Home of the Transactions of the Wessex Institute, the WIT electroniclibrary
provides the international scientific community with immediate and permanent
access to individual papers presented at WIT conferences. Visit the WIT eLibrary
at http://library.witpress.com
WIT Press publishes leading books in Science and Technology.
Visit our website for the current list of titles.
www.witpress.com
WITPRESS
Power Supply,
Energy Management
and Catenary Problems
This page intentionally left blank
Power Supply,
Energy Management
and Catenary Problems
Editor: Eduardo Pilo
Universidad Pontificia Comillas de Madrid, Spain
Published by
WIT Press
Ashurst Lodge, Ashurst, Southampton, SO40 7AA, UK
Tel: 44 (0) 238 029 3223; Fax: 44 (0) 238 029 2853
EMail: witpress@witpress.com
http://www.witpress.com
For USA, Canada and Mexico
WIT Press
25 Bridge Street, Billerica, MA 01821, USA
Tel: 978 667 5841; Fax: 978 667 7582
EMail: infousa@witpress.com
http://www.witpress.com
British Library CataloguinginPublication Data
A Catalogue record for this book is available
from the British Library
ISBN: 9781845644987
Library of Congress Catalog Card Number: 2010920855
The texts of the papers in this volume were set
individually by the authors or under their supervision.
No responsibility is assumed by the Publisher, the Editors and Authors for any injury
and/or damage to persons or property as a matter of products liability, negligence or
otherwise, or from any use or operation of any methods, products, instructions or
ideas contained in the material herein. The Publisher does not necessarily endorse
the ideas held, or views expressed by the Editors or Authors of the material contained
in its publications.
© WIT Press 2010
Printed in Great Britain by MPG Book Group.
All rights reserved. No part of this publication may be reproduced, stored in a
retrieval system, or transmitted in any form or by any means, electronic, mechanical,
photocopying, recording, or otherwise, without the prior written permission of the
Publisher.
Editor: Eduardo Pilo
Universidad Pontificia Comillas de Madrid, Spain
Contents
Par t A. Ener gy Management in the Tr ain Oper ation
Reducing power peaks and energy consumption in rail transit systems
by simultaneous train running time control
T. Albrecht..................................................................................................... 3
Power management control in DCelectrified railways for the
regenerative braking systems of electric trains
Y. Okada, T. Koseki & K. Hisatom............................................................. 13
Impact of train model variables on simulated energy usage and
journey time
P. Lukaszewicz............................................................................................ 25
A study of the power capacity of regenerative inverters in a DC electric
railway system
C. H. Bae, M. S. Han, Y. K. Kim, S. Y. Kwon & H. J . Park....................... 35
Train operation minimizing energy consumption in DC electric
railway with onboard energy storage device
K. Matsuda, H. Ko & M. Miyatake.............................................................. 45
Computeraided design of ATO speed commands according to
energy consumption criteria
M. Dominguez, A Fernandez, A.P. Cucala & L.P. Cayuela........................ 55
Charge/discharge control of a train with onboard energy storage devices
for energy minimization and consideration of catenary free operation
M. Miyatake. K. Matsuda & H. Haga.......................................................... 65
Evaluation of energy saving strategies in heavily used rail networks by
implementing an integrated realtime rescheduling system
M. Luethi ..................................................................................................... 75
Par t B. Power Supply System Analysis, Design and Planning
Online temperature monitoring of overhead contact line at the new
German highspeed rail line CologneRhine/Main
N. Theune, T. Bosselmann, J . Kaise, M. Willsch, H. Hertsch
& R. Puschmann.......................................................................................... 87
Electric traction energy metering on German Railways and the impact
of European standardisation on the energy billing process in Germany
K. Weiland................................................................................................... 95
Development of feeder messenger catenary with the auxiliary wire
K. Nishi, Y. Sato & T. Shimada................................................................. 101
Catenary and autotransformer coupled optimization for 2x25kV systems
planning
E. Pilo, L. Rouco & A. Fernandez............................................................. 113
Investigation into the computational techniques of power system
modelling for a DC railway
A. Finlayson, C. J . Goodman & R. D. White............................................. 123
Optimal design of power supply systems using genetic algorithms
J .R. J imenez Octavio & E. Pilo.................................................................. 135
Application of linear analysis in railway power system stability studies
S. Danielsen, T. Toftevang & O.B. Fosso.................................................. 145
Fast estimation of aggregated results of many load flow solutions in
electric traction systems
L. Abrahamsson & L. Söder ...................................................................... 157
DC protection calculations – an innovative approach
R. Leach, D. Tregay & M. Berova............................................................. 171
Author index............................................................................................. 187
Pr eface
In recent years, energy consumption has become a crucial concern for every
transportation mode. However, it is in electrified railways where energy savings
have shown a bigger potential due to (i) regenerative braking, allowing the
conversion of kinetic energy into electric power, and (ii) vehicle interconnection,
which permits other trains to use regenerated power. In the future, increasing
energy efficiency and the emission reductions could lead railways to a significant
gain of modal share. Hence, an important effort has been done by the industry,
the operators, the research centers and governments to face this challenge. The
proceedings of the last editions of COMPRAIL conferences on railways clearly
reflect this sustained effort and main achievements of the past years.
This book gathers selected research papers published in the Computer in
Railways (COMPRAIL) series (IX, X and XI), which have been updated for this
edition. Although the book is focused on infrastructure, in many cases it is not
possible to analyze separately the train operation and the infrastructure’s
behaviour, particularly when the overall energy efficiency is taken into
consideration. The analysis of the impact of regenerative braking is a good
example of that, as it depends on all theses aspects: the onboard electronic
system and its control, the way the train is driven, the other trains in the area
(scheduling), the electrical characteristics of the traction network, the presence of
reversible substations (substations with inverters) and energy storage devices, etc.
Accordingly, a number of papers describing important issues related to energy
management and train operation have also been included.
This book is organized in two parts. The first focuses on energy management
issues in train operation and spans topics such as train driving, scheduling,
regenerative braking and onboard energy storage; the second deals with
infrastructure including topics such as catenary design and monitoring, traction
power systems analysis, computational issues in simulations and optimization.
Readers will find in this volume important papers dealing with a variety of
topics of current interest.
Finally, I would like to thank the authors for their revision of the papers as
well as the team of WIT Press that has worked in the edition of this book.
The Editor
This page intentionally left blank
Part A
Energy Management in the
Train Operation
This page intentionally left blank
Reducing power peaks and energy consumption
in rail transit systems by simultaneous train
running time control
T. Albrecht
“Friedrich List” Faculty of Transportation Sciences,
Chair of Trafﬁc Control and Process Automation,
Dresden University of Technology, Germany
Abstract
Costs for traction energy in electric rail transit systems do not only depend on
the energy actually consumed by the single trains. Other major factors affecting
the energy bill are power peaks, which stand for investment and sometimes for
operating costs and the efﬁcient use of energy regenerated during braking, which
can contribute to reducing peaks and energy consumption. For constant headway
operation on a single line, the headway itself and the interval between the depar
ture times of two trains from the two different terminus stations (synchronization
time) strongly inﬂuence energy consumption and power peaks. But these factors
are mostly not ﬁxed in favour of reducing energy costs but determined by trafﬁc
demand and operational restrictions.
This paper examines the possibilities of train running time modiﬁcation in order
to reduce power peaks and energy consumption for any situation of given headway
and synchronization time. The problem can be described as the search for an opti
mal distribution of a train’s running time reserve along its ride. The application of
Genetic Algorithms is proposed.
A case study is carried out for a German DC electric rapid rail system, where
different cost functions are examined. Simulation studies are performed taking
into account stochastically varying station dwell times. It is shown that using train
running time modiﬁcation, improvements in overall energy consumption can be
achieved and power peaks can be reduced signiﬁcantly.
Keywords: energy saving train control, coordinated train control, regenerative
braking, genetic algorithm.
Energy Management in the Train Operation 3
1 Introduction
Minimizing energy consumption in electric railways systems is not only a question
of minimizing the train’s energy needs for tractioning but also of efﬁciently using
regenerative energy. This topic is of special importance in DC systems with non
inverting substations. Here, energy billing is mostly realized at substation level
and the efﬁcient use of regenerative energy can directly contribute to reducing the
amount of energy to be purchased. But energy costs are not only determined by
the energy itself, power peaks often also inﬂuence the energy bill. According to a
UITP survey of underground railway system operators [1], more than 80% of the
operators paid a capacity price for the ﬁxed cost of energy supply, which depends
on the effective value consumed during a ﬁxed time period, e.g. 15 min.
Since the availability of fast and precise network simulators for modelling the
effects of the power supply system including regenerative braking, some
approaches have been taken to more efﬁciently using regenerative energy by means
of coordinated train control. Most of them deal with train dwell time control as a
method for improving the usage of regenerative energy. Control methods applied
are fuzzy control [2], search techniques [3] and heuristics [4, 5, 6]. They all have
the goal of providing decision safety, if and how long a train about to be starting
shall wait at its station, so that no high power peaks occur during its acceleration
and a big part of the energy needed for accelerating the train can be taken from
trains braking at the same instant. This approach suffers from mainly two points:
1. As long as operating personal is responsible for the clearance of the train,
precise timekeeping in the order of seconds can not be guaranteed. Passen
gers arriving during the additional dwell time trying to board the train will
not be denied their wish in most cases for reasons of customer satisfaction,
but the optimal departure time passes by.
