industrial, Commercial and Domestic Equipment in the
United Kingdon,» i Engineering Recommendation P28, Issue 1, 1989.

[16] «Federal Information Processing Standards Publications (FIPS) no. 94».

Page 73
Transformer Energization in Power Systems: A Study Guide

[17] M. Nagpal, T. Martinich, A. Moshref, K. Morison and P. Kundur, «Assessing and Limiting Impact of
Transformer Inrush Current on Power Quality,» IEEE Trans. On Power Delivery, vol. 21, nr. 2, pp. 890-896,
April 2006.

[18] D. Durbak, «Temporary overvoltags following transformer energizing,» Siemens PTI Newsletter, nr. 99, pp. 1-
3, Sept. 2006.

[19] M. Adibi, R. Alexander and B. Avramovic, «Overvoltage control during restoration,» IEEE Trans. on Power
Systems, vol. 7, nr. 4, pp. 1464-1470, Nov.1992.

[20] J. Bowles, «Overvoltages in HVDC transmission systems caused by transformer magnetizing inrush
currents,» IEEE Trans. on Power Apparatus and Systems, vol. 93, nr. 1, pp. 487-493, Jan./Feb. 1974.

[21] D. Lindenmeyer, H. Dommel, A. Moshref and P. Kundur, «Analyis and control of harmonic overvoltages
during system restoration,» International Conference on Power System Transients (IPST), Budapest, 1999.

[22] G. Morin, «Service restoration following a major failure on the Hydro-Quebec power system,» IEEE Trans. on
Power Delivery, vol. 2, nr. 2, pp. 454-463, April 1987.

[23] C. Cheng and S. Chen, «Simulation of resonance over-voltage during energization of high voltage power
network,» Electric Power Systems Research, vol. 76, pp. 650-654, 2006.

[24] A. Ketabi, A. Ranjbar and R. Feuillet, «Analysis and control of temporary overvoltages for automated
restoration planning,» IEEE Trans. on Power Delivery, vol. 17, nr. 4, pp. 1121-1127, Oct. 2002.

[25] O. Bourgault and G. Morin, «Analysis of a harmonic overvoltage due to transformer saturation following load
shedding on Hydro-Quebec - NYPA 765 kV interconnection,» IEEE Trans. on Power Delivery, vol. 5, nr. 1,
pp. 397-405, Jan. 1990.

[26] M. Rioual, H. Digard, X. Legrand, C. Martin, H. Ito and Y. Corrodi, «Field application of a synchronous
controller based on the measurements of the residual fluxes for the energization of hydraulic step-up
transformers,» IEEE General Meeting, Detroit, Jul 2011.

[27] H. Ito, «Controlled switching technologies, state-of-the-art,» Proc. Transmission and Distribution Conference
and Exhibition 2002: Asia Pacific. IEEE/PES, vol. 2, p. 1455–1460, Oct. 2002.

[28] J. H. Brunke and K. J. Frohlich, «Elimination of transformer inrush currents by controlled switching. I.
Theoretical considerations,» IEEE Trans. Power Del., vol. 16, nr. 2, p. 276–280, April 2001.

[29] J. H. Brunke and K. J. Frohlich, «Elimination of transformer inrush currents by controlled switching. II.
Application and performance considerations,» IEEE Trans. Power Del., vol. 16, nr. 2, p. 281–285, April 2001.

[30] U. Krusi, K. Frohlich and J. H. Brunke, «Controlled transformer energization considering residual flux-
implementation and experimental results,» Proceedings of the IASTED International Conference Power and
Energy Systems, Anaheim, CA, USA, 2001.

[31] H. Bronzeado, S. O. Pinto, P. Jonsson, J. C. d. Oliveira and M. L. R. Chaves, «Transformer Inrush is Over:
An Experience with a 100MVA, 230/138 kV Three-phase Transformer Controlled Energizing,» International
Conference on Power Systems Transients (IPST), Kyoto, 2009.

[32] A. Ebner, M. Bosch and R. Cortesi, «Controlled switching of transformers - effects of closing time scatter and
residual flux uncertainty,» Proc. 43rd International Universities Power Engineering Conference UPEC 2008,

Page 74
Transformer Energization in Power Systems: A Study Guide

Sep. 2008.

[33] CIGRE WG A3.07, «Controlled switching of HVAC circuit breakers - Planning, Specifications and Testing,»
CIGRE TB 263, Jan. 2004.

[34] H. S. Bronzeado, J. C. Oliveira, R. Apolonio and A. B. Vasconcellos, «Transformer inrush mitigation - Part I:
modeling and strategy for controlled switching,» Cigré international technical colloquium C. SCA3, Rio de
Janeiro, Brazil, Sep. 2007.

[35] H. S. Bronzeado, S. O. Pinto, J. C. Oliveira, M. L. R. Chaves and P. Jonsson, «Transformer inrush mitigation
- Part II: Field test on a 100MVA threephase transformer,» Cigré international technical colloquium, C. SCA3,
Rio de Janeiro, Brazil, Sep. 2007.

[36] A. Mercier, E. Portales, Y. Filion and A. Salibi, «Transformer controlled switching taking into account the core
residual flux - a real case study,» CIGRE Session, paper 13-201, 2002.

[37] H. Kohyama, K. Kamei and H. Ito, «Application of controlled switching system for transformer energization
taking into account a residual flux in transformer core,» CIGRE SC A3&B3 Joint Colloquium, Tokyo, 2005.

[38] H. S. Bronzeado and J. C. de Oliveira, «The influence of tap position on the magnitude of transformer inrush
current,» International Conference on Power System Transients (IPST), Budapest, 1999.

[39] P. C. Y. Ling and A. Basak, «Investigation of magnetizing inrush current in a single-phase transformer,» IEEE
Trans. Magn., vol. 24, nr. 6, p. 3217–3222, Nov. 1988.

[40] S. G. Abdulsalam and W. Xu, «A sequential phase energization method for transformer inrush current
reductiontransient performance and practical considerations,» IEEE Trans. Power Del., vol. 22, nr. 1, p. 208–
216, Jan. 2007.

[41] L. Prikler, G. Banfai, G. Ban and P. Becker, «Reducing the magnetizing inrush current by means of controlled
energization and de-energization of large power transformers,» Electric Power Systems Research, vol. 76, nr.
8, p. 642–649, May 2006.

[42] N. Chiesa and H. K. Høidalen, «Novel Approach for Reducing Transformer Inrush Currents: Laboratory
Measurements, Analytical Interpretation and Simulation Studies,» IEEE Trans. Power Del., vol. 25, nr. 4, pp.
2609-2616, Oct. 2010.

[43] M. Martínez Duró, F.-X. Zgainski and B. Caillault, «Transformer Energization Studies with Uncertain Power
System Configurations,» International Conference on Power System Transients (IPST), Vancouver, 2013.

[44] CIGRE TB 39, «Guidelines for Representation of Network Elements when Calculating Transients,» CIGRE
WG 33.02, 1990.

[45] IEC TR 60071-4:2004, «Insulation co-ordination – Part 4: Computational guide to insulation co-ordination and
modelling of electrical networks,» 2004.

[46] R. Iravani (chair), A. Chandhury, I. Hassan, J. Martinez, A. Morched, B. Mork, M. Parniani, D.


Shirmohammadi and R. Walling, «Modeling Guidelines for Low Frequency Transients,» Modeling and
Analysis of System Transients using Digital Systems, IEEE Special Publication, TP-133-0, 1998.

[47] CIGRE WG 33.10, «Temporary overvoltages. Test case results,» Electra, vol. 188, 2000.

Page 75
Transformer Energization in Power Systems: A Study Guide

[48] N. Chiesa, A. Avendano, H. K. Høidalen, B. A. Mork, D. Ishchenko and A. P. Kunze, «On the ringdown
transient of transformers,» International Conference on Power System Transients (IPST), Lyon, France,
2007.

[49] M. R. Iravani, A. K. S. Chaudhary, W. J. Giesbrecht, I. E. Hassan, A. J. F. Keri, K. C. Lee, J. A. Martinez, A.


S. Morched, B. A. Mork, M. Parniani, A. Sharshar, D. Shirmohammadi, R. A. Walling and D. A. Woodford,
«Modeling and analysis guidelines for slow transients. III. The study of ferroresonance,» IEEE Trans. Power
Del., vol. 15, nr. 1, p. 255–265, 2000.

[50] J. A. Martinez and B. A. Mork, «Transformer modeling for low- and mid-frequency transients - a review,»
IEEE Trans. Power Del., vol. 20, nr. 2, p. 1625–1632, 2005.

[51] H. W. Dommel, S. Bhattacharya, V. Brandwajn, H. K. Lauw and L. Marti, Electromagnetic Transients Program
Reference Manual (EMTP Theory Book), Portland, OR: Bonneville Power Admin., 1992.

[52] C. P. Steinmetz and E. J. Berg, Theory and calculation of alternating current phenomena., New York:
Electrical World and Engineer, inc., 1897.

[53] E. P. Dick and W. Watson, «Transformer models for transient studies based on field measurements,» IEEE
Trans. Power App. Syst., vol. 1, pp. 409-419, Jan. 1981.

[54] J. Martinez, R. Walling, B. Mork, J. Martin-Arnedo and D. Durbak, «Parameter determination for modeling
system transients-Part III: Transformers,» IEEE Trans. Power Del., vol. 20, nr. 3, pp. 2051- 2062, July 2005.

[55] X. Chen, «Negative inductance and numerical instability of the satura-ble transformer component in EMTP,»
IEEE Trans. Power Del., vol. 15, nr. 4, p. 1199–1204, Oct. 2000.

[56] T. Henriksen, «How to avoid unstable time domain responses caused by transformer models,» IEEE Trans.
Power Del., vol. 17, nr. 2, p. 516–522, 2002.

[57] V. Brandwajn, H. W. Dommel and I. I. Dommel, «Matrix representation of three-phase n-winding transformers
for steady-state and transient studies,» IEEE Trans. Power App. Syst., Vol. %1 av %2PAS-101, nr. 6, p.
1369–1378, June 1982.

[58] S. V. Kulkarni and S. A. Kapade, Transformer Engineering. Design and Practice, New York: Marcel Dekker,
Inc., 2004.

[59] N. Chiesa, B. A. Mork and H. K. Høidalen, «Transformer model for inrush current calculations: Simulations,
measurements and sensitivity analysis,» IEEE Trans. Power Del., vol. 25, nr. 4, p. 2599–2608, Oct. 2010.

[60] B. A. Mork, F. Gonzalez and D. Ishchenko, «Leakage inductance model for autotransformer transient
simulation,» International Conference on Power System Transients (IPST), Montréal, 2005.

[61] H. W. Dommel, A. Yan and S. Wei, «Harmonics from transformer saturation,» IEEE Trans. Power Del., vol. 1,
p. 209–215, 1986.

[62] N. Chiesa and H. K. Hoidalen, «Analytical algorithm for the calculation of magnetization and loss curves of
delta connected transformers,» IEEE Trans. Power Del., vol. 25, nr. 3, pp. 1620-1628, July 2010.

[63] A. Gaudreau, P. Picher, L. Bolduc and A. Coutu, «No-load losses in transformer under overexcitation/inrush-
current conditions: tests and a new model,» IEEE Trans. Power Del., vol. 17, nr. 4, pp. 1009- 1017, Oct 2002.

Page 76
Transformer Energization in Power Systems: A Study Guide

[64] M. Rioual and C. Crepy, «Modelling of transformers in saturated conditions: determination of the air-core
reactance in the air by analytical formulas and its validation by on site tests,» IEEE PowerTech, Bucarest,
Romania, 2009.

[65] M. Rioual, C. Crepy and Y. Guillot, «Determination of the air-core reactance of transformers by analytical
formulas for different topological configurations and its comparison with an electromagnetical 3D approach;
discussions,» IEEE General Meeting, Calgary, 2009.

