IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO.

2, APRIL 2012

699

A Monovoltage Equivalent Model of

Bi-Voltage Autotransformer-Based
Electrical Systems in Railways
Eduardo Pilo, Member, IEEE, Luis Rouco, Member, IEEE, Antonio Fernndez, and
Lars Abrahamsson, Member, IEEE

AbstractThis paper presents an equivalent model that allows

representing bi-voltage autotransformer-based systems (such as
2 25 kV ac, 2 15 kV ac) as if they were monovoltage systems.
This model can be used for symmetrical (such as 2 25-kV 50-Hz
systems) and unsymmetrical (such as 12/24-kV 25-Hz systems)
configurations. It is based on two simplifying hypotheses that
establish relationships between currents and voltages in the positive and negative phases. These hypotheses are discussed and the
accuracy of the model is evaluated by comparing the results with
a detailed conventional model of power-supply systems.
Index TermsPower system modeling, rail transportation
power system.

I. INTRODUCTION
RADITIONALLY, dc systems have been extensively used
in electrified railways. However, the advances in power
electronics since the 1950s made the later development of commercial high-speed railways possible in the 1980s that extended
the use of ac power-supply systems. High-speed lines involved
higher power consumptions, becoming very convenient in using
higher voltages.
However, the success of the high-speed railways led to electrical capacity saturation in some intensively used high-speed
corridors. Therefore, a newer system of using autotransformers
(AT) (in this paper, it will be referred to as bi-voltage system)
was specifically designed for lines with these high-power
requirements. In a few words, with this system, the catenary
is fed with a higher voltage which is reduced by autotransformers placed along the catenary [1], [2]. By raising the
feeding voltage, this system reduces the necessary current for
transporting the power needed by rolling stock.
More recently, the same principle has been used to implement bivoltage dc systems [3], [4], where power-electronic converters are used instead of autotransformers (see Fig. 3). SNCF
has been testing the system in a 14-km 3-MW demonstration

Manuscript received March 05, 2011; revised June 19, 2011; accepted October 23, 2011. Date of current version March 28, 2012. Paper no. TPWRD00185-2011.
E. Pilo, L. Rouco, and A. Fernndez are with the Instituto de Investigacin
Tecnolgica (IIT), ETSI de Ingeniera ICAI, Univ. Pont. Comillas de Madrid,
Madrid 28015, Spain (e-mail: eduardo.pilo@iit.upcomillas.es; luis.rouco@iit.
upcomillas.es; antonio.fernandez@iit.upcomillas.es).
L. Abrahamsson is with the School of Electrical Engineering, Royal Institute of Technology (KTH), Stockholm SE-100 44, Sweden (e-mail: lars.abrahamsson@ee.kth.se).
Digital Object Identifier 10.1109/TPWRD.2011.2179814

line and its implementation in commercial lines is now under

study. The proposed model could eventually also be used for
these dc systems.
Bi-voltage catenaries are two-phase lines. Mainly because of
the magnetic coupling, both phases have to be considered simultaneously in calculations. Furthermore, autotransformers mesh
the circuit and, thus, link voltages and currents in both phases.
For that reason, the analysis of bi-voltages circuits is normally
more complicated since the entiree circuit has to be solved to
determine its behavior. In some cases, this is a major concern,
especially when circuit analysis techniques are combined with
other techniques to formulate multidisciplinary problems (i.e.,
optimization problems).
This paper presents an equivalent model that allows representing bi-voltage systems as if they were monovoltage systems.
To do so, linear relationships are identified between voltages and
current of both phases. Then, one of the phases can be eliminated and the equivalent monovoltage model is obtained.
The resulting equivalent monovoltage model of bi-voltage
systems drastically reduces the number of buses and makes the
topology radial. In addition, linearity is conserved and the superposition principle (and all other liner circuit theorems) can
be used in the resulting model.
Without loss of generality, in this paper, it has been assumed
that each section of the electrified railway is fed only from one
substation. This hypothesis is actually true when railways are
fed from a three-phase grid, which is the most common situation. However, the principle of superposition enables also using
this model to the other cases.
The proposed model was originally designed to be included
in an optimization model to determine the topology of the
power-supply system (substation location, catenary selection,
autotransformer location, etc.) [5]. Since this model was formulated as a mixed integer programming (MIP) problem, technical
restrictions, such as voltages drops, had to be evaluated in a
simple, linear manner. Since that work, it has been widely used
for other applications requiring a simplified representation of
the power-supply system.
This paper is organized as follows. Section II introduces the
feeding systems used in railway power-supply systems. Section III describes previous models used in the literature to analyze bi-voltage systems. In Section IV, the proposed equivalent
model is presented and justified. Section V presents two study
cases used to analyze the accuracy of the proposed model. Finally, Section VI presents the conclusions of this paper.