2. Train travel time reserve used as additional dwell time could also have been
used on earlier stages of the train’s ride along the line as running time reserve
for longer coasting phases. This effect is independent of the mode of opera
tion of the train (manual or automatic).
To overcome these two obstacles, this paper proposes an approach using train run
ning time control instead of train dwell time control for synchronizing acceleration
and braking phases. The differences between the two approaches are illustrated in
ﬁgure 1.
In the next section, the problem of distributing train running time reserve along
a line is examined and the solution for minimizing a single train’s energy con
sumption is brieﬂy presented. For the minimization of system energy consumption
in constant headway operation, the use of Genetic Algorithms (GA) is proposed
in section 3. Section 4 examines the potential of the proposed method by means
of a case study for a German DC rapid railway system. The results for multitrain
coordination obtained using Genetic Algorithms are compared to the timetable
with minimal energy consumption for the single train.
4 Power Supply, Energy Management and Catenary Problems
time t
time t
necessary
dwell time
at station
additional
dwell time
for optimal
synchronization
additional
running time
allows additional
energy saving
improved usage
of regenerative
energy by delayed
departure
power P
power P
a) dwell time control
b) running time control
power curve
of second train
E <E
1 2
Figure 1: Dwell time modiﬁcation (a) vs. running time modiﬁcation (b) for
improved usage of regenerative energy.
2 Train running time modiﬁcation using Dynamic
Programming
The problemof distributing train running time reserve along a line may be regarded
as multistage decision problem, because at each stop it has to be decided, how
much reserve to spend on the next section of the ride. For many cost functions,
including the single train’s energy consumption, this problem can be solved using
Dynamic Programming [7].
Travel time reserve already spent when reaching an intermediate stop presents
the current system state, the transition between two succeeding stations (stages
of the process) is realized by a train ride with a certain amount of running time
reserve. The optimal distribution of running time reserve is computed recursively
from the terminus station with all reserve used up to the ﬁrst station, so an optimal
decision is computed for every possible process state. This makes the algorithm
suitable for online control.
Energy Management in the Train Operation 5
3 Using Genetic Algorithms for train running time control in
constant headway operation
To ﬁnd an optimal combination of timetables for the two directions in constant
headway operation can not be regarded as multistage decision problem, as the
decisions have to be made simultaneously for many trains.
The application of Genetic Algorithms (GA) is proposed here for the solution of
this problem. This universal solving tool can be used for practically any problem
that can be coded into binary form.
For coding, each unit of running time reserve (e.g. 1 unit = 1 s) makes up one
gene. The information the gene contents is the section of the track on which this
particular unit of running time reserve is to be spent. This coding results in a
binominal distribution of the different timetables favouring timetables with equally
distributed running time reserve. This contributes to ﬁnding reasonable and not
extreme solutions.
The initial population is created randomly except for one individual, which
presents the timetable with minimal energy consumption for the single train.
The cost function to be minimized can be chosen freely. During simulation stud
ies the minimization of energy consumption and of 15minaverage power for all
or selected substations have been used.
The size of the search space N for the particular problem of distributing k units
of running time reserve among n sections of the line is equal to a combination with
repetitions
N =
n + k −1
k
. (1)
For a typical problem like the one presented in the next section the solution
can be found using only 25 inviduals in one population for 50 generations, this
is extremely fast taking into account the size of the search space N ≈ 10
14
. The
solution of one such problem takes about 60  90 mins using a MATLAB imple
mentation on a 2.4 GHz Standard PC.
4 Case study
A case study has been carried out for one line of the Berlin SBahn network. It
consists of a track of 18 kms length with 14 stations (30 s dwell time at every
station). Power supply is realized by 4 substations situated at kms 0, 8.6, 11.8 and
18 [8]. The different sections are electrically coupled. The vehicle used for the
simulations is a BR 481 EMU. Energyoptimal train control between two consec
utive stations is realized using the controller presented in [7]. The quality criteria
are computed using a network simulator based on the solution of the nodal voltage
equations, speciﬁcities of DC systems are taken into account as proposed in [9].
At ﬁrst, the inﬂuence of the parameters headway and synchronization time are
examined. Then, the results of train running time modiﬁcation using Genetic Algo
rithms are presented. The obtained distribution of train running time reserve is used
6 Power Supply, Energy Management and Catenary Problems
300 600 900 1200 1500
160
180
200
220
headway in s
energy consumption in kWh
300 600 900 1200 1500
0
25
50
75
100
headway in s
regenerative rate in percent
Figure 2: Energy consumption and regenerative rate for different headways.
as timetable to keep in simulations. The same simulation is carried out for a con
troller using Dynamic Programming and the minimization of the energy consumed
by a single train as a target function. The both control strategies are compared.
4.1 Variation of headway
To examine the inﬂuence of the chosen headway on the energy consumed in the
network, a constant headway operation in only one direction of a line was sup
posed. It can be measured, how good the trains travelling in one direction are coor
dinated for themselves. It was assumed, that all trains travel with the timetable
causing minimal energy consumption for the single train. As ﬁgure 2 shows, there
are headways, which allow almost perfect reception of regenerated energy by the
trains travelling in only one direction, e.g. at 300 s. Receptivity of the network
decreases with increasing headway, simply due to the fact of less trains operating.
The increase of overall energy consumption is connected with it. The frequencies
visible in the function plots depend on track geometry and vehicle properties.
4.2 Variation of synchronization time for a given headway
When operating at headways with inherent receptivity, the synchronization time
between the two directions does hardly inﬂuence energy consumption or receptiv
ity of the line. For all other headways, this factor is of major importance. Here, a
headway of 600 s was chosen, being typically operated on the Berlin network dur
ing peak hours. Although this headway is a local minimumof energy consumption,
the regenerative rate is far below ideal values.
In ﬁgure 3 the results obtained for energy consumption, 15minaverage power
and line receptivity are presented for a range of synchronization times for the given
headway.
Energy Management in the Train Operation 7
0 50 100 150 200
365
375
385
395
405
synchronization time in s
energy consumption in kWh
(sum of all substations)
Minimal energy cons.
for single train
Opt. criterion
15−min−av. power
Opt. criterion
energy consumption
0 50 100 150 200
3
3.5
4
synchronization time in s
15−min−average power in MW
(sum of all substations)
Opt. criterion
15−min−av. power
Opt. criterion
energy consumption
Minimal energy cons.
for single train
0 50 100 150 200
70
80
90
100
synchronization time in s
regenerative rate in percent
Opt. criterion
energy consumption
Opt. criterion
15−min−av. power
Minimal energy cons.
for single train
Figure 3: Energy consumption, 15minaverage power and regenerative rates for
different synchronization times and a headway of 600 s.
4.3 Variation of train running times for given headway and synchronization
time
Choosing synchronization time is not only a question of energy consumption, the
choice is also inﬂuenced by the number of trains and, e.g. connections to other
lines. For a range of possible synchronization times in a 600 s headway situation,
it was examined, what beneﬁts can be achieved using train running time control.
The application of Genetic Algorithms as proposed in section 3 was realized here
for two different cost functions. The results are plotted in ﬁgure 3.
It can be seen, that the values of energy consumption and 15minaverage power
are much smaller for the timetables optimized for system energy and power than
with the initial timetable. It must furthermore be recognized, that the values
8 Power Supply, Energy Management and Catenary Problems
obtained for the different cost functions do in general not differ too much, but still
signiﬁcantly. For an operator the optimal compromise can be found if its actual
cost function is used for optimization.
As an example for a situation with a remarkable potential of train running time
modiﬁcation, the situation for 180 s synchronization time will be examined closer.
In ﬁgure 4 the initial timetable (optimized for energy consumption of the single
train) is compared to a timetable optimized using GA with 15minaverage power
as cost function. The latter timetable leads to energy savings of 4% and a reduction
of the sum of 15minaverage power of all substations of 17%.
Part a) shows the different distributions of running time reserve along the sec
tions of the line for both solutions. Whereas in the initial solution running time
reserve is almost equally distributed among the sections, this is not the case for the
system optimized timetable. It can already be seen from the resulting train trajec
tories in part b) of the ﬁgure, that there is more overlap of starts and stops in the
optimized timetable compared to the synchronous movement of the trains in the
middle sections with the initial timetable.
In part c) the sum of the demanded power, the power regenerated from braking
and the regenerative power not used in the network but wasted in braking resis
tances are plotted over time. The differences in the plots of these powers, serving
for the calculation of regenerative rates, are clearly visible: In the timetable opti
mized for multiple train operation the power peaks are much smaller and fewer
energy is wasted in the braking resistances. Part d) shows the reduction of the
effective power measured in the single substations by plotting the timedependent
curves.
4.4 Simulation studies taking into account stochastically varying station
dwell times
All results shown before were computed under the assumption of constant dwell
times in the stations. Here it will be examined, if and how the optimal timetables
can be realized in practical operation with stochastically varying dwell times. For
every scenario to be described, 200 simulations were realized with varying dwell
times at all stations.