[66] F. de Leon and A. Semlyen, «Detailed modeling of eddy current effects for transformer transients,» IEEE
Trans. Power Del., vol. 9, nr. 2, pp. 1143-1150, Apr 1994.

[67] F. de Leon and A. Semlyen, «Time domain modeling of eddy current effects for transformer transients,» IEEE
Trans. Power Del., vol. 8, nr. 1, pp. 271-280, Jan 1993.

[68] F. de Leon and A. Semlyen, «A simple representation of dynamic hysteresis losses in power transformers,»
IEEE Trans. Power Del., vol. 10, pp. 315-321, Jan. 1995.

[69] H. K. Hoidalen, N. Chiesa, A. Avendano and B. A. Mork, «Developments in the hybrid transformer model –
Core modeling and optimization,» International Conference on Power System Transients (IPST), Delft, 2011.

[70] W. Enright, O. B. Nayak, G. D. Irwin and J. Arrillaga, «An Electromagnetic Transients Model of Multi-limb
Transformers Using Normalized Core Concept,» International Conference on Power System Transients
(IPST), Seattle, 1997.

[71] W. Enright, N. Watson and O. B. Nayak, «Three-Phase Five-Limb Unified Magnetic Equivalent Circuit
Transformer models for PSCAD V3,» International Conference on Power System Transients (IPST),
Budapest, 1999.

[72] B. A. Mork, F. Gonzalez, D. Ishchenko, D. L. Stuehm and J. Mitra, «Hybrid Transformer Model for Transient
Simulation—Part I: Development and Parameters,» IEEE Trans. Power Del., vol. 22, nr. 1, pp. 248-255, Jan.
2007.

[73] B. A. Mork, F. Gonzalez, D. Ishchenko, D. L. Stuehm and J. Mitra, «Hybrid Transformer Model for Transient
Simulation—Part II: Laboratory Measurements and Benchmarking,» IEEE Trans. Power Del., vol. 22, nr. 1,
pp. 256-262, Jan. 2007.

[74] B. Mork, D. Ishchenko, F. Gonzalez and S. Cho, «Parameter Estimation Methods for Five-Limb Magnetic
Core Model,» IEEE Trans. Power Del., vol. 23, nr. 4, pp. 2025-2032, Oct. 2008.

[75] H. K. Hoidalen, B. A. Mork, F. Gonzalez, D. Ishchenko and N. Chiesa, «Implementation and verification of the
Hybrid Transformer model in ATPDraw,» Electric Power Systems Research, vol. 79, nr. 3, pp. 454-459, 2009.

[76] N. Chiesa, H. K. Hoidalen, M. Lambert and M. M. Duró, «Calculation of Inrush Currents – Benchmarking of
Transformer Models,» International Conference on Power System Transients (IPST), Delft, 2011.

[77] J. A. Martinez, B. Gustavsen and D. Durbak, «Parameter Determination for Modeling System Transients—
Part I: Overhead Lines,» IEEE Trans. Power Del., vol. 20, nr. 3, pp. 2038-2044, 2005.

[78] M. Martínez Duró, «Damping Modelling in Transformer Energization Studies for System restoration: Some
Standard Models Compared to Field Measurements,» IEEE PowerTech, Bucarest, 2009.

[79] A. Imece, D. Durbak, H. Elahi, S. Kolluri, A. Lux, D. Mader, T. McDermott, A. Morched, A. Moussa, R.
Natarajan, L. Rugeles and E. Tarasiewicz, «Modeling Guidelines for Fast Front Transients,» A. Gole, J.A.

Page 77
Transformer Energization in Power Systems: A Study Guide

Martinez-Velasco and A.J.F. Keri (Eds), Modeling and Analysis of System Transients using Digital Systems,
IEEE Special Publication, TP-133-0, 1998.

[80] J. C. Das, Transients in Electrical Systems; Analysis, Recognition, and Mitigation, New York: McGraw-Hill,
2010.

[81] CIGRE WG 33.210, «Temporary overvoltages: causes, effects and evaluation,» CIGRE Session, Paris,
France, 1990.

[82] H. W. Dommel, «Digital computer solution of electromagnetic transients in single and multiphase networks,»
IEEE Transactions on Power Apparatus and Systems, Vol. %1 av %2PAS-88, pp. 388-399, April 1969.

[83] J. Marti, «Accurate Modeling of Frequency Dependent Transmission Lines in Electromagnetic Transient
Simulations,» IEEE Trans. On Power Apparatus and Systems, Vol. %1 av %2PAS-101, pp. 147-157, 1982.

[84] J. R. Martí, H. Dommel, L. Martí and V. Brandwajn, «Approximate transformation matrices for unbalanced
transmission lines,» PSCC, 9th Power System Computation Conference, Butterworths, London, Aug 1987.

[85] CIGRE C4.502, «Power System Technical Performance Issues Related to the Application of Long HVAC
Cables,» to be published.

[86] A. Robert and T. Deflandre, «Guide for assessing the network harmonic impedance,» Electra, vol. 167, Aug
1996.

[87] D. Durbak, A. Gole, E. Camm, M. Marz, R. Degeneff, R. O’Leary, R. Natarajan, J. Martinez-Velasco, K. Lee,
A. Morched, R. Shanahan, E. Pratico, G. Thomann, B. Shperling, A. Keri, D. Woodford, L. Rugeles, V.
Rashkes and A. Sharshar, «Modeling guidelines for switching transients,» Modeling and Analysis of System
Transients using Digital Systems, A. Gole, J.A. Martinez-Velasco and A.J.F. Keri (eds), IEEE Special
Publication, TP-133-0, 1998.

[88] U. Annakkage, N. Nair, Y. Liang, A. Gole, V. Dinavahi, B. Gustavsen, T. Noda, H. Ghasemi, A. Monti, M.
Matar, R. Iravani and J. Martinez, «Dynamic system equivalents: A survey of available techniques,» IEEE
Trans. on Power Del., vol. 27, nr. 1, pp. 411-420, Jan 2012.

[89] J. Arrillaga and et al., «AC system modelling for AC filter design - An overview of impedance modelling,»
Electra, CIGRE Joint TF 36.05.02/14.03.03, vol. 164, Feb 1996.

[90] L. Marti, «Low-order approximation of transmission line parameters for frequency-dependent models,» IEEE
Trans. on Power Apparatus and Systems, vol. 102, nr. 11, pp. 3582-3589, Nov 1983.

[91] T. Noda, «Identification of a multiphase network equivalent for electromagnetic transient calculations using
partitioned frequency response,» IEEE Trans. on Power Del., vol. 20, nr. 2, pp. 1134-1142, April 2005.

[92] T. Noda, «A binary frequency-region partitioning algorithm for the identification of a multiphase network
equivalent for EMT studies,» IEEE Trans. on Power Del., vol. 22, nr. 2, pp. 1257-1258, April 2007.

[93] J. Hong and J. Park, «A time-domain approach to transmission network equivalents via Prony analysis for
electromagnetic transients analysis,» IEEE Trans. on Power Systems, vol. 10, nr. 4, pp. 1789-1797, Nov
1995.

[94] H. Singh and A. Abur, «Multiport equivalencing of external systems for simulation of switching transients,»
IEEE Trans. on Power Del., vol. 10, nr. 1, pp. 374-382, Jan 1995.

Page 78
Transformer Energization in Power Systems: A Study Guide

[95] A. Abur and H. Singh, «Time domain modelling of external systems for electromagnetic transients programs,»
IEEE Trans. on Power Systems, vol. 8, nr. 2, pp. 671-679, May 1993.

[96] A. Morched, J. Ottevangers and L. Marti, «Multi-port frequency dependent network equivalents for the
EMTP,» IEEE Trans. on Power Delivery, vol. 8, nr. 3, pp. 1402-1412, July 1993.

[97] V. Do and M. Gavrilovic, «An iterative pole removal method for synthesis of power system equivalent
networks,» IEEE Trans. on Power Apparatus and Systems, vol. 103, nr. 8, pp. 2065-2070, Aug 1984.

[98] V. Do and M. Gavrilovic, «A synthesis method for one-port and multi-port equivalent networks for analysis of
power system transients,» IEEE Trans. on Power Systems, vol. 1, nr. 2, pp. 103-113, April 1986.

[99] B. Gustavsen and A. Semlyen, «Rational approximation of frequency domain responses by vector fitting,»
IEEE Trans. on Power Del., vol. 14, nr. 3, pp. 1052-1061, July 1999.

[100] B. Gustavsen, «Computer code for rational approximation of frequency dependent admittance matrices,»
IEEE Trans. on Power Del., vol. 17, nr. 4, pp. 1093-1098, Oct 2002.

[101] B. Gustavsen and A. Semlyen, «Application of vector fitting to state equation representation of transformers
for simulation of electromagnetic transients,» IEEE Trans. on Power Del., vol. 13, nr. 3, pp. 834-842, July
1998.

[102] B. Gustavsen and A. Semlyen, «Enforcing passivity for admittance matrices approximated by rational
functions,» IEEE Trans. on Power Systems, vol. 16, nr. 1, pp. 97-104, Feb 2001.

[103] D. Deschrijver, B. Gustavsen and T. Dhaene, «Advancements in iterative methods for rational approximation
in the frequency domain,» IEEE Trans. on Power Delivery, vol. 22, nr. 3, pp. 1633-1642, July 2007.

[104] Y. Vernay and B. Gustavsen, «Application of Frequency-Dependent Network Equivalents for EMTP
Simulation of Transformer Inrush Current in Large Networks,» International Conference on Power System
Transients (IPST), Vancouver, 2013.

[105] B. Gustavsen and C. Heitz, «Fast realization of the modal vector fitting method for rational modeling with
accurate representation of small eigenvalues,» IEEE Trans. on Power Delivery, vol. 24, nr. 3, pp. 1396-1405,
July 2009.

[106] B. Gustavsen and T. Noda, «Techniques for the Identification of a Linear System from its Frequency
Response Data,» Power System Transients. Parameter Determination, J.A. Martinez-Velasco (ed.), Boca
Raton (Fl), CRC Press, 2009.

[107] CIGRE WG 36-05, «Harmonics, characteristic parameters, methods of study, estimates of existing values in
the network,» Electra, vol. 77, pp. 35-59.

[108] EPRI Report TR-107332, «Evaluation of Distribution Capacitor Switching Concerns,» Electrotek Concepts,
Inc., Oct. 1997.

[109] CIGRE WG 13.01, «Practical application of arc physics in circuit breakers. Survey of calculation methods and
application guide,» Electra, vol. 117, 1988.

[110] CIGRE WG 13.01, «Application of black box modelling to circuit breakers,» Electra, vol. 194, 1993.

[111] IEC 60071-1:2006, «Insulation co-ordination – Part 1: Definitions, principles and rules,» 2006.

Page 79
Transformer Energization in Power Systems: A Study Guide

[112] IEC 60099-4:2009, «Surge arresters – Part 4: Metal-oxide surge arresters without gaps for a.c. systems,»
2009.

[113] IEEE C62.22-2009, «IEEE Guide for the Application of Metal-Oxide Surge Arresters for Alternating-Current
Systems,» 2009.

[114] A. R. Hileman, Insulation Coordination for Power Systems, Boca Raton (Fl): CRC Press, 1999.

[115] IEC 60099-5:2000, «Surge arresters – Part 5: Selection and application recommendations,» 2000.

[116] J. A. Martinez-Velasco (ed.), Power system transients: parameter determination, Boca Raton (Fl): CRC,
2010.

[117] IEC 60071-2:1996, «Insulation co-ordination – Part 2: Application guide,» 1996.

[118] P. M. Balma, R. C. Degeneff, H. R. Moore and L. B. Wagenaar, «The Effects of Long Term Operation and
System Conditions on the Dielectric Capability and Insulation Coordination of Large Power Transformers,»
IEEE Trans. Power Del., vol. 14, nr. 3, Jul 1999.