0885-8977/$31.00 2012 IEEE

700

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

Fig. 1. Monovoltage system.

Fig. 4. Example of an asymmetrical bi-voltage system.

Fig. 2. Bi-voltage system.

is the cell where the considered train is located. The transmission cells are those which are in the section between the substation and the first autotransformer of the cell of the train. Finally,
downwards cells are those which are located further away from
the substation than the cell of the train.
III. CONVENTIONAL MODELS OF BI-VOLTAGE SYSTEMS
A. Base Magnitudes

Fig. 3. DC bi-voltage system.

II. OVERVIEW OF RAILWAYS POWER-SUPPLY SYSTEMS

Even if there are other power-supply systems, monovoltage
and bi-voltage systems are probably the most popular ones. As
shown in Fig. 1, in monovoltage systems, the conductors are
directly set to the voltage used to feed the trains. In other words,
of the overhead line is the same as the
the voltage level
used to feed the train.
voltage level
On the other hand, a bi-voltage system is derived from the
is set between
substation so that: 1) a higher voltage
is used to feed
feeding conductors and 2) a lower voltage
the trains. To ensure a proper power flow between both voltage
levels, autotransformers are distributed along the catenary (see
Fig. 2). Typical values for the distance between autotransformers are from 8 to 15 km. As normal, the central conductor
used to feed the train
is grounded, and the voltage level
is normally called positive voltage.
As mentioned in the introduction, the same principle can be
adapted to dc lines by using electronic power converters instead
of autotransformers (see Fig. 3).
In this paper, the adjective symmetrical is used when
. The 2 25-kV system used in many new
lines is a typical symmetrical bi-voltage AT-based system [6].
SEPTAs autotransformer system is a very illustrating example
of an asymmetrical bi-voltage system in which 12/24-kV
autotransformers [7] are used (see Fig. 4).
Also, the term cell refers to the portion of catenary located
between two consecutive autotransformers. The cell of the train

For simplicity, the per-unit system has been used in this paper.
In order to set up the base magnitudes, the circuit can be divided
into three zones based on their nominal voltage: 1) high-voltage
zone (transmission or distribution network); 2) positive zone;
has to be chosen and
and 3) negative zone. The base power
is common to all of the zones. (A typical value is 10 MVA.) Furthermore, base voltages have to be selected for the three zones. If
base voltages are exactly the voltages of every zone in a scenario
without any kind of load, transformation ratios take values of 1
and 1. Base impedance and base currents can be determined
from the base power and voltage of each zone.
Although conversion from real scalar magnitudes to per-unit
scalar magnitudes is trivial, some matrix magnitudes may involve different bases at the same time, in a similar way as
(1)
where
and
are the positive and negative voltages, and
are the positive and negative currents,
and
are the
and
positive and negative self-impedances, and, finally,
are the positive and negative mutual impedances, respectively.
Bold letters are used to refer to phasorial magnitudes in contrast
to normal letters used for scalar magnitudes.
In (2), per-unit voltages and currents are introduced to rewrite
(1)