At ﬁrst, it is assumed that, given a certain timetable, the strict keeping of this
timetable is obligatory. The reserve to spend on the next section t
res
is calculated
with
t
res
= scheduled arrival time −shortest travel time −actual departure time.
(2)
When negative t
res
occur, timeoptimal driving is applied. This corresponds to
a very simple Pcontroller.
With an assumed variation of 10 s of station dwell time the calculated amount of
energy saving and power reduction can also be realized under practical conditions.
It can be seen that the absolute value of energy consumption is 6% higher than the
theoretical value (see ﬁg. 5a), which obviously results from the situations, where
Energy Management in the Train Operation 9
2 4 6 8 10 section no.
0
20
40
60
sec
2 4 6 8 10 section no.
0
20
40
60
sec
0 500 1000 1500 s
20
40
60
km/h
0 500 1000 1500 s
20
40
60
km/h
0 200 400 s
2
4
6
MW
demanded
power
used
regenerated
power
wasted
regenerated
pow.
0 200 400 s
2
4
6
MW
demanded power
wasted
regenerated
power
used
regenerated
power
0 200 400 600 800 s
1
1.5
2
MW
SS1
SS2
SS3
SS4
0 200 400 600 800 s
1
1.5
2
MW
SS4
SS1
SS3
SS2
a) Distribution of running time reserve along the sections of the line.
b) Vehicle speed over time in the two directions.
c) Demanded power and regenerated power used and wasted over time.
d) Mean effective power curves for the four substations (SS) over time interval.
Figure 4: Comparison between initial timetable on the left and timetable optimized
for 15minaverage power (headway 600 s, synchronization time 180 s).
10 Power Supply, Energy Management and Catenary Problems
10s 15s 20s 25s
390
400
410
420
simple controller
Dynamic
Programming
controller
10s 15s 20s 25s
3.2
3.3
3.4
3.5
3.6
single train optimization
multi−train coordination
10s 15s 20s 25s
65
70
75
80
85
90
in kWh
a) Energy consumption
in MW (sum of all subst.)
b) 15minav. power c) Regenerative rate
in percent
Maximal deviation of station dwell times (equal distribution)
Figure 5: Energy consumption, 15minaverage power and regenerative rates for
different variations of dwell time.
only few or none of the running time reserve is left and timeoptimal driving has
to be applied in order to keep the timetable.
As mentioned earlier, the results of the optimization with Dynamic Program
ming can easily be used for online control. Compared to the strict timekeeping
control, energy consumption is reduced drastically and almost reaches the value
of multitrain optimization. With increasing dwell time variation, the advantage of
this controller shows up clearly: Energy consumption as well as 15minaverage
power decrease with this controller whereas with the simple controller and the
multitrain optimized timetable the results rise fairly stronger. On the other hand,
the regenerative rate remains higher for all examined cases with the multitrain
optimized timetable.
As the GA optimized timetable fulﬁls its purpose by optimally coordinating
starts and stops in the order of seconds, exact timekeeping is the only possibility
to reach this under stochastically varying dwell times. Whereas for smaller varia
tions this can be reached by the simple controller, higher variations call for a more
sophisticated controller combining the philosophies of energy saving of the single
train and coordination of starts and stops. The development of such a controller is
part of future work.
5 Conclusions
The paper presents a new approach to train running time control in order to achieve
energy cost reductions.
Given an optimal combination of headway and synchronization time, it is suf
ﬁcient to apply a controller based on the minimization of a single train’s energy
using Dynamic Programming. When these conditions can not be met, the modiﬁ
Energy Management in the Train Operation 11
cation of train running times can contribute to signiﬁcantly reducing power peaks
and energy consumption and thereby reducing energy costs in rail transit systems.
Acknowledg ments
This paper contains parts of the author’s doctoral thesis to be submitted to Dresden
University of Technology. It was elaborated within the research project ”intermobil
Region Dresden”, which is funded by the German Federal Government, the Min
istry of Research and Eduction (BMBF) under the project no. 19 B9907 A8. The
author wishes to thank Prof. H. Strobel for his helpful advice during the research
and the elaboration of this paper. He is also very grateful to Prof. H. Biesenack and
Prof. A. Stephan for supporting the analysis of the railway power supply system.
References
[1] UITP, Reducing energy consumption in Underground systems  an important
contribution to protecting the environment. Proc. of the 52
nd
International
Congress, Stuttgart 1997.
[2] Chang, C.S., Phoa, Y.H., Wang, W. & Thia, B.S., Economy/ regularity fuzzy
logic control of DC railway systems using eventdriven approach. IEE Proc.
Electr. Power Appl., 143(1), pp. 917, 1996.
[3] Firpo, P., & Savio, S., Optimized train running curve for electrical energy sav
ing in autotransformer supplied AC railways. Proc. of the IEE Conference
Electric Railways in a United Europe, pp. 2327, 1995.
[4] Gordon, S.P. & Lehrer, D.G., Coordinated train control and energy manage
ment control strategies. Proc. of the 1998 ASME/ IEEE Joint Railroad Confer
ence, pp. 165176, 1998.
[5] Guo, H.J., Ohashi, H. & Ishinokura, O., DC electric train trafﬁc scheduling
method considering energy saving  Combination of train trafﬁc parameters for
larger regenerative power (In Japanese). Transactions IEE Japan, 199D(11),
pp. 13371344, 1999.
[6] Sans` o, B. & Girard, P., Instantaneous power peak reduction and train schedul
ing desynchronization in subway systems. Transportation Science, 31(4), pp.
312323, 1997.
[7] Albrecht, T. & Oettich, S., A new integrated approach to dynamic schedule
synchronization and energy saving train control. J. Allan, R.J. Hill, C.A. Breb
bia, G. Sciutto, S. Sone, J. Sakellaris (eds.), Computers in Railways VIII, WIT
Press, pp. 847856, 2002.
[8] Biella, W., Die rechnergesteuerte adaptive Fahrkennlinienvorgabe zur Ener
gieoptimierung bei DCNahverkehrsbahnen (Diss.) TU Berlin, 1988.
[9] Cai, Y., Irving, M.R. & Case, S.H., Iterative techniques for the solution of
complex DCrailtraction systems including regenerative braking. IEE Proc.
Gener. Transm. Distrib., 142(5), pp. 445452, 1995.
12 Power Supply, Energy Management and Catenary Problems
e
Power management control in DCelectrified
railways for the regenerative braking systems
of electric trains
Y. Okada
1
, T. Koseki
1
& K. Hisatomi
2
1
The University of Tokyo, Japan
2
ShinKeisei Electric Railway Co. Ltd., Japan
Abstract
Most electric trains in DCelectrified railways are presently equipped with a
regenerative braking system. On braking, the traction controller of a train can
convert kinetic energy into electrical energy during deceleration of the train only
when other powering trains consume the electrical energy as electrical loads for
the regenerating train in the electrical circuit. Therefore, the traction controller of
the braking train must reduce the electrical power following squeezing control of
regenerative power when the electrical loads are too small in the electrical
circuit, because there are, typically, no other devices to absorb the regenerated
energy in the electrical circuit. However, actual traction controllers have often
reduced regenerative power excessively because they do not recognize the states
of the electrical circuit, which include positions of other trains and substations
and power consumption/regeneration of other trains in the electrical circuit.
In this paper, the authors discuss an improvement of the squeezing control of
regenerative power based on information of the electric circuit. The information
includes voltage at the pantograph, estimated positions and power
consumption/regeneration of other trains etc.
1 Regenerative braking in DCelectrified railway
Fig.1 shows the typical power flow on braking in a DCelectrified circuit. The
black solid arrows show the typical power flow in the present system, in which
only the powering train consumes the power regenerated from a braking train.
Therefore, the braking train must reduce the electrical power following
squeezing control of regenerative power when the power consumption of
powering trains is too small since there is, typically, no other device to absorb
Energy Management in the Train Operation 13
the regenerated energy in the electrical circuit. However, there are many possible
solutions for effective usage of regenerative braking. For example, brake
choppers with resistances on board or in the electrical circuit contribute to
maintenance reduction of trains. Another method is to install energy storage
devices which include flywheels, batteries and double layer capacitors on board
or in the electrical circuit, and commutated rectifiers at substations contribute to
efficient energy usage. In addition, reduction of voltage regulation at substations
and of feeding resistance can contribute to effective regenerative braking.
However these methods require additional hardware, which mean additional
cost. The other solution, which does not cause excessive cost, is to improve the
squeezing control of regenerative power, which can enhances the performance of
regenerative braking.
Power
converter
M otor
Pow er
converter
M otor
Pow ering train Regenerating train
Energy storage by fly w heels, batteries
and double layer capacitors
Pow er consum ption w ith brake chopper
and resistance
Substation
Pow er system Loads
Im provem ent of
squeezing control of
regenerative pow er
Reduction of
line resistance
Reduction of
voltage regulation
Introduction of
com m utated rectifier
Typical pow er flow in present system
System s for efficient energy usage and
m aintenam ce reduction
System s for only m aintenance reduction
Energy storage by fly w heels, batteries
and double layer capacitors
Pow er consum ption w ith D C chopper
and resistance
Figure 1: Typical power flow on braking.