[119] N. e. a. Antonova, «Temporary overvoltages and their influence upon the insulation level of the equipment,» i
CIGRE Session, Paris, France, 1990.

[120] M. Martínez Duró and A. Tanguy, «Lack of Information about Transformer Dielectric Withstand to Harmonic
Temporary Overvoltages,» CIGRE Session, Paris, France, 2010.

[121] M. Martínez Duró and R. Denis, «Parameter uncertainty assessment in the calculation of the overvoltages
due to transformer energization in resonant networks,» CIGRE Session, paper C4-204, Paris, France, 2012.

[122] E. d. Rocquigny, N. Devictor and S. Tarantola (eds.), Uncertainty in Industrial Practice: A Guide to
Quantitative Uncertainty Management, Wiley & Sons Publishers, 2008.

[123] J. Brunke, «Elimination of Transient Inrush Currents When Energizing Unloaded Power Transformers,»
Doctoral Dissertation, Swiss Federal Institute of Technology, Zurich, 1998.

[124] R. Holmgren, R. Jenkins and J. Riubrugent, «Transformer Inrush Current,» CIGRE Session, Paris, France,
1968.

[125] D. P. Kroese, T. Taimre and Z. Botev, Handbook of Monte Carlo Methods, New York: John Wiley & Sons,
2011.

[126] P. Glasserman, Monte Carlo Methods in Financial Engineering, Springer, 2004.

[127] C. Robert and G. Casella, Monte Carlo Statistical Methods, New York: Springer-Verlag, 2004.

[128] J. Hanley and A. Lippman-Hand, «If nothing goes wrong, is everything all right? Interpreting zero
numerators,» JAMA, vol. 35, pp. 1743-1745, 1983.

[129] J. Martinez Velasco and J. Martin-Arnedo, «Voltage Sag Stochastic Prediction Using an Electromagnetic
Transients Program,» IEEE Trans. Power Del., vol. 19, pp. 596- 602, 2004.

[130] M. Martínez Duró, «Cost-effective uncertainty treatment,» CIGRE Session, Paris, France, 2012.

[131] W. Schmidt, «Ueber den einschaltstrombeidrehstromtransformatoren - Switch-on current in 3-phase a-c

Page 80
Transformer Energization in Power Systems: A Study Guide

transformers,» ElektrotechnischeZeitschrift – Edition A, vol. 82, nr. 15, p. 471–474, Jul. 1961.

[132] G. Bertagnolli, Short-Circuit Duty of Power Transformers, Second Revised Edition, Zurich: ABB, 1996.

[133] T. R. Specht, «Transformer magnetizing inrush current,» AIEE Trans, vol. 70, p. 323–328, 1951.

[134] J. F. Holcomb, «Distribution transformer magnetizing inrush current,» AIEE Trans, Part III (Power Apparatus
and Systems), vol. 80, nr. 57, p. 697–702, 1961.

[135] T. H. Fawzi and P. E. Burke, «The accurate computation of self and mutual inductances of circular coils,»
IEEE Trans Power App. Syst., vol. 97, nr. 2, p. 464–468, 1978.

[136] H. Wheeler, «Inductance formulas for circular and square coils,» Proceedings of the IEEE, vol. 70, nr. 12, pp.
1449- 1450, Dec 1982.

[137] J. Brunke and K. Fröhlich, «Elimination of transformer inrush currents by controlled switching,» IEEE Trans.
Power Del., vol. 21, nr. 2, pp. 890-896, Apr 2006.

[138] G. Kumbhar and S. Kulkarni, «Analysis of sympathetic inrush phenomena in transformers using coupled field-
circuit approach,» IEEE PES General meeting, Tampa, USA, Jun. 2007.

[139] A. Villa, «Sympathetic interaction between step-up transformers 13,2kV/400kV of MACAGUA hydroelectric
complex from CVG Edelca,» CIGRE Session, Paris, 2006.

[140] A. M. Gole, J. A. Martinez-Velasco and A. J. F. Keri, «Modeling and Analysis of System Transients Using
Digital Programs,» IEEE PES Special Publication, 1999.

[141] CIGRE WG A3.07, «Controlled switching of HVAC circuit breakers - Planning, Specifications & Testing,»
CIGRE TB 264, 2004.

Page 81
Transformer Energization in Power Systems: A Study Guide

Annexes

ANNEX A ANALYTICAL CALCULATION OF INRUSH CURRENT


Standard analytical formulas for the calculation of inrush peak and rate of decay are derived from single-phase
transformer theory. The inrush current on a three-phase transformer can be calculated analytically based on the
analytical formulas for a single-phase transformer and an empirical scaling factor. This factor accounts for the
number of phases, core construction and coupling of the transformer [1, 58, 131].

A relatively simple equation based on a sustained exponential decay of the inrush current is given by Bertagnolli
[132]:

2 2 +
( )= Eq. A-1
2 2 2
+

2
= Eq. A-2

with:
U = voltage across the winding phase [V]
RW = winding resistance [ ]
Lair core = air-core winding inductance [H]
BN =rated flux density in the core (peak value) [T]
BR = residual flux density in the core [T]
BS = saturation flux density (approx. 2.0-2.05 T) [T]
tn = time variable [s]
=time constant of exponential decay [s]

This equation is useful for rapid hand calculations due to its simplicity.

The analytical formula proposed by Specht [133] is somewhat more accurate as the decay of the DC component of
the flux (BR) is considered only during saturation (B > BS):

2 ( )
( )= 1 Eq. A-3

( )= ( 1) 2(sin cos ) Eq. A-4

with the saturation angle: = cos ( )

Page 82
Transformer Energization in Power Systems: A Study Guide

An improved analytical equation is recommended by Holcomb [134]:

2 ( )
( )= sin( ) sin( ) Eq. A-5
2 2 2
+

1
= tan Eq. A-6

where ts is the time when the core begins to saturate (B(t) >Bs). tS is calculated for the first period from:

( )= + (1 cos ) Eq. A-7

thereafter for the following period from:

( )= + (cos 0 cos ) Eq. A-8

with t0 the time when the inrush current reaches zero at each cycle calculated from Eq. A-5 for i(t0)=0. It is assumed
that the inrush current is different from zero only between tS and t0.

Eq. A-1 and Eq. A-3 calculate only the envelope of the inrush current peaks, not the actual waveform. Eq. A-5 can
be used to calculate analytically an approximate waveform of the inrush current. These three approaches are
compared in Figure A-1 using a common set of parameters. The air-core inductance Lair-core of a winding can be
calculated as:

2
= 0 Eq. A-9
,

with heq_HV being the equivalent height of the winding including fringing effects. The equivalent height is obtained by
dividing the winding height by the Rogowski factor KR (< 1.0) [58]. This factor is usually determined empirically and
is a function of the height, mean diameter, and radial width of a winding. More advanced equations for the
calculation of the air-core inductance takes into account the thickness and diameter of the winding in addition to the
height for the calculation of the equivalent height [135, 136].

400
Bertagnolli
Specht
300 Holcomb
Current [A]

200

100

0
0 50 100 150 200
time [ms]

Figure A-1 Comparison of equations for analytical estimation of inrush current.

Page 83
Transformer Energization in Power Systems: A Study Guide

These types of formulas are commonly used by manufacturers for the estimation of the first peak of the inrush-
current and the rate of decay. In-house developed experience factors are used to fine tune these analytical
formulas for the design and materials of each specific transformer. In particular, those factors are used to tune the
calculation of the slope of the saturation characteristic in the fully saturated region, the maximum current, and the
rate of decay. In addition, the choice of the empirical formula for the calculation of the transformer reactance Lair-
core affects the final result. It is reported in [133] that an accuracy of 40% should be expected from analytical
formulas for the calculation of inrush current.

Page 84
Transformer Energization in Power Systems: A Study Guide

ANNEX B TRANSFORMER MODELLING: CALCULATION OF THE


SATURATION CURVE FROM THE NO-LOAD TEST
Transformer data: 66MVA 238/13.8kV star-delta transformer three-leg stacked core (60Hz)

Leakage inductance: 10% (at 66MVA)

Air-core inductance: 26% from HV winding (at 66 MVA)

Table B-1 No-load Test Report(a)

Voltage (%) Current (%) Losses (kW)

90 0.049 28.01
95 0.055 32.02
100 0.067 36.73
105 0.101 43.61
110 0.204 53.06
(a)
performed on the LV side (delta-connected)

The core equivalent resistance at 100% rated excitation is:


2
138002
= =3 = 15 554 Eq. B-1
36730
The calculation of nonlinear saturation curve is more complex. First the loss current is calculated for each excitation
point:

( ) 100 ( ) 103 100


= = Eq. B-2
(%) 3 13.8 10 3 (%)
The phase no-load current for the delta-connected winding is:

(%) (%) 66 106


= = Eq. B-3
100 100 3 13.8 103
and is calculated for each excitation point. Then the loss current is subtracted from the total no-load current to
obtain the magnetization current. Note that the loss current is resistive and the magnetization current is inductive,
therefore 90 degree shifted:

= 2 2 Eq. B-4

Page 85
Transformer Energization in Power Systems: A Study Guide

Table B-2 No-load current components calculated from test report

No-load current, Loss current, Magnetization


Voltage (%) Current (%) Losses (kW)
RMS (A) RMS (A) current, RMS (A)

90 0.049 28.01 0.787 0.752 0.235


95 0.055 32.02 0.878 0.814 0.331
100 0.067 36.73 1.061 0.887 0.583
105 0.101 43.61 1.610 1.003 1.260
110 0.204 53.06 3.264 1.165 3.049

The flux-linkage is calculated from the test voltages:

2 2 13.8 10^3 (%)


= = Eq. B-5
2 60 100
The RMS magnetization current is then converted to the peak flux-current nonlinear characteristic [62].

The final slope of the saturation curve from the LV side (for core model connected to the LV terminals) is:

26 10 3 (13.8 103 )2 1
= = = 3.67 Eq. B-6
100 66 106 2 60
Since the transformer is energized from the delta-connected winding, the triplen harmonics circulate in the delta
winding and do not contribute to the magnetization of the core. This effect should be corrected to obtain a more
accurate saturation curve. The effect of the air-core inductance and triplen harmonic correction [62] on the
saturation curve is shown in Figure B-1.

Table B-3 Saturation curve points

Flux-linkage Peak Current Peak Current(a)


(Wb-t) (A) (A)

46.5914 0.3318 0.3318


49.1798 0.8983 0.9628
51.7682 1.3082 1.4579
54.3566 3.2919 3.8583
56.9450 8.1679 9.8087
(b)
56.9778 16.3359 18.7346
(a)
with triplen-harmonic elimination [62]
(b)
additional point to set final slope

Page 86
Transformer Energization in Power Systems: A Study Guide

(a) (b)
Figure B-1 Effect of Air-core inductance on the final slope and triplen harmonic correction.
Saturation curve (a) without and (b) with triplen harmonic correction.

Page 87
Transformer Energization in Power Systems: A Study Guide

ANNEX C CASE STUDY EXAMPLES & SIMULATION RESULTS VS. FIELD


MEASUREMENTS

C.1 RMS-Voltage Drop


The example given below relates to energization of two banked generator transformers at Substation A shown in
Figure C-1 and in a more detailed single line electrical diagram in Figure C-2. The depression in voltage took place
upon closure of circuit breaker X190 while circuit breaker X290 was open and the bus coupler circuit breaker was
closed.

Sub
K

Sub D Sub
Sub L
E

Sub
Sub G M
Sub Sub Sub Sub
C B A F
Sub
N

Figure C-1 Schematic diagram of the area.