(2)
where lowercase letters refer to per-unit magnitudes,
and
are the positive and negative base voltages, and
and
are the positive and negative base currents.
Once all of the base currents have been expressed as a func, (3) and (4)
tion of base voltages and apparent power

PILO et al.: MONOVOLTAGE EQUIVALENT MODEL OF BI-VOLTAGE AUTOTRANSFORMER-BASED ELECTRICAL SYSTEMS

701

Fig. 5. Phase model of bi-voltage systems.

are the final expressions for impedance matrices in the per-unit

system and for the current-voltage relation

(3)

Fig. 6. Fortescue-equivalent model of bi-voltage systems.

(4)

cuit [13]. Equations (5) and (6) are used to make this transformation

Since nominal voltages are usually chosen as base voltages,

when symmetrical bi-voltage systems are considered,
and
are normally identical. In this situation, positive
and
are
and negative phases will be balanced only if
similar, which is usually true if similar physical conductors have
been used in both phases.
When asymmetrical bi-voltage systems are considered, the
nominal negative voltage can be significantly higher than the
positive (see the example in Fig. 4). As a consequence, even
if similar physical conductors are used in the positive and the
and
may be
negative phases, per-unit self impedances
very different (see (3)) and, consequently, the system may be
quite unbalanced.

and

(5)
(6)

where

and the subindices 0 and 1 refer to the

zero- and direct-sequence magnitudes, respectively.

For all of the parts normally used in railway electrification
(catenary, transformers, etc.), the impedance matrix is always
. For that reason, when this transforsymmetrical
mation is used, (4) becomes

B. Conventional Two-Phase Models

Two different kinds of models have been used for representing two-phase circuits: 1) phase models and 2) Fortescue
equivalent models.
In phase modes, both positive and negative phases are explicitly represented. In Fig. 5, the equivalent models of the main
parts of the circuit (high-voltage grid, substation, catenary, and
autotransformer) are represented [1], [8]. In equations and figures, the symbol refers to per-length unit magnitudes.
In this model, the catenary is represented as a set of mutually coupled equivalent conductors (typically positive, negative,
and ground, which can be represented explicitly or not). The
grid is represented by its Thevenin equivalent. In order to represent substation transformers and autotransformers, standard
three-windings and two-windings transformer models are considered, including the short-circuit impedance and neglecting
the magnetizing branch. Similar models have been considered
in [9][11]. A refinement of these models has been described in
[12].
In Fortescue equivalent models, the original circuit is transformed into a zero-sequence circuit and a positive-sequence cir-

(7)
In these cases, the equivalent zero-sequence and direct-sequence model are represented in Fig. 6.
Sometimes, positive and negative phases of the catenary are
and (7) becomes
designed to be balanced
(8)
In this case, positive- and zero-sequence circuits are uncoupled and the Fortescue-equivalent representation becomes very
useful because of its simplicity.
However, frequently phases are not balanced, especially in
unsymmetrical bi-voltage systems, where positive and negative
voltages are very different [see (3)].
C. Simplified One-Phase Model
Just like the model presented in this paper, the simplified
models use one of the aforementioned two-phases models and
reduce it by grouping or neglecting elements.

702

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

Fig. 7. Typical real current loops in a bi-voltage system section.

Fig. 8. Simplified behavior of a bi-voltage system.

In [13], a simplified model is formulated based on the

Fortescue-equivalent model. As far as the catenary is concerned, both the positive and the negative phases are assumed
to be balanced and, thus, the zero-sequence and the direct-sequence circuits are supposed to be uncoupled. Based on this
hypothesis, the total impedance between the train and the
substation is calculated.
For a specific position of the train, the total impedance seen
by the train can be calculated and then transformed into phase
magnitudes as follows:
(9)
and
are, respectively, the total zero-sewhere
quence and direct-sequence impedances seen from the train.
and
can be calculated as
In the Fig. 6 example,
shown