In this paper, the authors discuss improvement of the squeezing control of
regenerative power with information of the electrical circuit and brake choppers
with resistances.
2 Problems of squeezing control of regenerative power
On braking, the braking train converts kinetic energy to electrical energy. And
other powering trains consume the electrical energy as electrical loads in the
electrical circuit. Therefore, when electrical loads are too small in the circuit, the
braking trains must reduce regenerative power following the characteristic
shown by the solid line in Fig.2 to avoid excessive voltage at the pantograph.
This control is called squeezing control of regenerative power.
Computers in Railways IX, J. Allan, C. A. Brebbia, R. J. Hill, G. Sciutto & S. Sone (Editors)
© 2004 WIT Press, www.witpress.com, ISBN 1853127159
14 Power Supply, Energy Management and Catenary Problems
1650 1750
Voltage of pantograph[V]
M
o
t
o
r
c
u
r
r
e
n
t
1900
C onventional characteristic
M axim al line voltage
0
Figure 2: Typical characteristic of squeezing control.
However, actual traction controllers often reduce regenerative power
excessively
[1]. The reasons for the excessive reduction are as follows;
1. traction controllers reduce regenerative power excessively in lowspeed
range because they reduce AC motor current directly instead of their DC
current,
2. traction controllers reduce motor current at lower voltage than maximal
voltage limit of feeding circuit as shown by the solid line in Fig.2 and,
3. actual traction controllers often reduce motor current at lower voltage than
the conservative voltage limit shown by the solid line in Fig.2.
In these problems, squeezing DC current of traction controller instead of AC
motor current can solve the problem in 1(above). However, a traction controller
needs to recognize the state of the electrical circuit in which braking trains exist
to solve the problems in 2 and 3. When the traction controller cannot recognize
the states of the electrical circuit, it must control regenerative power with
statically conservative characteristic shown by the solid line in Fig.2 to avoid
excessive voltage at the pantograph, because the voltage at the pantograph rises
when a powering train, which exists in the electrical circuit, reduces its power
consumption. The faster the reduction of power consumption is, the higher the
voltage of the pantograph rises. Therefore, the traction controller must squeeze
regenerative power regarding reduction of power consumption of register
controlled trains, which reduce their power consumption faster than any other
train, in the electrical circuit. However, the reduction of power consumption of
trains controlled by VVVFinverters, armature choppers or a field chopper is
slower than that of resistorcontrolled trains. Traction controller squeezes,
consequently, regenerative power excessively when powering trains controlled
by these methods to reduce their power consumption.
3 Improvement of squeezing control
The improvement of electrical circuits, power management with data
communication in an electrical circuit etc. are proposed to improve squeezing
control of regenerative power
[1], [2], [3]. In this paper, the authors propose
squeezing control of regenerative power whose characteristics vary according to
states of the electrical circuit. It is necessary to know the behaviour of the
Energy Management in the Train Operation 15
pantograph voltage rising quickly at the stop of the power consumption of
powering trains in the same electric circuit for improving the squeezing control
of the braking train. For that purpose, the traction controller must have the
following information;
1. the position and velocity of the trains, voltage at pantograph, DC current of
traction controller and power regeneration of the regenerating train,
2. running profile of the line on which the regenerating train exists,
3. control method of every train on the line,
4. the time when powering trains in the electrical circuit reduce their power
consumption and
5. distance between the braking train and the powering trains.
In the above, the information in 1 can easily be measured, and the information
in 2 and 3 can be stored on board as data of traction controller. However, the
information given in 4 and 5 needs to be estimated from the information in 1, 2
and 3. And, the characteristics of squeezing control of regenerative power must
be determined, based on the information.
One must propose how to estimate the information in 4 and 5 and how to
determine the characteristics of squeezing control of regenerative power. The
voltage regulation at the pantograph in the case of powering trains with various
control methods, reduce their power consumption for determining characteristics
of squeezing control of regenerative power in the following part of this paper.
Filter
capacitor
Filter
reactor
Internal
resistance
Feeder line
(variable)
Substation
Feeder line
(1km )
Braking train
Ｅ
fc
Ｉ
0
Ｅ
ｓ
Powering train
Squeezing control of
regenerative pow er
Traction
controller
Ｉl
Ｉ
ｓ
Ｉ
r
Ｅ
fc
Ｉ
0
: C urrent operation from braking operation
Ｉ
ch
Ｅ
fc
Brake chopper operation
Braking
resistor
Ｉ
ch
Ｉ
r
Figure 3: Electrical circuit for examination.
4 Voltage regulation at the pantograph
4.1 Electrical circuit for examination of voltage regulation
Fig.3 shows the electrical circuit to calculate voltage regulation at the
pantograph. The electrical circuit consists of a substation, a powering train and a
braking train controlled by VVVFinverter. The powering train is controlled by
16 Power Supply, Energy Management and Catenary Problems
VVVFinverters, fieldcurrent choppers or resistor controllers. Fig.4 shows the
equivalent circuits of the powering train and Fig.5 shows characteristics at
reduction of power consumption at the powering train. The line voltage at the
electrical circuit is limited up to 1900V. The authors will monitor the voltage at a
filter capacitor of a braking train instead of that at pantograph.
Filter reactor
Filter capacitor
Traction
controller
Traction
controller
(b) Field chopper control
Resistor control
(a) VVVFinverter control
Ｉ
a
Ｉ
l
Ｉ
l
Ｉ
a
Figure 4: Equivalent circuits of a powering train.
Ia[A]
1600 1600
Tim e[s] Tim e[s]
50
0 2.5 3.5 5
800
50
0 2.5 2.55 2.60 2.65
（ａ）VVVFInverter control （c）Resistor control
5
1600
Tim e[s]
50
0 2.5 3.1 5
（b）Field chopper control
Ia[A] Ia[A]
1.0s 0.6s 50m s
Figure 5: Characteristics at reduction of power consumption.
E
ｆｃ
C haracteristic of
squeezing control
I
0
I
00
>I
0
→ I =I
00
I
00
<I
0
→ I =I
0
I
00
I
10V
E
max
E
ｆｃ
[V]
I
00
[A]
2000
0
1
1+Ts
(T=30[m s])
I
r
(a) O peration logic for squeezing control of regenerative power (b) C haracteristic of squeezing control
Figure 6: Squeezing control for VVVFinverter and chopper controlled train.
4.2 Voltage regulation at pantograph of the braking train
4.2.1 Case of VVVFinverter controlled powering train
Fig.6 (a) shows the operation logic of how to reduce the regenerative power
when the powering train controlled by VVVFinverter stops its power
consumption. In this logic, the VI characteristic in Fig.6 (b) is assumed as the
“characteristic of squeezing control” in Fig.6 (a). The first order delay, the time
constant of which is assumed T=30[ms], represents the response of the traction
Energy Management in the Train Operation 17
motor current. In addition, the distance between the powering and the braking
trains is 2 km.
Fig.7 shows voltage at the filter capacitor of the braking train. It also shows
that the braking train can keep electric braking action by reducing its
regenerative power continuously for avoiding excessive pantograph voltage,
even if the other train stops its powering in various cases from E
max
=1600[V] up
to 1850[V]. In addition, Fig.8 demonstrates the relation between the voltage at
the filter capacitor and the DC current from the braking train while the powering
train reduces its power consumption in the case that E
max
is 1850V. And Fig.8
illustrates that the traction controller of the braking train can reduce its
regenerative power following the design of its squeezing control.
Figure 7: Voltage at the filter capacitor (VVVFinverter).
Figure 8: Following characteristic of squeezing control (VVVFinverter).
4.2.2 Case of powering train controlled by fieldcurrent chopper
Fig.6 (a) shows operation logic for squeezing control of regenerative power
when a powering train controlled by a fieldcurrent chopper stops its power
consumption. In addition, the distance between the powering and the braking
trains is 2 km.
18 Power Supply, Energy Management and Catenary Problems
Fig.9 shows voltage at the filter capacitor of the braking train. It also shows
that the braking train can keep electric braking action by reducing its
regenerative power continuously for avoiding excessive pantograph voltage,
even if the other train stops its powering in various cases from E
max
=1600[V] up
to 1850[V]. In addition, Fig.10 illustrates the relation between voltage at the
filter capacitor and the DC current of the traction controller of the braking train
while the powering train reduces its power consumption in case that E
max
is
1850V. And Fig.10 illustrates that traction controller of the braking train can
reduce its regenerative power following the design of its squeezing control.
Figure 9: Voltage at the filter capacitor (Chopper).
Figure 10: Following characteristic of squeezing control (Chopper).
4.2.3 Case of resistercontrolled powering train
Fig.11 shows operation logic for squeezing control of regenerative power in case
the powering train, which is resistercontrolled, reduces its power consumption.
In addition, the first order delay, whose time constant is 1.0 ms, is used to
suppress vibration of I
00
and the other first order delay, whose time constant is 30
ms, indicates characteristic of response of current at traction motor. Moreover,
Energy Management in the Train Operation 19
the limiter 1 makes its output zero when its input is negative and the Limiter 2
makes its output zero when its input is positive.