Page 88
Transformer Energization in Power Systems: A Study Guide

to Sub B to Sub G

to Sub B to Sub G

X190
X290

400/19 kV 400/21 kV 400/19 kV


345MVA 415MVA 345MVA
Ynd1 Ynd1 Ynd1

Figure C-2 Single line diagram of Substation A.

Figure C-3 shows the captured RMS voltage traces for the three phases at Substation A following the energization
of the two transformer banks. Pre-energization RMS voltages were recorded as 237.1, 236.8 and 235.1 kV for
phases A, B and C respectively and those RMS voltages at 30 ms following energization were measured as 231.7,
221.2 and 227.8 kV. The latter figures represent a voltage depression of around 2.28%, 6.59% and 3.11%
respectively. It is of particular interest to note the effect of nearby SVCs (at Substations F and L) and relocatable
SVC (at substation B) following the dip at voltage. It is important to note that RMS voltages have also been
depressed in the surrounding substations following the energization but not to the same extend as in Substation A.
The corresponding combined inrush current is shown in Figure C-4 where the peak value is 2115.9 A on phase B
(notice that at approximately eight cycles after the energization the saturation of the current transformers (CT) is
apparent in phases B and C5). It should be noted that, due to the interaction and magnetic core characteristic, the
combined peak current is not double the value expected had each transformer been energized individually (1300 to
1600 A).

5
The CT saturation can be visually detected due to the sudden variation of the DC offset in the current waveforms and the
atypical shape of the half-period that in the first periods is close to zero. CT saturation is caused by the asymmetrical current
waveforms and exhibits in a distortion of the measured currents. If the CT saturation is evident after few cycles, the value of
the current measured in the first cycle can be assumed quite accurate. When reproducing these waveforms in an EMT
program, it is of no meaning to reproduce the current waveforms beyond the first few cycles without also modeling the CTs.
This behaviour is common for general purpose substation CTs. Rogowski coils are also affected by similar issues: even though
they have a linear behaviour and do not saturate, they cannot measure DC. Only special current transducers with a
bandwidth from DC to some kHz can be used to accurately measure inrush current transients without introducing
measurements errors. Laboratory measurements performed with such current transducers shows that half inrush current
cycle remains nearly zero for the entire duration of the energization transient.

Page 89
Transformer Energization in Power Systems: A Study Guide

X: 0.11
238
Y: 237.1

X: 0.11
236 Y: 236.8

X: 0.11
234 Y: 235.1

X: 0.14
Y: 231.7
232
Voltage rms (kV)

230

X: 0.14
Y: 227.8
228

226

224

X: 0.15
222
Y: 221.2

220
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Time (s)

Figure C-3 Measured RMS voltage traces at Substation A during energization.

As the measurements can only be representative of a particular system condition, the phenomenon was modelled
using an EMT program to check for the presence of other possible worst case conditions as well as the effects of
point-on-wave switching, transformer residual flux and possible system contingencies in the area. The area is not
very strong from the perspective of the available short-circuit power. The model was extended to include the area
shown in Figure C-1. Transformers at substations B, G and F all had their saturation characteristics modelled. As
well, the SVC and mechanically switched capacitor bank connected on the tertiary of some transformers were
explicitly modelled. Various loads present in the given area were modelled as parallel R and L equivalents to
represent the MW and Mvar loading at the time. All line models, based on the geometrical data, were distributed
parameter lines. The result of the simulated RMS voltages at Substation A is given in Figure C-5. Considerable
effort was made to include the voltage control logic of SVCs to get results as close to measurements as possible,
with extremely good agreement. It should be noted that pre-energization RMS voltages on each phase are
somewhat higher than the nominal system line voltage of 400kV as it is part of the operational strategy for the area.

Sensitivity analysis for various system topologies were also checked to find the best conditions for subsequent
energization in the short term. POW switching was ruled out as it would require significant time to install. The effect
of sympathetic inrush was not found to be a major contributor to the observed voltage dips in the simulation. Best
results were obtained by setting transformer taps to their maximum value as this gives considerably less onerous
voltage dip. This, along with the use of voltage control plant in the area, was recommended for subsequent
energization.

Page 90
Transformer Energization in Power Systems: A Study Guide

Figure C-4 Combined transformer inrush current.

240
X: 0.09
Y : 237.8
238
X: 0.09
Y : 237.7
236 X: 0.09
Y : 236.7

234
Voltage RMS (kV)

232 X: 0.13
Y : 232.3

230

X: 0.13
228 Y : 229.2

226

224

X: 0.13
222 Y : 223.2

220
0.1 0.2 0.3 0.4 0.5 0.6 0.7
Time (s)

Figure C-5 Simulated RMS voltage traces at Substation A during energization.

Page 91
Transformer Energization in Power Systems: A Study Guide

C.2 RMS-Voltage Drop and Inrush Currents

C.2.1 Background
In the year 2000, a new independent power producer (IPP) with a 250 MW gas turbine generator connected to the
132 kV network, between a 150 MW hydro-electric generating station (designated JHT) and a 200 MW pulp and
paper mill (PPM) having power quality sensitive loads. There are three 89 km long 138 kV transmission lines
connecting JHT to DMR station, which is a strong bus in the utility system due to interconnection to the 500 kV
system. There are two 3.7 km 138kV lines from JHT to supply the mill. The IPP is located between JHT and the mill
and the IPP transformer is tapped into each of these two circuits. The IPP generator step-up transformer is rated
315 MVA 138/21 kV, 14.9% leakage impedance, having HV grounded-star LV delta windings. Since the IPP does
not have black-start capability, the transformer must be energized from the grid to supply auxiliary load so that the
unit can be started.

Proximity of the IPP to a very large customer, where power quality is important, necessitated the assessment of
voltage dip due to inrush current during energization of the IPP step-up transformer to determine if, and how, it
needed to be limited. The short circuit level of the system at the 132 kV point of interconnection (POI) is
approximately 1850 MVA. Using a “back-of-the-envelope” calculation [17], voltage dip was estimated as 29% as a
worst case. This indicated that energization of the IPP transformer could pose power quality problems for the
nearby mill. The utility’s restrictions on the voltage dip caused by the operation of IPPs limits the RMS voltage dip
at the POI to between 3% and 6%, on the most affected phase, with frequency of occurrence not to exceed once
per eight hours. Subject to prior approval by the utility, the IPP may be permitted to impose a voltage sag of up to
9%. It therefore became apparent that a more accurate and rigorous assessment of voltage dip due to inrush would
be necessary if the IPP was to be required to make possibly costly changes to his plant in order to comply with the
IPP interconnection requirements.

C.2.2 Model Validation


In June, 2000, as part of the IPP commissioning tests, the 315 MVA generator transformer was energized several
times by random closing of the three-phase energizing breaker during a time when the mill was shut down for
annual maintenance. The inrush currents and the 132 kV phase-to-ground voltages at the transformer bushings
were recorded on one working day during four energizations of the unloaded transformer. These were recorded at
64 samples per cycle using a metering device with remote access. Prior to the first energization, the transformer
had been in a de-energized state for eight months since leaving the factory. It was expected that the residual flux in
the core would have been determined from the previous de-energization following testing at the factory. The
highest inrush current was recorded during the third energization. These field recordings provided a useful
benchmark with which to test EMT transformer models.

An EMT model of the relevant part of the power system and the 315 MVA transformer was developed. A simplified
single-line diagram of the network model is shown in Figure C-6. Positive and zero sequence Thevenin
impedances of the power system at DMR were calculated to represent the remaining network and the
interconnection to 500 kV and the main grid, not explicitly modeled. Transmission lines were modeled as 60 Hz
constant parameter lines. The JHT generating units were modeled as ideal 60 Hz sinusoidal voltage sources
behind subtransient reactance X”d. The three-phase energizing breaker was modelled by three ideal switches.

Page 92
Transformer Energization in Power Systems: A Study Guide

Figure C-6 Simplified Single-line diagram of the system modelled in EMT program.

The IPP step-up transformer is physically a two-winding three-phase transformer having a three-leg core. The EMT
representation assumed three single-phase transformers. This representation, using inter-connected single-phase
ideal transformers, is the traditional “T”-equivalent model that has been commonly used for low frequency transient
studies. This model is not adequate for transformers that do not have a delta winding to produce inter-phase
coupling, which is not the case for the IPP generator transformer of concern here. In the EMT model, the single-
phase transformers were connected appropriately, as grounded-star on the 132 kV side and delta on the 21 kV
side. The transformer short-circuit test report from the manufacturer was used for calculating the series resistances
and leakage reactances (Rp, Xp, Rs, and Xs) shown in Figure C-7. The transformer leakage impedance was
assumed to be equally divided between primary and secondary sides of the ideal transformers. From the
transformer manufacturer’s open-circuit test of applied voltage versus exciting current, an EMT auxiliary program
was used to calculate the instantaneous flux versus instantaneous exciting current. This was input into a Type “98”
pseudo nonlinear reactor model, initially used to represent the saturable iron core (i.e. residual flux was ignored
since the saturation characteristic passes through the origin).

Figure C-7 Equivalent circuit of a single-phase T-equivalent transformer


(p.u. values refer to the transformer rating).

Page 93
Transformer Energization in Power Systems: A Study Guide

The EMT program was used to simulate the energizations of the 315 MVA generator transformer. The sinusoidal
fundamental frequency voltage sources in the system model and the closing point-on-wave were adjusted for the
simulation to reflect, as near as possible, the actual initial conditions at the IPP during the field tests prior to
energization. Comparison of simulation to measurement showed a reasonable agreement for the inrush currents
for the first energization but unacceptable match for subsequent energizations.

Lack of agreement between simulation and field tests, except for the initial energization, indicated that the effects of
residual flux on the inrush currents could not be ignored in the modelling. For the first energization, residual flux
may not have been important, therefore using a transformer model which ignored it but accounted for saturation
effects would have produced acceptable results. For the subsequent energizations, however, residual flux was
present and its effects had to be included in the modelling. The polarity and magnitude depended on the
transformer condition following the preceding de-energization.

The transformer model was then modified to include the hysteresis characteristics of the core. The Type 98 reactor
models in the T-equivalent transformers were replaced by EMT Type “96” hysteretic pseudo nonlinear reactors.
Upon request, the transformer manufacturer supplied a section of the lower part of the major hysteresis loop,
shown in the box on Figure C-8. Based on this data, the remaining section of the lower hysteresis loop was
estimated and input to the Type 96 model. By considering symmetry, the upper hysteresis characteristic is known
once the lower curve is defined. The residual flux at zero current was assumed to be the maximum flux that the
transformer would retain after de-energization. Residual flux in each hysteretic reactor was selected based on the
requirement that fluxes in the three legs must sum to zero. For the third energization, Phase A was modeled with
maximum residual flux, having negative polarity and the other two phases were modeled with half of this flux, of
positive polarity.

Figure C-8 Open circuit characteristic of the IPP transformer. The lower part of the hysteresis
characteristic (inset) was provided by the transformer manufacturer.

C.2.3 Comparison Between Simulation and Field Test for the Third Energization
Figure C-9 to Figure C-11 compare the transformer inrush currents simulated by the EMT program for the third
energization with the recordings made during the field tests for the same event. Recorded and simulated data are
represented by the blue and the red lines, respectively. This case resulted in the highest inrush current measured
for all four energizations. The crest value of Phase A current is approximately 2500 A, or 1.34 p.u., for both
simulation and field measurement. There is also good agreement with Phase B inrush current. However, the
Phase C simulated current is significantly smaller than the measured current, likely due to differences in the closing
point-on-wave for this phase and residual flux assumed for Phases B and C. Figure C-12 compares the simulated
to the measured Phase A-to-ground voltages at the HV side of the transformer.

Page 94
Transformer Energization in Power Systems: A Study Guide

C.2.4 Conclusions
The T-equivalent model for the 315 MVA two-winding three-phase generator transformer, employing Type 96
hysteretic pseudo nonlinear reactors is adequate for predicting the worst case inrush current and voltage dip due to
it.