Since a per-unit system is used, the voltage drop along a

cell in the positive and in the negative equivalent conductors has the same values and different sign.
Since typical short-circuit impedance is about 1%, autotransformers are supposed to be ideal. Therefore, there are
current flows only in the autotransformers that are immediately adjacent to the considered train.
It should be noted that currents flowing through ground conductors of the transmission cells (denoted as ) have not been
assumed to be zero.
B. Branch Currents
In the transmission cells, the voltage drops assumption can be
expressed as

(10)
(11)
where operator refers to the total impedance of two parallelconnected impedances.
Since there are no shunt branches in the positive-sequence circuit, it should be noted that the total direct-sequence impedance
depends linearly on the distance between the train and the
. On the other hand, the zero-sequence
substation
total impedance depends on the autotransformers located in
both sides from the train.
The aforementioned model is a useful and very elegant
method for obtaining a simplified representation of bi-voltage
systems. All of the elements of the circuits (transformer,
catenary, and autotransformers are accurately represented.
However, the model is quite complicated to use, especially
when calculating zero-sequence total impedance. In addition,
) may be
assuming that catenaries are balanced (and
too restrictive in many cases, and not assuming it complicates
the expressions even more due to zero-sequence and direct-sequence coupling.
IV. PROPOSED MONOVOLTAGE EQUIVALENT MODEL OF
BI-VOLTAGE SYSTEMS

(12)
Eliminating
be obtained:

from (12), the following equation can

(13)

where
(14)
In the transmission cells, the coefficient
expresses
the ratio between positive and negative currents. In many simplified models, both currents are assumed to be equal and, thus,
1. However, (14) shows the dependency toward the
catenary parameters: in quite balanced positive/negative systems, it tends to values similar to 1, while not in very unbalanced
systems (where currents tend to take the lowest impedance path,
either the positive or the negative).
In the cell of the train, the voltage drops assumption can be
expressed as

Fig. 7 shows the current loops in a bi-voltage system section

when a train is consuming power.
A. Assumptions
Two different assumptions have been used to simplify the
model (see Fig. 8) as follows.

(15)
where represents the relative position of the train in its cell
and takes values in the range
.

PILO et al.: MONOVOLTAGE EQUIVALENT MODEL OF BI-VOLTAGE AUTOTRANSFORMER-BASED ELECTRICAL SYSTEMS

Eliminating
be obtained:

from (15), the following equation can

(16)

703

C. Bus Voltages
,
,
, and
Voltage drops in the impedances
(see Fig. 5) of the traction substation can be, respectively,
expressed as

where

(29)
(30)
(31)
(32)

(17)
Furthermore, currents
and
can be calculated
consumed by the train because
from the current
(18)

Thus, the total voltage drop from the equivalent grid to the
positive terminal of the substation can be calculated as

From (17) and (18), the following equation can be obtained:

(33)

(19)
(20)
where
(21)

where
(34)
In the transmission cells, expressions of voltage drop along
the positive and negative conductors can, respectively, be obtained from (12)

(22)

(35)
(36)

From (18)(20), it can also be established that

with
(23)
(37)
By applying Kirchhoff law for currents in buses A and B (see
and expressing
and
as
Fig. 8), eliminating
(see (13))
a function of
(24)
(25)
where
(26)
(27)

is the total length of the transmission cells.

where
The equivalent impedance of the catenary
can be
written as
(38)
It should be noted that this equivalent impedance: 1) depends
only on the line parameters of the catenary and, thus, 2) is independent of the separation between autotransformers.
In the cell of the train, the voltage drop between the first autotransformer AT1 and the train itself corresponds to the first part
of (15):