E
ｆｃ
I
0
I
00
I
d
dt
Proportional gain
0.3
Lim iter 1
Lim iter 2
+
+
C haracteristic of
squeezing control
I
00
>I
0
→ I =I
00
I
00
<I
0
→ I =I
0
I
r
1
1+T
1
s
(T
1
=1[m s])
1
1+T
2
s
(T
2
=30[m s])
I
00
Figure 11: Operation logic for squeezing control (2).
Fig.12 shows voltage at the filter capacitor of braking train when the E
max
indicated in Fig.6 (b) is 1600V and the distance between the powering and the
braking trains is 2km. Fig.12 also illustrates that the filter capacitor of braking
train rises drastically because the powering train spontaneously reduces its power
consumption in several milliseconds. Therefore, E
max
must be less than 1600V so
that traction controller can reduce regenerative power conservatively when
resistercontrolled train, instead of a VVVFinverter controlled train or a field
chopper controlled one, cuts off its power.
Figure 12: Voltage at the filter capacitor (3).
In addition, Fig.13 shows maximal voltage at the filter capacitor of the
braking train when the distance between the powering and the braking trains
varies if E
max
is 1600V. This figure means that the longer the between the
powering and the braking trains is, the lower the maximal voltage at the filter
capacitor of the braking train is, since the line resistance proportional to the
distance between the two trains restricts the power to be transferred from the
braking to the powering train.
Fig.13 also demonstrates that the longer the distance between the powering
and the braking trains is, the higher the E
max
can be. Fig.14 shows maximal E
max
to avoid excessive voltage at the filter capacitor of the braking train. This figure
means the longer distance between the powering and the braking trains allows
20 Power Supply, Energy Management and Catenary Problems
higher E
max
, since the influence from the action of the powering train is
substantially smaller when the distance between the two trains is longer. The
logic indicated by Fig.14 (b) determines the possible E
max
to avoid excessive
voltage at the filter capacitor of the braking train.
Figure 13: Voltage rise.
START
E
max
=1600[V]
C ircuit sim ulation
M axim al E
fc
< 1900V
(during sim ulation)
E
max
= E
max
10
End
E
max
= E
max
+10
No
Yes
(a) The possible E
max
to avoid excessive voltage (b) The logic to determ ine the possible E
max
Figure 14: Possible E
max
to avoid excessive voltage.
1850 0 1860
I
ch
[A]
E
fc
[V]
150
Figure 15: VI characteristic for a choppercontrol of a braking resistor.
If the braking train has supplemental braking resistor, whose characteristic for
operation is assumed as Fig.15, E
max
=1850[V] is possible for all the investigated
Energy Management in the Train Operation 21
train distance, since the braking resistor can effectively absorb the power
deviation from the spontaneous action of the powering train. In this case,
maximal power consumption of the braking resistor at all the investigated train
distance is 220kW, which is approximately 7% of maximal power consumption
of typical electric train on powering.
5 Conclusion
In this paper, the authors have proposed squeezing control of regenerative power
whose characteristics vary according to states of electrical circuit. They have
examined the voltage at the filter capacitor of the braking train when the
different three kinds of powering trains stop their power consumption. They have
concluded:
1. when a powering train, which is controlled by VVVF inverter or field
chopper, stops its power consumption, braking train can successfully
reduce its regenerative power with squeezing control whose E
max
is close to
maximal voltage limitation,
2. the controller of the braking train must reduce its regenerative power
conservatively when a resistercontrolled powering train close to the
braking train stops its power consumption,
3. longer distance between the powering and the braking trains allows higher
E
max
, since the influence from the action of the powering train is
substantially smaller when the distance between the two trains is longer,
and
4. the braking resistor, whose power consumption approximately 7% of the
maximal power consumption of typical electric train on powering enables
E
max
to be 1850[V] for all the investigated train distance.
6 Future work
The authors have studied only the squeezing control of regenerative power on
board. However, they must also investigate how to estimate and use the
following information to introduce a better squeezing control of regenerative
power whose characteristics vary according to the states of electrical circuit;
1. the time when powering trains in electrical circuit stop their power
consumption, and
2. distance between the braking train and the powering train which cuts off its
power consumption.
Acknowledgements
and cooperation in the investigation in this paper.
Mr. Hideki Iida at ShinKeisei Electric Railway Co., Ltd. for their assistance
The authors are grateful to Prof. Satoru Sone at Kogakuin University and
22 Power Supply, Energy Management and Catenary Problems
References
[1] S. Sone, Reexamination of Feeding Characteristics and Squeezing
Control of Regenerative Trains, Joint Technical Meeting Transportation
and Electric Railway and Linear drives, TER0249/LD0264, 2002.
[2] Y. Okada, T. Koseki, Evaluation of maximal reduction of electric energy
consumed by DCfed electric trains, NATIONAL CONVENTION
RECORD I.E.E. JAPAN, 5219, pp307308, 2003.
[3] Y. Okada, T. Koseki, S. Sone, Energy Management for Regenerative Brakes
on a DC Feeding System, STECH’03, pp 376380, 2003.
Energy Management in the Train Operation 23
This page intentionally left blank
Impact of train model variables on simulated
energy usage and journey time
P. Lukaszewicz
Aeronautical and Vehicle Engineering, KTH, Stockholm, Sweden
Abstract
Several train model input variables, such as running resistance, line voltage,
adhesion, braking release time and braking gain time, are studied. An analysis is
performed on how variations in the variables impact relatively on calculated
energy usage and running time of trains. The study shows that for the calculation
of energy usage the simulations are most sensitive to variations in running
resistance, followed by line voltage, adhesion, braking release time and braking
gain time. For the running time, the study shows that variation in mechanical
rolling resistance and air drag has a relatively small influence provided that the
tractive force is big enough. If the line voltage and adhesion, which affect here
the tractive force, drop below certain levels the running time increases
dramatically. The braking release and gain times have little influence on the
running time. The results also show which variables should be paid extra
attention to, when constructing a train model.
Keywords: train modelling, train data, sensitivity, power consumption, energy
usage, running time, simulations, ERTS.
1 Introduction
The correctness of computed results of energy usage and running time of trains
in a railway network is dependent upon the chosen train model and input data.
Therefore it is of interest to examine quantitatively how much the results can
differ from each other if the input data used by the same train model varies and
which data should be paid extra attention to.
By means of sensitivity analysis, the impact of the following variables is
studied for a SJ Rc4 loco hauled freight train:
Energy Management in the Train Operation 25
 Running resistance, which is the total force acting against the travel
direction.
 Adhesion.
 Tractive force (due to variation in catenary voltage).
 Braking gain time, which is the time it takes to obtain the desired
braking force, from when the driver starts braking.
 Braking release time, which is the time it takes to reduce the braking
force to zero, from when the driver stops braking.
Section 2 describes the method and models. The results are presented in section
3 and are discussed in section 4.
2 Method and models
This sensitivity analysis on how variation in input data affects the final results on
computed energy usage and running time is here performed by means of the
Energy and Running Time Simulator, ERTS. ERTS is a simulation program
developed by KTH and has verified models and data, versus fullscale
measurements, of trains and drivers. The verification shows that the discrepancy
between calculated and measured train energy usage is within the measurement
error of approx. 2% [1].
The train models are detailed especially with respect to braking and tractive
forces, electrical efficiency, running resistance, adhesion and slippage.
The driver models in ERTS are developed from fullscale measurements [2].
Observations were made on how real drivers are handling the trains especially
with respect to track profile, signalling and type of train and service. The
developed driver models, not included here, can drive a train as an average driver
would drive, or drive in an optimised way with respect to energy usage or
running time.
The driver model in this study is constant and set to drive the train strictly in
accordance with the signalled speed. The acceleration is performed at maximal
powering. Braking is performed as late as possible with respect to the braking
ability which is set to 1/3 of the maximal braking force of the train. This level of
the braking ability is obtained from observations on how the trains are driven in
reality. The models are described in [1].
2.1 The train model
The train model represents a loco hauled freight train of mixed consist.
The locomotive is of type SJ Rc4 and the tractive force diagram for two
different catenary voltages and powering levels is shown in Figure 1, together
with the tractive force limit,
α
F , due to adhesion as it is modelled in ERTS.
The calculated magnitude of the tractive force, F
w
, takes into account the
powering level, effect of speed, catenary voltage and the tractive force,
α
F ,
available with respect to adhesion. In this study, no wheel slippage is present.
26 Power Supply, Energy Management and Catenary Problems
0 5 10 15 20 25 30 35 40
0
50
100
150
200
250
F
t
(
k
N
)
v (m/s)
Limit due to adhesion
(ERTS)
Notch 9, 15 kV
Notch 3, 15 kV
Notch 9, 12 kV
Notch 3, 12 kV
F
a
Figure 1: Tractive force diagram. Notch 9 is the maximal powering level.
This means that the train speed is the same as the tangential speed at the
peripheral of the wheels of the locomotive. The tractive force at the wheels, is
calculated by:
) , min(
α
F F F
t w
= (1)
The total energy usage, of the train is calculated at the pantograph level for two
cases; E
1
, when a tractive force is present and the train is moving, and E
2
when
the train is coasting, braking or not moving.