The effects of residual flux in the core plays an important role in the magnitudes of the individual inrush currents
and must be accounted for in the transformer model.

Subsequent to the field testing, the IPP replaced his original three-pole 132 kV breaker with an independent pole
breaker and point-on-wave switching controlled by a commercially available relay. This relay estimates the residual
flux based on voltage measurements made during the previous de-energization and also accounts for the closing
characteristics of the energizing breaker to arrive at optimal closing times.

Figure C-9 Phase A inrush currents from field test (blue line) and EMT simulation (red line),
for the third energization.

Page 95
Transformer Energization in Power Systems: A Study Guide

Figure C-10 Phase B inrush currents from field test (blue line) and EMT simulation (red line),
for the third energization.

Figure C-11 Phase C inrush currents from field test (blue line) and EMT simulation (red line),
for the third energization.

Page 96
Transformer Energization in Power Systems: A Study Guide

20 000 0

15 000 0

10 000 0
so urce s id e volta ge (v olts)

5 000 0

V 1-ICP
V a-S

0
60 80 1 00 12 0 1 40 16 0 180 2 00

-5 000 0

- 10 000 0

- 15 000 0

Tim e (m ili- sec onds )

Figure C-12 Phase A-to-ground voltages at the 132 kV side of the transformer
from field test (blue line) and EMT simulation (red line), for the third energization.

Page 97
Transformer Energization in Power Systems: A Study Guide

C.3 Sympathetic Interaction: On-Site Tests Measurements and Simulations

C.3.1 Background
A new 375 km 287 kV transmission system is being planned to integrate a proposed new run-of-river 300 MW
independent power producer (IPP) into the utility’s grid. The first 335 km circuit from the grid to the point-of-
interconnection (and future switching station) will be constructed, owned and operated by the utility. The remaining
40 km circuit to the IPP main generating station will be owned and operated by the IPP. The main interconnecting
station would be a 500/287 kV substation at the end of a 450 km radial 500 kV shunt-compensated transmission
system to the “backbone” 500 kV system. An existing radial 287 kV intertie to a customer-owned 800 MW
hydroelectric generating station is terminated at the 500/287 kV interconnecting substation. This self-generation
supplies the electrical power requirements to a large aluminium plant with surplus power being exported to the grid.
It also incidentally helps to support the fault levels in the area.

The new IPP will consist of a main 200 MW generating station with four 13.8 kV hydro units connected to the
287 kV line via two 116 MVA generator transformers. In the future, there will be an additional 100 MW of
generation from two other generating plants interconnected from a 69 kV system. As is sometimes the case with
run-of-river IPPs, this one will not have black-start capability. This means, of course, that the main generator step-
up transformers have to be picked up from the grid. The electrical utility imposes restrictions on the disturbances
caused by the operation of transmission level customers and IPPs, including limits on voltage dip, voltage flicker,
and harmonic injections. As for limits on voltage dip, IPPs are limited to RMS voltage dips of between 3% and 6%,
on the most affected phase, with frequency of occurrence not to exceed once every eight hours. Subject to prior
approval by the utility, the IPP may be permitted a voltage dip of up to 9%. Voltage dip limits, applied at the point
of interconnection (POI), do not address the duration of the dip, similar to the ITIC-CBEMA [16] under-voltage
characteristic. Preliminary “back-of-the-envelope” calculations indicated a significant probability of voltage dip
problems should the IPP pick up a 116 MVA transformer by randomly closing the 287 kV breaker. It therefore
became apparent that a more rigorous assessment of voltage dip due to inrush using EMT simulations would be
necessary should the IPP be required to make possibly costly unplanned investment in new equipment in order to
comply with the utility’s interconnection requirements.

This example summarizes the EMT modelling and simulations of high transformer inrush currents, where the
traditional “T” equivalent model of the transformer has been used. The hysteresis characteristic of the transformer
core and the air core inductance assumed for the modelling was based on actual data provided by a major
transformer manufacturer for another generator transformer. Simulations demonstrated that unacceptable voltage
dip and high temporary overvoltages could be produced by uncontrolled closing of the energizing breaker when a
large intertie transformer is energized from the grid, with or without the presence of an already energized
companion transformer.

C.3.2 System Data


Figure C-13 shows a simplified electrical single-line diagram of the proposed new 287 kV transmission system from
the existing utility 500/287 kV substation to the 300 MW IPP. Not shown are the configuration of the breakers at
the line terminals and the very small 287 kV station service transformer at the POI switching station. Station “S”
denotes the existing 500/287 kV interconnecting substation having a three-phase symmetrical fault level of
2520 MVA at the 287 kV bus. Station “B” denotes the 287 kV point of interconnection (POI), located 40 km from
the IPP. Generating station “F” is the IPP 287 kV intertie station having two generator transformers, four
generators and a future interconnection to a 69 kV system containing two smaller generating plants.

Page 98
Transformer Energization in Power Systems: A Study Guide

Figure C-13 Simplified equivalent single-line diagram of 287 kV system with new IPP.

The 335 km utility-owned circuit will comprise a single flat configuration transposed line from the 500/287 kV
substation to the POI. The phase conductors will comprise double-bundle 477 MCM ACSR hawk. A 65 Mvar
287 kV reactor at the interconnecting substation and a 30 Mvar line reactor at the remote end will compensate for
most of the line-charging capacitance. Since this circuit will be operated in single-pole trip and reclose mode for
single-line-to-ground faults, the shunt reactors will be grounded through neutral reactors. A 50 Ohm series
capacitor bank will be located at Station B between the POI and the utility end of the line. The bank will provide
40% series compensation for the 335 km 287 kV transmission line. For the purposes of this study, the series
capacitor bank is assumed to be bypassed during conditions of low line current. The 40 km circuit from the POI to
the generating station F, of similar construction to the main circuit, will have two switchable 25 Mvar line reactors.
Both of these reactors would be required to be in service when the IPP generation is out of service. These reactors
will have solidly grounded neutrals. Station F will comprise both a generating facility and a transformation station.
There will be two 116 MVA 287/13.8 kV HV grounded-star LV delta generator transformers (one transformer
connects two 71 MVA generators and the second unit connects one 71 MVA and one 25 MVA generator). A
287/69 kV autotransformer to interconnect a future 69 kV system was not included in the scope of the system
model. This study is focussed on the expected severe case of energizations of the two large transformers at
Station F and the expected RMS voltage dip and recovery as well as temporary overvoltages.

Table C-1 contains the positive-sequence and zero-sequence parameters for the 335 km transmission line from
Station S to the series capacitor bank as well as the 40 km line to the IPP. Impedances are given in Ohms at
60 Hz and admittances are given in micro-Siemens (µS).

Page 99
Transformer Energization in Power Systems: A Study Guide

Table C-1 287 kV Transmission Line 60 Hz Parameters for IPP Intertie

Positive Sequence Zero Sequence


From To Length R1 X1 Y1 R0 X0 Y0
Bus(a) Bus(a) (km) Ohms Ohms µS Ohms Ohms µS

S B 335 20.59 123.3 1495. 77.69 475.3 783.

B K 40 2.46 14.7 178. 9.28 56.8 93.5


(a)
Refer to Figure C-13

Table C-2 provides the details of the 287 kV line reactors located at Stations S, B, and F. The reactors on the
335 km line are fixed whilst the two 25 Mvar reactors at the IPP are switchable, depending upon the number of
generating units on line.

Table C-2 287 kV Line Reactor Details (at 60 Hz)

Reactor Location Line Reactor Neutral Reactor

Line End Mvar Ohms/Phase Ohms

S 65 1267. 800

B 30 2746. 1600

K 2 x 25 3295. (each) Grounded Neutral

C.3.3 116 MVA Transformers Data


Table C-3 shows the basic transformer data that was provided by the IPP in his application for transmission
connection to the grid. Since the actual transformer had not been constructed, a manufacturer’s test report
providing information required for modelling the unit was not available at the time of the studies. As the core
material, dimensions of the core, and other data on the transformer were not known, the hysteresis characteristic,
saturation curve and air-core inductance were estimated, based on known data for a similar (reference) generator
transformer already connected to the grid.

Page 100
Transformer Energization in Power Systems: A Study Guide

Table C-3 116 MVA Transformer Data)

Natural Cooled (ONAN) Rating 116.5 MVA

Voltage of HV Winding 287.0 kV

Voltage of LV Winding 13.8 kV

Connection of HV Winding Grounded Star

Connection of LV Winding Delta

Pos. Seq. Leakage Impedance 14.5 %

Air-Core Inductance(a) 43.5%


(a)
Based on Manufacturer’s data for another generator
transformer

C.3.4 Transformer and Network Modelling


The 287 kV transmission system to the IPP, shown in Figure C-13, together with the 450 km 500 kV single-circuit
transmission from the main grid to Station “S” and the 287 kV system to the 800 MW generating station and
aluminium plant (details not shown in the Figure) were included in the overall EMT system model. Transmission
lines were represented as frequency-dependent lines. Phase and neutral reactors, including the two 25 Mvar
reactors at Station F, were modelled as linear inductors (the saturation characteristics of shunt reactors typically
have a knee point of around 1.5 p.u. and the air-core inductance is quite high – about one-third of the unsaturated
value). The 287 kV 50 Ohm series capacitor bank was assumed to be bypassed, as would occur in practice, since
the transmission system is lightly loaded for these studies. Surge arresters were not modelled for these
simulations considering the levels of the overvoltages compared to the arrester conduction characteristics. The
IPP generators were modelled as Park’s machines rather than the simpler representation of three-phase ideal
sinusoidal voltage sources behind d-axis subtransient reactance.

The two-winding three-phase transformers under study have three-limb cores, as is common for generator
transformers. The flux in each leg of a multi-leg core is not independent but is magnetically coupled to the other
legs. For transformer energization studies it is important to account for dynamic core flux, the effects of residual
flux in the core from the previous de-energization, and for core flux saturation. The EMT model for the IPP step-up
transformer assumed an interconnection of three single-phase transformers. Each single-phase transformer was
modelled by the traditional “T”-equivalent, employing ideal transformers where the primary and secondary windings
were connected to achieve the grounded-star winding configuration on the 287 kV side and delta configuration on
the 13.8 kV side. Hysteretic pseudo-nonlinear reactors were used to model the magnetizing characteristics of each
leg of the core. This model also enables the residual fluxes in each leg of the core to be predefined up to the limit
of the specified hysteresis characteristic, prior to switching. These nonlinear models were connected across the
grounded primary windings of the ideal transformers. The transformer leakage impedance was divided between
the primary and secondary sides of the ideal transformers. From zero-sequence considerations, the component of
leakage impedance on the LV side was brought inside the delta winding. The T-equivalent model has been used
for many years in steady state studies and some low frequency transients such as magnetizing inrush currents and
their effects.

Page 101
Transformer Energization in Power Systems: A Study Guide

Favourable comparison in the past between simulation and field recordings of inrush currents provides some
justification for use of the T-equivalent model to predict worst-case inrush current, voltage dip and temporary
overvoltages due to harmonic resonance with the network. EMT simulations of a field test, when several
energizations of a 315 MVA 138/13.8 kV generator transformer were recorded, demonstrated that the inrush
currents derived from the T-equivalent model with reasonable assumptions produced acceptable agreement with
measured worst case inrush currents [17].