In the transportation cells, the coefficient

and
express the ratio between positive or negative currents, respectively, and the current consumed by the trains. In many simplified models, both currents are assumed to be half of the current
and
0.5). However,
of the train (giving
(26) and (27) show their dependency toward the catenary parameters: in quite balanced positive/negative systems, it tends to
values similar to 0.5, and not in very unbalanced systems (where
currents tend to take the lowest impedance path, either the positive or negative).
Finally, from (26) and (27), it can be established that

where is the voltage in the bus I (see Fig. 5)

can be expressed as a function of
Voltage drop
and grouped into two terms: 1) the voltage drop that depends
only on the catenary parameters (in a similar way to (35)) and
2) the residue. The second term expresses the voltage drop due
to the distance between autotransformers. For that reason, the
second term is called AT equivalent impedance

(28)

(41)

(39)
(40)

704

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

Fig. 10. Proposed monovoltage equivalent model (1 train).

Fig. 9. Voltage drops along the positive phase.

with
(42)
In the cell of the train, the voltage drop between the first and
second autotransformers corresponds to (15). If currents
and
are expressed as a function of
, the positive and
negative voltage drops can be expressed as

(43)
(44)
It should be noted that these voltage drops depend mainly
on the distance from the train to the previous autotransformer
and the equivalent impedance of the catenary
.
Finally, in the downward cells, there are no additional voltage
drops as both positive and negative currents are zero.
Fig. 9 shows the voltage drops along the positive phase in
bi-voltage systems.
takes place in the substation.
The initial voltage drop
Then, the voltage drop along the transmission cells has a con. This slope increases when
stant slope proportional to
the cell of the train is reached, due to current concentration.
and
The new slope includes the effect of the impedance
reaches its minimum value where the train is located. At the end
of the cell of the train, the voltage level recovers the value that
was considered from the
would have been reached if just
substation to the train position. Finally, in the downward cells,
there are not any additional voltage drops.
can be calFrom a practical point of view, the term
culated or approximated in different ways, depending on the required simulation accuracy as follows.
The impedance can be calculated using the complete expression defined in (42). However, to do so, the relative
position of the train has to be known.
If the dependency on the position of the trains has to be
eliminated, several options are available:
(it can be
Calculating an upper bound
occurs when the
proven that maximum value of
). Thus, the
train is in the middle of the cell
upper bound can be calculated as follows:

Fig. 11. Superposition of different trains voltage drops (overview).

(45)
An average estimation can also be calculated by integrating (42) and the

(46)
Finally, if autotransformers are close enough to each
can even be neglected.
other, the term
As an example, Fig. 10 shows the proposed monovoltage
equivalent model obtained from the Fig. 5 case.
D. Scenarios With Several Trains
When several trains are fed from one substation, the superposition principle can be used to calculate voltages in the electric
circuit (see Fig. 11).
It has been previously established that voltages in the downward cells do not depend on the autotransformer distance, position, or technical characteristics (see (43)). Thus, the voltage
at a train can be calculated by superposing the effect of all

PILO et al.: MONOVOLTAGE EQUIVALENT MODEL OF BI-VOLTAGE AUTOTRANSFORMER-BASED ELECTRICAL SYSTEMS

705

Fig. 14. Proposed monovoltage equivalent model (several trains, same cell).
Fig. 12. Superposition of different trains voltage drops.

TABLE I
MODEL PARAMETERS OF THE CATENARY

Fig. 13. Proposed monovoltage equivalent model (several trains, different

cells).

For a train , the coupling with other trains of the same cell
can be neglected and all of the calculations done as if was
the only train in C.
In addition, if the dependency upon the position of the trains
has to be eliminated, the aforementioned assumptions (using the
upper bound, using the average value, or even neglecting the
) can also be considered.
terms
V. EXAMPLE

the trains. As can be observed in Fig. 12, the voltage drop due
depends on all of the trains sharing the cell
to the term
A. Description