( )
2 1
1
) ( 0 2
1
6
) (
1
0 or 0 ;
0 , 0 ;
10 6 . 3
1
) , (
) 1 (
E E
F v t P E
F v
v p
t v a K F
E
E
n
i
w i i
n
i
w
i
i i i j i w
tot
+ =
= = ∆ =
> >
⋅
∆ + +
=
=
∑
∑
=
=
η
ζ
(2)
where, E
tot
is total energy usage in kWh, n is the total number of time steps t ∆
during a simulation. K is a constant accounting for the rotational masses, a is the
acceleration, ζ is the slippage (=0), η is the efficiency of the locomotive as a
Energy Management in the Train Operation 27
function of power, p, and speed v, and P
0
is originating from the auxilliary
power. The total running time is calculated from
∑
=
∆ =
i
n
i
t T
1
(s), for v>0 (3)
The freight wagons in the train set have 2 axles/wagon and are of two types;
open type Oms and covered type Hbis. Basic data for the test train is shown in
Table 1:
Table 1: Nominal and basic data for the test train.
Length, incl. loco 418.5 m
Mass, gross incl. loco 1197 t
Mass of locomotive SJ Rc4 79 t
Axles, trailing 52
Max speed 100 km/h, 27.8 m/s
Axle load, average 21.5 t
Braking gain time, nominal 15 s
Braking release time, nominal 30 s
Braking level used 1/3 of max
The reason for choosing this train configuration is because of the existence of
measured data [1] on energy usage, running resistance, tractive force, efficiency,
braking ability and time lags in the tractive and braking systems.
2.2 Track model
The track model represents a tangent CWR. The length of the track is 88 km. A
simulation with nominal input data for the train model results in a running time
of 3597 s. The signalled speed restrictions are according to Table 2:
Table 2: Speed restrictions for the track model.
Distance (m) speed (km/h)
0 100
20490 40
21364 100
38152 70
39288 100
44106 70
44566 100
51322 40
52534 100
88000 100
28 Power Supply, Energy Management and Catenary Problems
3 Impact of variables on energy usage and running time
3.1 Simulation with nominal input data
Figure 2 shows the speed profile for the train obtained from simulation with
nominal input data. Table 3 shows the numerical results. This is the reference
case, with which all other results are compared with in this study.
Figure 2: Speed profile from simulation with nominal input data.
Table 3: Results from simulation with nominal input data.
Constant grade (‰) E
tot
(kWh) T (s) Mean speed (m/s)
0 1723.5 3597 24.47
5 3329.9 3758 23.42
3.2 Running resistance
The nominal running resistance,
0 R
F , of the train set is obtained from fullscale
measurements [1] and is calculated as a function of speed, v, by:
2
0
4 . 41 1 . 229 11961 v v F
R
+ + = (4)
The impact of variation of running resistance on energy usage and running time
is shown in Figure 3.
Energy Management in the Train Operation 29
Figure 3: Impact of variation of running resistance on energy usage and
running time.
In this case, the impact on running time is small, but big on the energy usage.
If the resistance has large errors from input data together with resistance
originating from grades, the tractive force of the locomotive might not be
sufficient. In this case severe delays will be present.
3.3 Adhesion
The available nominal adhesion is calculated in ERTS by the CurtiusKniffler
formula [3] which has been modified [1] to better suit fullscale test data.
0
7.5
0.9( 0.161)
44 3.6v
α = +
+
(5)
The results are shown in Figure 4. If the adhesion is higher than nominal, almost
no variation occurs. However, if the adhesion ratio for this case starts decreasing
below approx 0.7, the running time starts increasing due to insufficient tractive
power limited by the adhesion. Energy usage decreases mainly because of lower
average speed which reduces the aerodynamic drag.
3.4 Line voltage
The tractive force of the locomotive SJ Rc4 is affected by the line voltage, see
Figure 1. A voltage drop decreases the tractive force from the train speed of
17 m/s and up.
The variation of running time and energy usage due to variation of line
voltage is shown in Figure 5. The nominal voltage is 15 kV.
3.5 Braking gain time
The variation of braking gain time has for this studied case very small impact on
the running time and energy usage, as shown in Figure 6.
30 Power Supply, Energy Management and Catenary Problems
Figure 4: Impact of adhesion on energy usage and running time for grade 0 and
5‰.
Figure 5: Variation of energy usage and running time due to variation of line
voltage.
Energy Management in the Train Operation 31
Figure 6: Variation of running time and energy usage due to variation of
braking gain time.
3.6 Braking release time
The variation of braking release time has a slight impact on energy usage. If the
braking release time is reduced, compared with the nominal 30 s, a decrease in
energy usage is distinguished, Figure 7.
Figure 7: Variation of energy usage and running time due to variation of
braking release time.
4 Conclusions
This study shows in a quantitative way the importance of choosing correct input
data and their significance. It is therefore important to have up to date models, to
collect train data, maintain databases and to have information on how and for
which circumstances the data should be used.
Variation of running resistance has little effect on running time, provided
the tractive force is sufficient. The energy usage is strongly dependent upon the
running resistance.
When the available adhesion, as modelled in ERTS, drops under a certain
level the energy usage drops as well. The running time increases significantly.
32 Power Supply, Energy Management and Catenary Problems
When the line voltage drops and the tractive force is not sufficient, the
energy usage drops as well. The running time increases significantly.
Variation of the braking gain and release times showed little significance in
this study.
In this study, only the train model data is studied. An another important
factor is the driver behaviour which has a strong impact on energy usage .
References
[1] Lukaszewicz P., Energy Consumption and Running Time for train. KTH
Stockholm 2001. TRITAFKT 2001:25. ISSN1103470X.
[2] Lukaszewicz P., Driving describing parameters, energy consumption and
running time. Computers in Railways VIII. Comprail 2002 Lemnos.
[3] Andersson, E. Berg, M.., Railway systems and vehicles (in Swedish). KTH
Energy Management in the Train Operation 33
This page intentionally left blank
A study of the power capacity of regenerative
inverters in a DC electric railway system
C. H. Bae, M. S. Han, Y. K. Kim, S. Y. Kwon & H. J . Park
Korea Railroad Research Institute, South Korea
Abstract
This paper presents a method of determining power capacity and installation
positions of regenerative inverters installed in DC electric railway system. This
method uses the regenerative power data obtained from Train Performance
Simulation (TPS) and Power Flow Simulation (PFS). The simulation results of
TPS and PFS for Seoul subway lines 5 and 6 were applied, and suitable
substations where regenerative inverters should be installed and the suitable
power capacity to be installed were decided.
Keywords: regenerative inverter, electric railway system, train performance
simulation, power flow simulation.
1 Introduction
In a DC electric railway system, 22.9kV system voltage is converted into DC
1500V voltage through a 3phase silicon diode rectifier and supplied to traction
energy with railway motor cars. Since the regenerative power generated at the
regenerative braking of motor cars cannot be absorbed into the supply grid in the
case of diode rectifiers, this power should be used at nearby powering trains or
consumed as heat at resistances mounted on the cars. However, if a regenerative
inverter is installed in inverseparallel with the diode rectifier, it can absorb this
dump regenerative energy and feed it into an electric highvoltage grid for reuse.
Accordingly, the energy can be saved by reusing dump regenerative power
wasted away as heat, and the braking and ATO performance of motor cars can be
improved through enhancing the regenerative power absorption rate of catenary
lines. Despite these advantages, regenerative inverters cannot be installed at all
substations for electric railways because the manufacturing and installation cost
of regenerative inverters is higher than the benefit from the reuse of regenerative
Energy Management in the Train Operation 35
powers. Thus, they should be installed at sections with a long continuous slope
or where regenerative power loss in the resistor bank becomes a problem. In
order to determine the appropriate installation positions, number and capacity of
the regenerative inverter, it is necessary to calculate the accurate regenerative
power generated in a subway line.
This paper suggests determination schemes of the capacity and installation
positions of regenerative inverters installed in 1500V DC electric railway
system. We suggested a method that approximates using parameters related to
substations where regenerative inverters are installed, railway lines and operating
motor cars, and another that calculates using regenerative power obtained from
Train Performance Simulation (TPS) and Power Flow Simulation (PFS)
developed by Korea Railroad Research Institute for light rail transit system [1].
We carried out TPS and PFS for Seoul subway lines 5 and 6 and calculated the
regenerative power and decided the substations where regenerative inverters
should be installed and the suitable power capacity to be installed.
2 Power capacity of the regenerative inverter
Fig. 1 shows a diode rectifier and a regenerative inverter at an electric railway
substation. The 12pulse diode rectifier generates 1500V DC voltage and the
IGBT regenerative inverter detects the voltage rise of the catenary line caused by
the dump regenerative energy, absorbs the regenerative power, and transmit it to
a highvoltage grid for reuse. Since many trains can brake simultaneously in a
subway line, the peak power rating of the regenerative inverter needs to be
higher than that of industrial inverters. Thus, the regenerative inverter allows the
output AC current to limit at a certain level in constant current control mode in
general. However, since this current cannot increase infinitely due to the
limitations of the overhead line voltage, it is inevitable that the intermittent peak
power rating of the regenerative inverters increases as much as possible. In order
Figure 1: DC electric railway substation equipped with a regenerative
inverter.