No data was available for the instantaneous flux versus instantaneous exciting current hysteresis characteristic for
input to the hysteretic pseudo-nonlinear inductor model. However, this data was available for another two-winding
three-phase generator transformer: a 66 MVA 238/13.8 kV unit having a three-limb core. The peak magnetic flux
density from which the hysteresis characteristic was derived was 1.9 T. Scaling of these fluxes and the exciting
currents to account for a higher nominal primary voltage and MVA rating of the transformer being studied, the
major hysteresis characteristic for the hysteretic inductor model was calculated. Figure C-14 shows part of the
major hysteresis characteristic over a reduced range of exciting current than was actually input to the hysteresis
model. This is done to show more detail of the “loop”. The air-core inductance, which is the slope of the flux versus
exciting current over the linear part of the characteristic when the core is fully saturated, actually occurs for much
higher exciting currents than shown in the figure. As indicated on Figure C-14, this characteristic permits a
maximum residual flux in any one leg of 72% of nominal flux, which is reasonable as seen by comparison to the
data presented in Table 2.1 of [123]. The minimum value of the transformer air-core inductance was assumed to
be three times the leakage inductance, or 43%. This is in accordance with the manufacturer’s data for the 238 kV
reference transformer. Figure C-15 shows a simplified single-line diagram of the two-winding transformer model for
the generator transformer. The leakage impedance was apportioned between the primary and secondary sides of
each ideal transformer. 70% was assumed on the primary side and 30% on the secondary side. By consideration
of the actual construction of two-winding three-limb stacked core transformers, an argument can be made that it is
reasonable to place most of the transformer leakage impedance on the primary side.

The parameters for the model are given in Table C-4.

Table C-4 Transformer Model Parameters

R1 = 0.50 Ohms

X1 = 72.07 Ohms

R2 = 0.0036 Ohms

X2 = 0.2142 Ohms

Rm > 1. M

N = 12.00

Page 102
Transformer Energization in Power Systems: A Study Guide

800
Max. Residual Flux
600 Peak Flux
72% of Nominal

400
Air-Core Inductance
Flux (V-s) 200 0.816 H (43.5%)

-200

-400

-600

-800
-6 -4 -2 0 2 4 6
Current (Amp)

Figure C-14 Hysteresis Characteristic assumed for the 116 MVA transformer at peak magnetic flux
density of 1.9 T. Air Core inductance is 0.816 H (43%).

Figure C-15 Model (Single-Phase Equivalent) of the 116 MVA 287/13.8 kV Two-Winding
Transformer with Hysteresis and Saturation Effects Included.

C.3.5 Simulation of Transformer Energizations Producing High Inrush Current

C.3.5.1 Case #1: Energize the First Transformer T1


Figure C-16 to Figure C-19 show a simulation of the energization of the first 116 MVA transformer, T1, from the
287 kV bus when no other transformer is connected to the bus. The residual flux assumed in the core and the
closing point on the voltage wave was selected to produce the maximum inrush current in one phase. Since the
fluxes must sum to zero in a three-legged core [137], the residual flux distribution in the core was assumed to be
+0.72 p.u., -0.36 p.u., and -0.36 p.u. in the Phase A, Phase B, and Phase C core legs, respectively. The +0.72 p.u.
residual flux corresponds to the flux at zero current of the major hysteresis characteristic (Figure C-14) and was
assumed to be a high flux that the transformer could retain after de-energization. For these residual fluxes, the
closing point-on-wave was selected so that phase A closed at the zero crossing of the voltage wave when the

Page 103
Transformer Energization in Power Systems: A Study Guide

prospective flux would add to the residual flux (i.e. no flux cancellation) to produce a high resultant flux and inrush
current in Phase A. All three poles of the energizing switches were closed simultaneously.

Figure C-16 shows that the highest inrush current (902 A) occurs in the phase having the highest residual flux.
There is considerable distortion in the currents and the DC components appear to decay very slowly following the
initial few cycles. Figure C-17 shows the three instantaneous phase-to-ground voltages at the 287 kV energizing
bus at Station F. As the figure shows, there is a 1.55 p.u. temporary overvoltage (TOV) in Phase C within 7 cycles
of energization and this overvoltage appears to be sustained throughout the remaining 800 ms of the simulation.
Figure C-18 shows a one-cycle sample of the Phase C voltage waveform at approximately 0.24 s after
energization. The distortion in the waveform is quite pronounced. FFT analysis shows that this waveform, which
repeats for many cycles, contains 53% second harmonic, 29% third harmonic and 20% fourth harmonic plus
diminishing fifth, seventh and higher harmonics. The high instantaneous overvoltage on Phase C, the most affected
phase, occurs for positive half cycles only. Finally, the individual RMS voltages calculated from the previous
instantaneous phase voltages, using a one-cycle sliding window, appear in Figure C-19. The initial voltage dip on
the most affected phase (Phase A) is 35% whereas the voltage dips on Phases B and C are slightly less than one-
half of Phase A. The behaviour of the recovery of the RMS voltages indicated on this figure is interesting. Phase A
stays below 90% of the nominal RMS voltage throughout the entire post-energization period of the simulation
whereas Phases B and C recover to pre-disturbance levels within 100 to 200 ms, and then proceed into a modest
TOV.

Some discussion of the effects of the TOV on the surge arresters protecting the IPP transformers and entrance
circuit breakers is in order. Surge arresters at Station F were not modelled because ones that are rated for a
grounded neutral 287 kV application would not have affected the magnitudes of the TOV to any noticeable degree
over the duration of the simulations. However, in reality, even though the currents conducted through the phase C
arresters during the overvoltage peaks might be very small, the cumulative effects of the energy accumulation
might not be small. Even a small energy accumulated over each cycle during a TOV persisting for many seconds
could cause the total energy to reach the energy rating of the arrester. There could then be a concern of arrester
failure. Figure C-18 shows that the TOV waveform is not a fundamental frequency sinusoid, and that the
overvoltage occurs in only one polarity. The energy accumulated in the Phase C surge arrester at the Station F
287 kV bus due to the non-sinusoidal overvoltage can be estimated from the conventional TOV capability curve
(the TOV strength factor versus time characteristic) provided by all arrester manufacturers. To apply it to this
particular case, the TOV should be assumed to be purely sinusoidal at power frequency having amplitude equal to
the maximum instantaneous Phase C to ground voltage. If the maximum instantaneous phase-to-ground voltage
per cycle is assumed to be a constant 367 kVpeak then, for a 240 kVrated 2.5kJ/kVr surge arrester, the TOV strength
factor Tr is conservatively calculated as 1.08. When no pre-stress is assumed, the TOV withstand curve indicates
that the surge arrester would absorb rated energy in about 90 seconds. If the arrester had a pre-stress energy of
2.5 kJ/kVr, rated energy would be reached in about 50 seconds. If the decay of the TOV is to be accounted for, a
more refined approach would be to employ a series of sections of sinusoidal waveforms, with each section being of
successively reduced amplitude in order to characterize the TOV.

Page 104
Transformer Energization in Power Systems: A Study Guide

No Other Transformers are Connected


1.00E+00
902 Amp
T1-Ia
9.00E-01
T1-Ib
8.00E-01 T1-Ic

7.00E-01

Transformer T1 Inrush Current (kAmp) at 287 kV Bus


6.00E-01

5.00E-01

4.00E-01

3.00E-01

2.00E-01

1.00E-01

0.00E+00

-1.00E-01

-2.00E-01

-3.00E-01

-4.00E-01

-5.00E-01

-6.00E-01
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-16 Simulation of the Transformer Inrush Currents.

No Other Transformers are Connected


5.00E+02
Va
Vb 364 kVp (1.55 pu) T.O.V.
4.00E+02 Vc

3.00E+02
287 kV Energizing Bus Ph.-Gnd Voltages (kV)

2.00E+02

Distortion Voltages:
1.00E+02 53% 2nd Harmonic
29% 3rd Harmonic
20% 4th Harmonic
0.00E+00

-1.00E+02

-2.00E+02

-3.00E+02

-4.00E+02

-5.00E+02
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-17 287 kV Bus Phase-to-Ground Voltages at the Energizing Bus.

Page 105
Transformer Energization in Power Systems: A Study Guide

4.00E+02
367 kVp (1.56 pu)

3.00E+02
Distortion Voltages:
53% 2nd Harmonic
2.00E+02 29% 3rd Harmonic
20% 4th Harmonic

Phase-to-Ground Voltage (kV)


1.00E+02

0.00E+00
Vc

-1.00E+02

-2.00E+02

-3.00E+02

-4.00E+02
0.338 0.34 0.342 0.344 0.346 0.348 0.35 0.352 0.354 0.356 0.358
Time (s)

Figure C-18 One Cycle of the 287 kV Bus Phase C Voltage During the TOV.

10.0

Two Phases are Above 1 pu (Temporary Overvoltage)


5.0

0.0
RMS Voltage Drop (%) at 287 kV Energizing Bus

-5.0
One Phase is Still Below 1 pu After 1 Second

-10.0

-15.0

-20.0

-25.0

Phase A Phase B Phase C


-30.0

-35.0 35% Voltage Drop on Phase A

-40.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-19 RMS Voltage Drop at the 287 kV Energizing Bus.

C.3.5.2 Case #2: Sympathetic Interaction #1 - Energize Transformer T2 When T1 is


Already Energized but Unloaded
An EMT program was used to simulate the energization of 116 MVA transformer T2 against the already energized
identical and unloaded T1. The model for transformer T2 was identical to the previous case for T1, including the
same residual flux distribution in the core, hysteresis characteristics and air-core inductance. However, the model
for already energized T1 was simplified by modelling the saturation characteristics, and extending this to include
the fully saturated region rather than using the hysteresis characteristic. Experimentation demonstrated that this
modelling approach was completely satisfactory. The same value of air-core inductance (three times leakage

Page 106
Transformer Energization in Power Systems: A Study Guide

inductance) was used in both transformer models. The closing point on the voltage wave was selected to produce
a high inrush current in Phase A.

Figure C-20 to Figure C-24 show the results of the simulation of sympathetic interaction. In Figure C-20 the inrush
current into transformer T2 is shown and indicates a maximum Phase A inrush current of 959 Ap, which is actually
higher than the inrush current when T1 was energized alone (902 Ap). The T1 magnetizing current during the initial
6 cycles of steady state and the subsequent inrush current resulting from energizing T2 appear in Figure C-21. As
can be seen, the highest T1 inrush current reaches 363 Ap after a delay of about 12 cycles after energizaton occurs
and is of opposite polarity to the Phase A inrush current in T2. Figure C-22 provides the sum of the T1 and T2
inrush currents which are being supplied from the 287 kV line. The maximum of this total current is 895 Ap, which
is almost the same as the peak inrush current obtained just by energizing T1 by itself. The instantaneous phase-
to-ground voltages at the energizing bus appear in Figure C-23. There is a TOV of 1.46 p.u. on Phase C which
appears 100 ms after energization and is slightly lower than 1.55 p.u. TOV seen in Case 1. Finally, Figure C-24
provides the three RMS voltages at the energizing bus before and during the inrush. The voltage dip in the
Phase A voltage is 35%, and is the same magnitude as the dip when T1 was energized by itself. The figure shows
that all three RMS voltages are below 1 p.u. during the entire post energization period (900 ms) with little indication
of recovery. In fact, the Phase A RMS voltage exhibits a declining trend after reaching a recovery to 85% of
nominal voltage 150 ms after energization. All of these voltage sags occur coincidently with a significant temporary
overvoltage in two of the instantaneous voltage waveforms.

1.00E+00

959 Amp T2-Ia


T2-Ib
8.00E-01
T2-Ic

6.00E-01
Transformer T2 Inrush Current (kAmp)

4.00E-01

2.00E-01

0.00E+00

-2.00E-01

-4.00E-01

-6.00E-01

-8.00E-01
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-20 Simulation of Transformer T2 Inrush Currents.

Page 107
Transformer Energization in Power Systems: A Study Guide

2.00E-01
T1-Ia

T1 Magnetizing T1-Ib
Current T1-Ic
1.00E-01

Transformer T1 Current (kAmp)


0.00E+00

-1.00E-01

-2.00E-01

-3.00E-01

363 Amp
-4.00E-01
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-21 Inrush Currents at the Already Energized but Unloaded Transformer T1.