(47)
is the distance between the train and the substawhere
refers to the current of train , is the total number
tion,
is the AT equivaof trains fed by the substation,
lent impedance
of train t, is the set of trains sharing
is the relative position of the train in its cell
the cell, and
(as defined in (15)).
Equation (47) corresponds to the equivalent circuits represented in Fig. 13 (when there is, at most, 1 train per cell) and
Fig. 14 (when there are several trains sharing a cell).
From a practical point of view, the total voltage drop due to
can be calculated or approximated in different
the terms
ways, depending on the required simulation accuracy:
Use the complete expression defined in (47). However, to
do so, the position of all trains of the cell has to be known.
and the min function.)
( is needed to calculate

In order to evaluate its accuracy, the proposed model is used to

perform voltage and current analysis in two different scenarios:
1) a case with only 1 train (scenario A) and 2) a case with several
trains (scenario B). These currents and voltages are compared
to the results obtained with the simulation tool SILVIA, which
implements a conventional two-phase model [14] used as the
reference. The models used in this tool to represent catenary are
actually an implementation of well-validated Electromagnetic
Transients Program (EMTP) models for overhead lines, as they
are described in [15].
The base power used for all sectors is 10 MVA. The base
voltages are 27.5 kV for the positive and negative zones and
220 kV for the high-voltage side of the substation.
The studied case is a 50-km-long bi-voltage section. The series impedance matrix per length unit (expressed in per units per
meter) of the catenary is

Table I shows the equivalent impedance

and catenary
coefficients calculated from (26)(38) for this catenary. Table II
shows the nominal characteristics of the substation transformer.
Table III shows the nominal characteristics of the autotransformers, which are located every 10 km.

706

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

TABLE II
NOMINAL CHARACTERISTICS OF THE SUBSTATION TRANSFORMER

TABLE III
NOMINAL CHARACTERISTICS OF THE AUTOTRANSFORMERS

Fig. 17. Current distribution (scenario A).

Fig. 15. Conventional two-phase model (scenario A).

Fig. 16. Equivalent monovoltage model (scenario A).

B. Scenario A
In scenario A, only 1 train has been considered, located 35 km
from the substation and consuming 20 MW 2 MVAr (which
corresponds to a double train composition).
Fig. 15 shows the topology of the two-phase circuit that is
used as the reference. (bus names have been chosen for coherency with scenario B). Fig. 16 shows the equivalent monovoltage model of the same section.
In order to analyze the accuracy of the assumption, there are
current circulations only in the autotransformers adjacent to the
train, positive phase currents have been calculated by: 1) using
,
the complete two-phase model and 2) using coefficients
,
, and
(see Fig. 17).
As described in Fig. 7, all of the autotransformers are connected in parallel and, thus, have a current loop. As expected,
currents calculated with the proposed model are slightly different than the reference currents, especially in the autotransformers located closer to the cell of the train. The difference
between both models has been found to be lower than 6% of the
current supplied to the train.
Fig. 18 compares the voltages along the positive bus voltages obtained with the reference model and the proposed model

Fig. 18. Voltage drop profile (scenario A).

(with and without

terms). This analysis intends to estimate how the differences found in current distributions impact
voltage calculation.
The analysis of voltages shows that voltage profiles obtained
with the reference model and the proposed model are almost
, voltage errors are lower
identical. When neglecting
than 0.9%, whereas with the complete model, voltage errors are
lower than 0.11%.
C. Scenario B
In scenario B, a homogeneous train distribution, with trains
located every 9 km, has been considered (see Table IV). Fig. 19
shows the topology of the two-phase circuit that is used as the
reference. Fig. 20 shows the equivalent monovoltage model of
the same section.
As for scenario A, positive phase currents have been calculated by: 1) using the complete two-phase model and 2) assuming the simplified behavior (see Fig. 21).

PILO et al.: MONOVOLTAGE EQUIVALENT MODEL OF BI-VOLTAGE AUTOTRANSFORMER-BASED ELECTRICAL SYSTEMS

707

Fig. 19. Two-phase circuit (scenario B).

Fig. 22. Voltage drop profile (scenario B).

Fig. 20. Equivalent monovoltage model (scenario B) .