36 Power Supply, Energy Management and Catenary Problems
to estimate the correct power capacity of such a regenerative inverter installed at
substations for DC electric railways, it is desirable to block the regenerative
power loss in the breaking chopper and resistor of all trains on a subway line,
make a route for absorbing regenerative energy, and measure this surplus
regenerative energy. Although this method can measure the surplus regenerative
energy at a substation exactly, it requires regenerative power absorbing
equipment, such as a resistor bank, installed at a substation. However, the
additional installation of resistor banks at electric railway substations is not easy
due to insufficient underground capacity in general. There are other methods,
such as approximating based on variables related to the substation, operating
line, train condition and regenerative power in other lines and calculating using
TPS and PFS. However, because the level of regenerative power varies
according to the conditions of the line on which the regenerative inverter is
installed, the train condition and the operation condition, it is difficult to
determine the accurate capacity through approximation based on these major
variables. Accordingly, we need to calculate dump regenerative power in various
train operation conditions by conducting TPS and PFS under different conditions
of line, train and substation.
3 Approximation method
Fig. 2 shows the layout of a substation for a DC electric railway for calculating
the power capacity of a regenerative inverter, and table 1 shows the calculation
conditions. A regenerative inverter in charge of a 12kmlong regeneration
section is installed at substation B, and the number of trains running in the
section, n , is obtained by eqn. (1).
h v
l
n
s
·
·
=
60
[trains/hour] (1)
where b means an integer larger than b , distance (l ), headway ( h ) and
commercial speed (
s
v ) are represented as units of meters, minutes, and h km/ ,
Figure 2: DC 1500V electric railway power system.
Energy Management in the Train Operation 37
respectively. The total regenerative energy that takes place in a day in section l
can be approximated in the following equations. Maximum power consumption
per hour,
m
P , is calculated from the train tonkilo capacity as follows,
k a l w s n P
m
· + · · · · = ) 1 ( 2 [kW] (2)
Here, the coefficient 2 means a double track section, and a is the standard
deviation of power variation according to the train diagram. The power capacity
of the regenerative inverter can be estimated using a power regeneration rate and
a regenerative braking efficiency rate obtained from substations equipped with
regenerative inverters at different railway substations. The power regeneration
rate,
1
ì , means the ratio of absorbed regenerative power to the maximum power
consumption of substations with a regenerative inverter,
m
P . The regenerative
braking efficiency rate,
2
ì , means the ratio of absorbed regenerative power to
the total regenerative power generated within the section covered by a substation
with a regenerative inverter. Here, the total regenerative power includes the
regenerative power consumed by nearby accelerating trains and regenerative
power loss in the resistor bank. In general, power regeneration rate
1
ì ranges
from 0.23 to 0.20, and regenerative braking efficiency rate
2
ì from 0.67 to 0.63
[2]. Using these data, the capacity of a regenerative inverter can be calculated as
eqn. (3), where W denotes the total regenerative power generated from the
section covered by the regenerative inverter. W includes the regenerative power
consumed by nearby accelerating trains and power loss in the resistor bank.
Accordingly, the capacity of the regenerative inverter should be larger than
W considering the operation condition of the line.
2
1
ì
ì
× =
m
P W [kW] (3)
Braking force at deceleration rate,  , can be obtained as eqn. (4). The
braking electric power generated from the regenerative braking performance of a
train at speed of v [km/h] is calculated by eqn. (5).
w s ) r ( . F
b
× × ÷  × × = 31 8 9 [N] (4)
q ×
×
=
367
v F
P
b
b
[kW] (5)
The regenerative peak current,
b
I , can be calculated as follows.
inv
b
b
V
P
I = [kA] (6)
On the conditions of table 1, W is obtained as 1480[kW] and
b
I 3.5[kA].
Thus, the power capacity of the regenerative inverter can be approximated as
38 Power Supply, Energy Management and Catenary Problems
1.5MVA, 350% 1 minute. However, this approximated calculation method does
not consider the railroad and train operation conditions: grade, curvature, and
headway duration. It can be used only to review the total system capacity rather
than as a specification to install a regenerative inverter.
Table 1: Calculation conditions.
Item Value Item Value
Number of cars, s 8 (4M4T) Running resistance, r 10kg/ton
Headway,
h
t 2.5 min Maximum speed,
m
v 80km/h
Weight, w 48 ton/car Commercial speed,
s
v 35km/h
Decelerating rate, 
0.97 m/s
2
Regenerative operation
voltage,
inv
V
1650V
Train tonkilo capacity, k 50kW/1000ton․kmPower regeneration rate,
1
ì 0.20
Power delivery efficiency, q
0.85
Regenerative braking
efficiency rate,
2
ì
0.65
4 Power flow simulation method
This section explains how to determine the capacity of a regenerative inverter
using TPS and PFS. PFS is performed by changing the power capacity and the
installation number of regenerative inverters, and the regenerative power loss of
a railway line is calculated. The loss ratio of regenerative power means the ratio
of regenerative power consumed as heat on the train to the whole regenerative
power generated as shown in eqn. (7). After the optimal position and the number
of regenerative inverters are determined, as a way of reducing the calculated loss
ratio of regenerative power to the maximum, the root mean square of
regenerative power (RMS power) and peak power are calculated. The effective
regenerative power per hour calculated by eqn. (8) determines the continuous
rating of the regenerative inverter, and is used to determine the peak power rating
based on the maximum regenerative power rate and the braking time of motor
cars.
100 1 ×


.

\

÷ =
reg
inv
P
P
R (7)
where P
reg
denotes the 1hour average value of the regenerative power generated
in a subway line and P
inv
denotes the 1hour average output power of
regenerative inverters in a subway line. In order to decide the continuous and
intermittent peak power capacity of the regenerative inverter, the mean square
value of the regenerative power generated in a substation is calculated as eqn.
(8).
Energy Management in the Train Operation 39
}
=
2
1
2
) (
1
t
t
reg
s
dt t p
T
P (8)
Here, P is the root mean square of regenerative power ) (t P
reg
, and T
s
sets 1
hour from t
1
to t
2
. The determination method of the suitable installation location
and power capacity of the regenerative inverters to be installed is shown in the
block diagram in fig. 3, and the details are as follows.
1. Perform PFS for the case that regenerative inverters are installed in all
substations on the line.
2. Calculate the mean square of regenerative power of each substation, and
rank the substations according to regenerative power.
3. Perform PFS after removing the regenerative inverters from the two
substations with the lowest regenerative power.
4. Again calculate the root mean square of regenerative power of each station
with a regenerative inverter, and calculate the loss ratio of regenerative power
for the whole line.
5. Perform PFS while removing the regenerative inverters one by one from the
substations with the lowest regenerative power.
6. Draw the curve of the loss ratio of regenerative power according to the
number of regenerative inverters installed in substations, and select the curve
that shows the largest reduction in regenerative power loss.
Train Performance Simulation
DC Power Simulation
Decrease Installation Number of
Regenerative Inverter
Calculate Loss Rate
of Regenerative Power
Calculate
Maximum and Root Mean Square
value of Regenerative Power
Decide installation
substation
Decide Power Rating of
Regenerative Inverter
Figure 3: Flowchart for substation selection.
40 Power Supply, Energy Management and Catenary Problems
Figure 4: Flowchart for regenerative inverter capacity.
10 12 14 16 18 20 22 24 26 28 30
1400
1500
1600
1700
1800
l
i
n
e
v
o
l
t
a
g
e
[
V
]
10 12 14 16 18 20 22 24 26 28 30
4000
2000
0
2000
4000
6000
8000
c
o
m
s
u
m
e
d
p
o
w
e
r
[
k
W
]
Time[min]
Figure 5: Seoul line 6 substation 8 without a regenerative inverter.
Once the position and number of regenerative inverters to be installed are
determined, the rated capacity of the regenerative inverter and the peak power
capacity are calculated through the procedure in fig. 4. The rated capacity of a
regenerative inverter sets the root mean square value of regenerative power
obtained from the substations, and the peak power rating is determined by the
ratio of the peak regenerative power to the root mean square value of
regenerative power. In addition, because the time for the rise of catenary line
voltage caused by the dump regenerative power of the subway substations does
not exceed 1 minute, the peak power rating is assumed to continue for 1 minute.
We performed TPS and PFS using data on trains and lines of Seoul subway
lines 5 and 6. Figs. 5 and 6 show the catenary line voltage and the power
consumption waveform of substations according to whether a regenerative
inverter is installed or not. In fig. 5, the regenerative power generated by the
power braking of motor cars is increasing the catenary line voltage
instantaneously. Fig. 6 shows that regenerative power is absorbed by the
substation and the variation of catenary line voltage is reduced.