1.00E+00

895 Amp Total_Ia


Total_Ib
8.00E-01
Total_Ic

6.00E-01
Total Supply Current (kAmp) to T1 & T2

4.00E-01

2.00E-01

0.00E+00

-2.00E-01

T1 magn.
-4.00E-01 Current

-6.00E-01

-8.00E-01
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-22 Total Current into T1 and T2 at the 287 kV Energizing Bus.

Page 108
Transformer Energization in Power Systems: A Study Guide

4.00E+02
Va 343 kVp (1.46 pu)
Vb
Vc
3.00E+02

2.00E+02

287 kV Energizing Bus Ph.-Gnd Voltages


1.00E+02

0.00E+00

-1.00E+02

-2.00E+02

-3.00E+02

-4.00E+02
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-23 Instantaneous Phase-to-Ground Voltages at the Energizing Bus.

10.0

5.0

All Three Phases Are Still Below 1 pu After 1 Second


0.0
RMS Voltage Drop (%) at 287 kV Energizing Bus

-5.0

-10.0

-15.0

-20.0

-25.0

Phase A Phase B Phase C


-30.0

-35.0 35% Voltage Drop on Phase A

-40.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-24 The RMS Voltages at the 287 kV Energizing Bus.

C.3.5.3 Case #3: Sympathetic Interaction #2 - Energize T2 When T1 is Already


Energized and a 25 MVA Generator is Connected
This simulation of transformer energization considers another example of sympathetic interaction. Unlike the
previous case, the already energized transformer was assumed to be carrying load when the second transformer is
energized. A simulation considered the scenario where T1 is energized and carries power from a 25 MVA
generator connected to the 13.8 kV secondary windings. Transformer T2, modelled with high residual flux in one
core leg, is energized against T1 at the most unfavourable point on the voltage wave. The models for T1 and T2
were exactly the same as for Case 2, including hysteresis characteristics and residual flux distributions in the core

Page 109
Transformer Energization in Power Systems: A Study Guide

of T2 and the saturation characteristics of T1. For the purposes of this study, the generator was modelled as a
Park’s machine but machine dynamics and the effects of governor and AVR/Exciter actions were not represented
(i.e. constant mechanical torque and constant excitation). Experimentation demonstrated that replacing the Park’s
model with a model comprising 60 Hz sinusoidal voltage sources behind d-axis subtransient reactance (X”d), would
produce incorrect results during the voltage recovery period but acceptable results during the first few cycles of the
inrush.

Figure C-25 to Figure C-29 show the results of a simulation of this second example of transformer sympathetic
interaction. In Figure C-25 the inrush currents into transformer T2 are shown and indicate a maximum
instantaneous current (in Phase A) of 1049 A. This is a significantly higher peak current than those seen in the
previous cases. Figure C-26 shows the currents at the 287 kV side of T1, including the initial 6 cycles of load
current followed by the effects of the inrush currents when T2 is energized. The total currents (the sum of T1 and
T2 inrush currents) appear in Figure C-27. Due to the presence of the generator, the maximum current is 792 Ap,
which is less than the 895 Ap of Case 2. The 287 kV instantaneous phase-to-ground voltages (in Figure C-28)
exhibits a TOV of 1.37 p.u., which is also less than for Case 2 (1.46 p.u.) and for Case 1 (1.55 p.u.). Finally, Figure
C-29 shows the three RMS voltages at the energizing bus. Voltage dip of 32% and recovery of the RMS voltages
are little improved over the case when T1 was unloaded.

Some comment should also be made on the three RMS voltages of Figure C-29. All three voltages are from 5% to
12% below the nominal voltage and one can expect that the terminal voltages of the 25 MVA generator would be
similarly depressed. While generator AVR/exciter was not modelled in the simulation, it is not unreasonable to
expect that the effect of including AVR action would have been to boost the machine terminal voltages and
therefore the 287 kV voltages, thereby aggravating the TOV. The AVR would not have reduced the TOV.

1.20E+03
T2-Ia
1049 Amp
1.00E+03 T2-Ib
T2-Ic

8.00E+02
Transformer T2 Inrush Current (Amp)

6.00E+02

4.00E+02

2.00E+02

0.00E+00

-2.00E+02

-4.00E+02

-6.00E+02

-8.00E+02
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-25 Simulation of Transformer T2 Inrush Current.

Page 110
Transformer Energization in Power Systems: A Study Guide

3.00E+02
T1-Ia
T1-Ib
T1-Ic
2.00E+02
Load
Current

1.00E+02

Transformer T1 Inrush Current (Amp)


0.00E+00

-1.00E+02

-2.00E+02

-3.00E+02

-4.00E+02
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-26 Inrush Currents at the Already Energized Transformer T1.

1.00E+03
Total_Ia
792 Amp
Total_Ib
8.00E+02 Total_Ic
Total Supply Current at 287 kV Bus to T1 and T2 (Amp)

6.00E+02

4.00E+02
Load
Current

2.00E+02

0.00E+00

-2.00E+02

-4.00E+02

-6.00E+02
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-27 The Total Current Into T1 and T2 (Supplied From the Grid).

Page 111
Transformer Energization in Power Systems: A Study Guide

4.00E+05
Va
321 kVp (1.37 pu)
Vb
Vc
3.00E+05

2.00E+05

287 kV Energizing Bus Ph.-Gnd Voltages (V) 1.00E+05

0.00E+00

-1.00E+05

-2.00E+05

-3.00E+05

-4.00E+05
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-28 Phase-to-Ground Voltages at the Energizing Bus.

10.0

5.0

0.0
All Three Phases Are Still Below 1 pu After 1 Second
RMS Voltage Drop (%) at 287 kV Energizing Bus

-5.0

-10.0

-15.0

-20.0

Phase A Phase B Phase C


-25.0

-30.0
32% Voltage Drop on Phase A

-35.0

-40.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s)

Figure C-29 RMS Voltages at the 287 kV Energizing Bus.

C.3.6 Conclusions
The worst case voltage dip due to uncontrolled closing of the energizing breaker to pick up a 116 MVA
transformer can be expected to exceed the limits that are acceptable to the utility.
A severe voltage dip, as seen in the RMS phase voltages, results when there is a high residual flux in one
of the core legs combined with a closing point on the corresponding voltage wave such that the polarity of
the residual flux is in the direction of flux build-up after switching.
The voltage dip is aggravated by the high inrush current which has to be supplied through a long radial
transmission system.

Page 112
Transformer Energization in Power Systems: A Study Guide

A temporary overvoltage results when one of the harmonic components of the inrush current excites a
parallel resonance in the supply network, as seen from the energizing bus. The resonance can cause the
waveforms to be highly distorted and this condition can be sustained for seconds.
For the simulation of transformer energization reported here, it is unlikely that the TOV excited by the
transformer inrush current will result in the surge arrester at Station F absorbing rated energy unless it had
already been highly pre-stressed prior to transformer energization.
As demonstrated by Case 1, it is possible for the RMS voltages (one or two phases) to recover from the
initial voltage dip and go into a temporary overvoltage condition.
Also demonstrated by this study is the possibility of obtaining depressed RMS voltages in the supply at the
same time as a temporary overvoltage condition on an instantaneous basis due to the effects of high
magnitude and highly distorted inrush currents having large and slowly decaying DC offsets.
Energizing a transformer at the most unfavourable point on the voltage wave against an already energized
but unloaded transformer of similar MVA rating may result in a voltage dip as large as would be obtained
by energizing the transformer by itself. The post energization recovery of the voltages following a back-to-
back energization is also significantly slower.
Energizing a transformer against an already energized transformer of similar MVA rating having generation
connected on its secondary side can result in worst case voltage dips of about the same severity as
obtained for energizing the transformer alone. Recovery of the RMS voltages back to pre-energization
levels may be even more prolonged.
The use of Park’s model is required in EMT program simulations to represent synchronous generators that
are connected to an already energized companion transformer in order to obtain the correct voltage
recovery following transformer energization. If only the first few cycles of the voltage dip are of interest,
then representing the generator as three phase ideal sinusoidal voltage sources behind X”d will produce
reasonable results.
It is not inconceivable that, for Case 3 which shows a prolonged under-voltage of the three RMS voltages
following transformer energization, the real generator’s AVR/exciter, responding to the average of the three
RMS terminal voltages, could boost the voltages and thereby aggravate the TOV appearing in the external
network.
In simulations of transformer energization, for the modelling of transformers already energized, it is not
necessary to model the hysteresis characteristics of the cores of these transformers. It is certainly
sufficient to model their saturation characteristics provided that the characteristic includes the fully
saturated region (the air-core inductance).

Page 113
Transformer Energization in Power Systems: A Study Guide

C.4 Sympathetic Interaction: On-Site Tests Measurements

C.4.1 Introduction
Several digital differential protective relay operations caused generator trips in a hydroelectric power-plant. These
relays had been recently installed both on the groups and on the transformers.
The hydroelectric power-plant has four 80 MVA units. The step-up transformers T12 and T34 are common to a pair
of units.
Onsite tests were done in March 2012 in order to characterize the conditions of the transformer or generator
disconnections, caused by the differential protective relay operation.
The inrush currents of a transformer are very dependent on initial conditions, in particular the instants of closing of
the three poles of the circuit breaker and the values of the residual fluxes in the 3 phases of the transformer. These
conditions are not controllable and each switching operation gives different inrush currents. Due to the power
system configuration, characterised by its short circuit power, its voltage level and its first resonant frequency and
parallel resonance impedance maximum value, these inrush currents can lead to resonances, with harmonic
temporary over-voltages, but also to voltage dips. It can be noted that these transients depend also on the
transformer characteristics, mainly its impedance in the saturated area, and its tap position [38].

Multiple switching operations were carried out and recorded for the transformers T12 and T34, in various operating
configurations of the power-plant:

- energising one transformer alone,


- energising a transformer with the other transformer already connected and unloaded,
- energising a transformer with the other transformer already connected and operating at different levels of
power i.e. with a hydro-unit connected and operating.

Recordings made in the two latter configurations showed sympathetic interaction between the two transformers [6,
138]. Examples are presented in the following subsections.

C.4.2 Hydro Power-plant Brief Description


The step-up transformers T12 and T34 each have three windings; 235kV/10.3kV/10.3kV. Their rated apparent
power is 160 MVA.

For each transformer, there is a 225kV 290mm², 800 m long cable between the transformer and the line circuit
breaker which is operated.

In light of the operating experience over the last 4 years, each transformer is energised approximately 350 times
per year.

C.4.3 Energizing a Transformer, the Second One Being Connected and Unloaded
It should be noted that for each test, the phase-to-ground voltages and the currents were measured at the 225 kV
terminals of each transformer. The currents were measured with a protection CT.

Figure C-30 to Figure C-33 show currents and voltages during a switching operation of transformer T12. The
transformer T34 had been previously connected and is unloaded, in steady state. Before the T12 energization, T34
magnetizing currents are a few present of rated current.

Page 114
Transformer Energization in Power Systems: A Study Guide

Figure C-30 T12 inrush currents – unloaded T34 voltages and currents.

Figure C-31 T12 switching – unloaded T34 RMS phase-to-ground voltages dips.

Page 115
Transformer Energization in Power Systems: A Study Guide

Figure C-32 T12 inrush currents – unloaded T34 voltages and currents - First 400ms starting from t0

Figure C-33 T12 inrush currents – unloaded T34 voltages and currents - 400ms starting from t0+850ms.

Transient voltage dips occur on T34 terminals, caused by the T12 inrush currents.

Page 116
Transformer Energization in Power Systems: A Study Guide

Figure C-31 shows that the phase-to-ground RMS voltage dips measured on the 225kV T34 terminals are
asymmetrical and are related to the T12 inrush currents on each phase. The dips are from 1% to 2.5%, depending
on the phase. The duration of the transients is 15 seconds.