TABLE IV
CURRENTS CONSUMED BY THE TRAINS (SCENARIO B)

The analysis of voltages shows that current estimation errors have a negligible impact in voltage calculations. With the
complete model, voltage errors are lower than 0.48%. When ne, voltage errors are lower than 0.8%.
glecting
VI. CONCLUSION
This paper has presented a simplified model for representing
bi-voltage systems (such as the 2 25-kV 50-Hz system) in a
similar way to simpler monovoltage systems (such as the 1
25-kV 50-Hz system). This model was originally designed to
be implemented in an optimization procedure to determine the
topology of the power-supply system. Since the optimization
was time-consuming, efficient models were required to evaluate
technical constraints, such as voltages drops in a simple, linear,
and accurate manner. Since then, the proposed model has been
shown to be faster and easier to implement than other models.
In addition, the proposed model makes bi-voltage systems
behavior more comprehensive than conventional two-phase
models. Furthermore, qualitative and quantitative analysis can
be more easily performed, without sacrificing accuracy.
ACKNOWLEDGMENT
The authors would like thank Profs. E. Lobato and I. Egido
for their support in the preparation of this paper.

Fig. 21. Current distribution (scenario B) .

It should be noted that errors are quite small as values are

always lower than 2.3% of the current of one train. Compared
with scenario A, errors are significantly lower due to cancellations when summing errors with different signs.
Fig. 22 compares the voltages along the positive bus voltages
obtained with the reference model and the proposed model (with
terms).
and without

REFERENCES
[1] J. D. Glover, A. Kusko, and S. M. Peeran, Train voltage analysis for
AC railroad electrification, IEEE Trans. Ind. Appl., vol. IA-20, no. 4,
pp. 925934, Jul. 1984.
[2] H. Roussel, Power supply for the atlantic TGV high-speed line, in
Proc. Inst. Elect. Eng. Int. Conf. Main Railway Electrif., 1989, pp.
388392.
[3] P. Ladoux, F. Alvarez, H. Caron, G. Josse, and J. P. Perret, Une nouvelle structure dalimentation des catnaires 1500 V : Le systme 2
1500 V, Rev Gnrale Des Chemins Fer, pp. 2111, Jun. 2006.
[4] T. Schtte, B. Kindell, and G. Hammarstrm, A Plant for Feeding
Railway Vehicles, Int. Patent WO 00/21777, 2000, World Intellect.
Property Organiz.

708

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 27, NO. 2, APRIL 2012

[5] E. Pilo, L. Rouco, and A. Fernandez, An optimization procedure to determine the topology of AC railways power supply networks, in Proc.
ASME Conf., Jan. 1, 2007, vol. 2007, pp. 333340.
[6] C. Gourdon and C. Herce, The overhead system of the TGV-atlantique, in Proc. Int. Conf. Main Line Railway Electrif. (Conf. Publ.),
1989, pp. 393396.
[7] R. Natarajan, T. A. Neschke, W. Naqvi, M. A. Lopez-Lopez, J. Hong,
and R. M. Bucci, Short circuit currents of the SEPTA traction power
distribution system, IEEE Ind. Appl. Mag., vol. 3, no. 6, pp. 5259,
Nov./Dec. 1997.
[8] R. J. Hill and I. H. Cevik, On-line simulation of voltage regulation
in autotransformer-fed AC electric railroad traction networks, IEEE
Trans. Veh. Technol., vol. 42, no. 3, pp. 365372, Aug. 1993.
[9] A. Mariscotti, P. Pozzobon, and M. Vanti, Simplified modeling of
2 25-kV at railway system for the solution of low frequency and
large-scale problems, IEEE Trans. Power Del., vol. 22, no. 1, pp.
296301, Jan. 2007.
[10] L. Hanmin, O. Sehchan, K. Gildong, and A. Cheonheon, A study on
the equivalent conductor representation of AC electric railway system,
in Proc. Int. Conf. Inf. Multimedia Technol., 2009, pp. 3337.
[11] A. Mariscotti and P. Pozzobon, Genetic algorithm approach for the
determination of the electrical parameters of railway traction lines,
Measurement, vol. 41, no. 2, pp. 169177, 2008.
[12] M. Brenna and F. Foiadelli, Sensitivity analysis of the constructive
parameters for the 2 25-kV high-speed railway lines planning, IEEE
Trans. Power Del., , vol. 25, no. 3, pp. 19231931, Jul. 2010.
[13] G. Varju, Simplified Method for Calculating the Equivalent Impedance
of an AT System. Budapest, Hungary: Innotech Ltd. (Innovation Park
Tech. Univ. Budapest), 1996.
[14] E. Pilo, L. Rouco, A. Fernndez, and A. Hernndez-Velilla, A simulation tool for the design of the electrical supply system of high-speed
railway lines, in Proc. IEEE Power Eng. Soc. Summer Meeting, 2000,
pp. 10531058.
[15] H. W. Dommel, Electromagnetic Transients Program Reference
Manual (EMTP Theory Book). Vancouver, BC, Canada: Bonneville
Power Admin., 1986.