Based on information from the TSO, the value of the short-circuit power varied from 4000 MVA to 4800 MVA during
the tests.

Due to the voltage dips induced by the switched T12 inrush currents, the already connected transformer T34
undergoes a gradual variation of its magnetic flux, with each phase different from the others, and characterized by
a strong aperiodic component. The T34 currents thus gradually increase, from their no-load (magnetizing) values to
0.45 p.u., approximately one second later, although T12 inrush peak currents reach rather low values, 1.8 p.u. in
the present case.

Figure C-34 shows that the T34 higher currents correspond to the T12 higher currents, and that the currents in T12
and T34 have opposite and alternating variations in each power frequency cycle (20ms in this case), which reveals
the interaction between the two transformers.

Figure C-34 T12 and T34 currents –100ms starting from t0+850ms.

C.4.4 Energizing a Transformer, the Second One Being Connected and Loaded
In this case, T34 is switched on while generating unit n°2 is operating and delivering power through its step-up
transformer T12. The operating point of unit n°2 is P=46 MW and Q=-5 Mvar (at the generator terminals).

The recorded transients are plotted in Figure C-35 to Figure C-37 for several time scales.

Page 117
Transformer Energization in Power Systems: A Study Guide

Figure C-35 T34 inrush currents – T12 loaded voltages and currents.

Figure C-36 T34 inrush currents – T12 loaded voltages and currents - First 1.5s starting from t0.

Page 118
Transformer Energization in Power Systems: A Study Guide

Figure C-37 T34 inrush currents – T12 loaded voltages and currents - First 200ms starting from t0.

Before T34 is energised, the currents through T12 are in steady state and symmetrical, their value being 0.28 p.u.
(peak).

Following the T34 switching, the currents through T12 become asymmetrical, with differing DC offsets among the
phases (see Figure C-37), and reach 0.36 p.u. peak. The phase-to-ground RMS voltage dips reach 3% on the most
affected phase.

The duration of the transients is less than 10 seconds.

C.4.5 Conclusions
During the field tests, differential relay operation was encountered due to the sympathetic interaction between the
two transformers, even in the case when the second transformer is loaded. The harmonic filtering and delay
parameters had to be adjusted on both the generator and transformer differential relays in order to filter the second
harmonic component more efficiently. Such activation of differential relays due to sympathetic interaction between
transformers has been previously reported in [139].

Page 119
Transformer Energization in Power Systems: A Study Guide

C.5 Controlled Switching Eliminates Inrush Current

C.5.1 Introduction
An IPP developed two run-of-river hydroelectric projects in the same vicinity in western Canada that provide a total
capacity of 20 MW. Generating Stations #1 and #2 are each connected radially to a common collector station with
another IPP. From the collector station there is a 6.1 km circuit to the utility 138 kV bus, which is the point of
interconnection for both IPPs. Generating station #1, the subject of this example, was commissioned in mid 2012
while Generation station #2 followed later. Generating Station #1 includes a single 16 MVA 13.8 kV generator, a
step-up transformer and 7 km 138 kV overhead transmission line to the collector station. Generating Station #2,
having two smaller units, is located much closer to the collector station. As a consequence of a system impact
study, the utility determined that if the main 14 MVA GSU transformer at the first generating station is energized
from the network by uncontrolled closing of the main breaker, the resulting voltage sag would likely exceed the
permissible limits on voltage sag even during normal (no-contingency) conditions. To control the transformer inrush
currents, the customer opted to use a point-on-wave (POW) circuit breaker controller and an independent-pole
breaker. This section summarizes the commissioning report of the controller for Generating Station #1. Field
measurements are presented which demonstrate that a properly designed controller applied to independent pole
breakers is capable of reliably eliminating inrush currents.

C.5.2 Commissioning of the POW Controller


Figure C-38 shows a simplified electrical single-line diagram for Generating Station #1 with the incoming 138 kV
circuit, the main circuit breaker, the 14 MVA 13.8/138 kV delta-grounded star three-leg core generator step-up
(GSU) transformer, the generator breaker and 16 MVA generator. The POW controller and the associated voltage,
current, and ambient temperature transducers are also indicated. The circuit breaker closing operation is controlled
by the POW controller. For calculation of the residual flux in the core resulting from the previous de-energization of
the transformer, the voltages across the HV windings (the line-to-ground voltages) are monitored using a sensor
connected to the capacitive HV bushing taps of the transformer. The controller was set up for remote secure
access via an Internet connection. During the commissioning tests, this enabled field personnel to consult with
remote specialists so that problems could be diagnosed and solved efficiently. This facility also provides access to
an event log and to waveforms captured during each breaker operation in the future.

Page 120
Transformer Energization in Power Systems: A Study Guide

Figure C-38 Simplified Electrical Single-Line Diagram of IPP Generating Station #1.
Interconnections to the Point-on-Wave Controller are Shown.

The set up and commissioning of the POW controller required a number of steps to determine the parameters for
input to the device so that the full capabilities of POW control could be utilized. Upon completion of the
commissioning tests, the residual flux in the core could be reliably calculated and the optimal closing time of each
pole of the circuit breaker could be determined over a wide range of ambient temperatures and residual flux
patterns in the core. The commissioning tests of the POW controller for this IPP proceeded as follows:

Circuit Breaker Timing Tests


These tests established the mechanical operating speed of each pole of the circuit breaker and the timing
relationships of the main contacts with the auxiliary (52a) contacts. The DC power supply voltage and the ambient
temperature affect the operation of the breaker and are used as inputs to the POW controller. They were monitored
during the breaker timing tests.

Static Tests
Static Tests verified that the measurements performed by the controller agree with the results obtained from
external metering.

Circuit Breaker Integration Tests


Circuit Breaker Integration Tests verified that wiring to the controller was correct and that the predicted CB
operating time was correct. The HV isolating disconnect switches were open for this test.

First Closure
First Close was used to calibrate the sensors at the capacitive bushing taps to match with the measurements from
the line CVTs. The Rate of Decrease of Dielectric Strength (RDDS) of the breaker, one of the characteristics of the
breaker that is required by the controller, was not accurately known at the time but was roughly estimated. Hence,
this test can be considered to be an uncontrolled closing. Since the residual flux in the transformer core (i.e. the leg

Page 121
Transformer Energization in Power Systems: A Study Guide

having the highest residual flux and its polarity) was unknown, the controller closed the first pole at 90 electrical
degrees (peak of the line voltage waveform) and closing of the other two phases was delayed by several half
cycles. With this strategy, the controller rotates the first pole to be closed at each operation: pole A for the first
operation, then pole B for the next closing, and so on. The first close resulted in a maximum inrush current of
2.1 p.u..

Second Closure
When the residual flux is unknown, the default strategy of the POW controller is to close the first pole at 90
electrical degrees (maximum voltage) and to delay the remaining phases by several half cycles. However, the
RDDS was unknown and the closing of the first phase was done at the wrong instant. This test can also be
considered to be a pseudo random closing. The instantaneous inrush currents, plotted in Figure C-39, indicated a
peak inrush current (Phase B) of 3.4 p.u..

Six Additional Random Energizations (Using Rough Estimate of RDDS)


Figure C-40 shows a chart summarizing the maximum instantaneous inrush currents for eight pseudo-random
energizations of the transformer from the 138 kV bus. These tests were done to calculate the RDDS of the circuit
breaker based on the measurement of the mechanical operating time and the current conduction by the arc. Since
the operation of the controller was based on the rough estimate of the RDDS, the closing was not optimum. As can
be seen, the maximum inrush current was 3.4 p.u. (for the second energization) and the average peak inrush
current was 1.5 p.u..

4
IA
3.4 p.u.
3 IB
IC
2
Current (pu)

-1

-2

-3
0 100 200 300 400 500 600
Time (mSec)

Figure C-39 Second Energization With Random Closing Resulted in Maximum Inrush Current of 3.4 p.u..

Page 122
Transformer Energization in Power Systems: A Study Guide

3.5

Peak Current (PU)


2.5

1.5

0.5

0
1 2 3 4 5 6 7 8
Energization No.

Figure C-40 Maximum Inrush Current for Eight Random Energizations. Residual Flux Unknown and Rough Estimate
for RDDS – Average Value 1.5 p.u. Current.

Closing at Maximum Voltage with New RDDS Settings


Four transformer energizations were performed using revised RDDS settings determined from the previous eight
closings. The closing of the circuit breaker was done based on the maximum voltage, not taking into account the
residual flux. Using this strategy, the controller closes the first arbitrary pole at the peak of the line voltage and
delays the closing of the remaining phases by several half cycles. The results of the four transformer energizations
are shown in the chart of Figure C-41. The maximum inrush current for these tests was about 1.4 p.u. (Phase A)
and the average peak inrush current was 0.55 p.u. or about 40% of the average value using the original RDDS
settings (Figure C-40). It can already be concluded that the use of a point-on-wave controller is beneficial for the
reduction of the inrush current, even if it is not optimised.

1.6

1.4

1.2
Peak Current (PU)

0.8

0.6

0.4

0.2

0
1 2 3 4
Energization No.

Figure C-41 Maximum Inrush Current for Four Random Energizations with Revised RDDS Settings. Residual Flux is
Unknown – Average Value 0.55 p.u. Current.

Page 123
Transformer Energization in Power Systems: A Study Guide

Normal Point-On-Wave Closing Strategy


The controller’s optimum transformer energization strategy is to first close the circuit breaker pole having the
highest residual flux. Closing occurs at the instant when the prospective flux equals the residual flux (electrical
target angle corresponding to that flux). The other phases are closed simultaneously several half cycles later. Four
energizations of the IPP 14 MVA GSU transformer were performed using the optimum closing strategy. Figure
C-42 shows a recording of about 0.5 s of the energization currents from Test #3 which produced the highest
measured current. As can be seen, the maximum current is only 0.035 p.u., which is close to the steady state
magnetizing current. Thus, the POW closing has virtually eliminated the inrush current. This should be compared to
the uncontrolled closing of Figure C-39 where the maximum inrush current is almost one hundred times larger.
Figure C-43 provides a chart of the peak inrush current for all four test energizations and demonstrates that the
optimal closing strategy produces repeatable results.
Flux
0.04
IA
0.03
IB
0.02 IC

0.01
Current (pu)

-0.01

-0.02

-0.03
0.034 p.u.
-0.04
0 100 200 300 400 500 600
Time (mSec)

Figure C-42 Optimal Closing Showing Elimination of Inrush Current.

0.04

0.035

0.03
Peak Current (PU)

0.025

0.02

0.015

0.01

0.005

0
1 2 3 4
Energization No.

Figure C-43 Maximum Inrush Current for Four Energizations with Optimised Closing.

Page 124
Transformer Energization in Power Systems: A Study Guide

C.5.3 Conclusions
A properly designed point-on-wave closing scheme, as installed for the IPP generating station, applied to
independent-pole breakers having a predictable closing characteristic, is capable of eliminating inrush current when
three-leg core transformers are energized.

If the residual flux is unknown, it is still possible to reduce the inrush current by 50 percent or more compared to the
results obtained with random closing, based on the commissioning tests for the POW scheme for the IPP. For a
transformer with grounded HV windings, the strategy is to close the first pole at maximum line voltage and
simultaneously close the remaining phases after a delay of several half cycles. This requires that breaker timing
tests and static tests have been completed and that the RDDS has been determined for each pole at various
closing angles. Effects of variation in ambient temperature and DC supply voltage to the breaker have to be
accounted for by the controller.

To achieve virtually complete elimination of the inrush current, there is the further requirement that the residual flux
in each core leg be known from the voltages measured during the previous de-energization. The first pole to be
closed is the phase having the highest residual flux. Closing occurs at the instant when the prospective flux equals
the residual flux (electrical target angle corresponding to the residual flux). The other phases are closed
simultaneously several half cycles later.

Page 125

Das könnte Ihnen auch gefallen