Eduardo Pilo (M00) received the M.Sc. degree in

industrial engineering and the Ph.D. degree in electrical engineering from Comillas Pontifical University, Madrid, Spain, in 1997 and 2003, respectively,
from Pontifical University, Comillas, Madrid, Spain,
in 2003.
His Ph.D. thesis was titled "Optimal design of the
electrification of high-speed railways" and was focused on modeling and simulating ac traction power
systems and optimizing their topology by using a
mixed-integer programing (MIP) formulation of the
problem. From 2003 to 2010, he was a Researcher in the Institute for Research
in Technology (IIT, ICAI) in the Railways Group (ASF) and in the Power
System Modeling Group (MAC). In this period, he was responsible for the
research projects related to traction power systems. Since 2003, he has been a
Lecturer of several degrees as well as master and doctorate courses. Since 2010,

he has been the Technical Director of Multitest09, a consulting company that

focuses on the technology of power machinery and apparatus, and Associate
Professor in the Power Systems Department in ICAI.

Luis Rouco (M91) received the Ph.D. degree in industrial engineering from the Polytechnic University
of Madrid, Madrid, Spain, in 1990.
Currently, he is Full Professor of Electrical Engineering with the School of Engineering, Instituto de
Investigacin Tecnolgica (IIT), ETSI de Ingeniera
ICAI, Univ. Pont. Comillas de Madrid. He served
as Head of the Department of Electrical Engineering
from 1999 to 2005. He teaches undergraduate level
courses on electrical machines and graduate level
courses on power system analysis and power system
stability and control.
He has been Director of the Course on Power System Operation, jointly developed with Red Elctrica de Espaa from 2004 to 2007 and Director of the Master
on Electrical Technology jointly developed with Endesa from 2007 to 2011. He
develops his research activities at Instituto de Investigacin Tecnolgica (IIT).
He has supervised a number of research projects for public administrations, electric utilities, and other engineering and industrial companies. His areas of expertise are modeling, analysis, simulation and control of the steady state, as well as
dynamic and transient behavior of electric power systems.

Antonio Fernndez is a Tenured Assistant Professor

with the Instituto de Investigacin Tecnolgica (IIT),
ETSI de Ingeniera ICAI, Univ. Pont. Comillas de
Madrid, Madrid, Spain, where he is also a Research
Fellow with the Institute for Research in Technology.
His research interests include traffic regulation systems, efficiency in railways operation, and optimal
design of signaling and railways capacity.

Lars Abrahamsson (M06) was born in Lule,

Sweden, in 1979. He received the M.Sc. degree in
engineering physics from the Lule University of
Technology, Lule, Sweden, in 2005, and is currently
pursuing the Ph.D. degree in electric power systems
at the Royal Institute of Technology, Stockholm,
Sweden.