Energy Management in the Train Operation 41
Fig. 7 shows absorbed regenerative power according to the number of
substations with a regenerative inverter. Fig. 7 (a) shows the case that
regenerative inverters are installed in all substations. Regenerative power is
different among substations because of the grade differences of line, distance
between stations and train operation conditions. Figs. 7(b)–(f) show the
regenerative power of each substation while removing the regenerative inverters
one by one from the substations with the lowest regenerative power. As the
number of substations with a regenerative inverter decreases, the regenerative
power at nearby substations with a regenerative inverter increases to some
degree.
10 12 14 16 18 20 22 24 26 28 30
1400
1500
1600
1700
1800
l
i
n
e
v
o
l
t
a
g
e
[
V
]
10 12 14 16 18 20 22 24 26 28 30
4000
2000
0
2000
4000
6000
8000
c
o
m
s
u
m
e
d
p
o
w
e
r
[
k
W
]
Time[min]
Figure 6: Seoul line 6 substation 12 with a regenerative inverter.
(a) (b) (c)
(d) (e) (f)
Figure 7: RMS of regenerative power in Seoul line 6.
42 Power Supply, Energy Management and Catenary Problems
Figure 8: Loss rate of regenerative
power in Seoul line 5.
Figure 9: Loss rate of regenerative
power in Seoul line 6.
Table 2: Power simulation results of Seoul subway lines.
Line Substation
RMS of regenerative
power[kW]
Peak regenerative
power[kW]
Ratio
[%]
5
Euljiro 4ga 1449 7102 490
Haengdang 1284 5664 441
Majang 1350 6554 485
6
Eungam 1305 6780 520
Daeheung 1279 6481 507
Samgakji 780 3827 491
Shinnae 941 4833 514
Figs. 8 and 9 show the curve of loss ratio of regenerative power changing
according to the number of regenerative inverters in Seoul lines 5 and 6. As a
largecapacity regenerative inverter makes it possible to transmit more
regenerative power to the supply grid, the loss ratio of regenerative power is
reduced, and the curve of regenerative power loss goes down with the increase in
the number of regenerative inverters installed. However, the reduction rate of
regenerative power loss is not constant. This is because regenerative power is
different among substations. As shown in figs. 8 and 9, reduction in the loss ratio
of regenerative power decreases gradually with the increase in the number of
substations with a regenerative inverter.
In the case of Seoul line 6, the reduction in the loss ratio of regenerative
power is largest when regenerative inverters are installed at four substations.
Because a larger reduction in regenerative power loss is not expected from the
installation of more regenerative inverters, it is desirable to install four
regenerative inverters. As in fig. 7, the adequate capacity of the regenerative
inverters for substations 1 and 5 can be selected as 1.5MVA and 1MVA for
substations 6 and 12, respectively. However, it is economically more efficient to
install a regenerative inverter only at substation 5 than at both, because
substations 5 and 6 are neighboring to each other.
We performed PFS for Seoul subway lines 5 and 6, and present the results in
table 2. The suitable power capacity of the regenerative inverter is determined by
Energy Management in the Train Operation 43
estimating the rated capacity as larger than the root mean square of regenerative
power from each substation and determining the peak power rating using the
ratio of peak regenerative power to the rated capacity.
5 Conclusions
This paper presents the methods for determining the installation location and
power capacity of regenerative inverters in DC electric railway systems. Using a
simple approximated calculation based on the conditions of the substations and
train operation and the regeneration rate of other railway lines, the power
capacity of the regenerative inverter was calculated. Also, the loss ratio of
regenerative power and the root mean square of regenerative power for each
substation were obtained using TPS and PFS and the installation location and
number of regenerative inverters was decided. Applying TPS and PFS to Seoul
subway lines 5 and 6, we obtained the suitable installation location and the
power capacity of the regenerative inverters to be installed.
References
[1] S.K. J ung et al., “Right Rail transit system development”, Korea Railroad
Research Institute, 2002.
[2] Electric Railway DC Power Supply System Investigation Committee,
“Phenomena of Power Supply System Including Regenerative Cars and
Future Directions”, Technological Report No. 296 of J apanese Institute of
Electrical Engineers, 1989
44 Power Supply, Energy Management and Catenary Problems
Train operation minimizing energy
consumption in DC electric railway with
onboard energy storage device
K. Matsuda, H. Ko & M. Miyatake
Sophia University, Japan
Abstract
The optimal train operation which minimizes sum of supplied energy from
substations is presented in this paper. In recent years, the energy storage devices
have enough energy and power density to use in trains as onboard energy storage.
The electric double layer capacitor (EDLC) is assumed as an energy storage
device in our study, because of its high power density. The onboard storage
can assist the acceleration/deceleration of the train and may decrease energy
consumption. Many works on the application of the energy storage devices to
trains were reported, however, they did not deal enough with the optimality
of the control of the devices. On the other hand, our previous works were to
optimize acceleration/deceleration commands of the train for minimizing energy
consumption without the energy storage device. Therefore, we intend to optimize
acceleration/deceleration commands together with current commands through
energy storage devices as our next research target. The proposed method can
determine the optimal acceleration/deceleration and current commands at every
sampling point. For this purpose, the optimal control problemof the train operation
is formulated mathematically. It is generally difﬁcult to solve the problem because
the problem is composed of a largescale nonlinear system. However, the
Sequential Quadratic Programming (SQP) can be applied to solve the problem.
Two results with and without onboard energy storage device are compared. These
optimized results indicate that the total energy consumption is reduced by at least
0.35% by using the EDLC. The relation between internal resistance and energy
consumption is also revealed.
Keywords: electric double layer capacitor (EDLC), optimal control, energy saving
operation, SQP method.
Energy Management in the Train Operation 45
1 Introduction
In recent years, the energy storage devices have enough energy and power
density to use in trains as onboard energy storage. The devices are for instance,
a secondary battery and an Electric Double Layer Capacitor (EDLC). Above
all, the EDLC has advantages such as maintenance free, long lifetime, rapid
charge/discharge with large current and high efﬁciency. Therefore, the EDLCis the
most suitable to equip trains as an auxiliary power supply. The onboard EDLC is
useful because of the following two reasons. Firstly, it decreases the loss of circuit
resistance by compensating voltage drop. Secondly, it enables us to utilize and
recycle regenerative power efﬁciently and prevent regenerative failure.
Many works on the application of the energy storage devices to trains were
reported. However, from an energysaving point of view, they did not deal
enough with the optimality of the control of the devices. On the other hand,
our previous works [1, 2] was to optimize notch commands which determine the
acceleration/deceleration force in the train without energy storage devices. We
optimize notch commands together with charge/discharge commands with making
use of the experience of our previous study. It is signiﬁcant to investigate the
optimal charge/discharge command minimizing energy consumption in order to
maximize the effect of installing the EDLC.
In this paper, we intend to formulate the optimal control problem of the train
operation to ﬁnd notch and charge/discharge commands which minimize amount
of consumed energy, propose howto solve it, discuss the optimized results and ﬁnd
knowledge of the optimal operation. The knowledge will be applied to the future
charge/discharge controllers for EDLCs.
2 Modeling of DC feeding Circuit
We modeled a DC feeding circuit when there is only one train between substations.
The model circuit appears in ﬁg. 1. In this ﬁgure, V
s
and R
0
are the supply voltage
and the internal resistance at a substation respectively. The values of R
1
and R
2
are
wire resistances. These resistance values are proportional to the distance between
the train and substation position. Positions of substations and stations are shown
in ﬁg. 2. The constants C and R
c
are the capacitance and internal resistance in the
capacitor respectively.
It is necessary to convert voltage by using a chopper because the
voltage difference is high between the pantograph and capacitor. The chopper
characteristic is too complicated to be examined in detail here. Therefore, we
solved the circuit equation on the assumption that the chopper efﬁciency is 95%.
In addition, the energy consumption in the train is regarded as constant in
short time because acceleration/deceleration commands do not change often. The
motorinverters of the train are modeled as a current load that helps solving circuit
equations simply.
46 Power Supply, Energy Management and Catenary Problems
C
R
1
Chopper
R
0
R
0
V
S
V
S
Substation1 Substation2
R
C
R
2
Train
V
T
Figure 1: Circuit model with one train between substations.
SS1 AS DS SS2
L
L
a
L
b
AS:arrival station
DP:departure station
SS:substation
Figure 2: Positions of stations and substations.
3 Formulation of optimal control problem
We formulated the optimal control problem in this section. Here, variables are
deﬁned as follows. Control inputs n and u determine the acceleration/deceleration
force and charge/discharge current through the capacitor, respectively. They
are deﬁned as table 1. State variables x, v and V
c
indicate the train position,
speed and capacitor voltage, respectively. Variable V
T
is the voltage at the
pantograph. In fact, it is a state variable if control inputs are determined and the
circuit equation can be solved. However, we deﬁned V
T
as the auxiliary variable
because it is difﬁcult to solve circuit equations analytically. Additionally, these all
variables depend on time t. The optimal control problem is described as follows,
mathematically.
Table 1: Deﬁnition of control inputs n and u.
n or u Operation mode Current through the capacitor
1 maximum deceleration maximum charge
negative deceleration charge
0 coast wait
positive acceleration discharge
1 maximum acceleration maximum discharge
Energy Management in the Train Operation 47
Minimizing the objective function
J =
T
0
V
s
I
s
(x, V
T
)dt (1)
Subject to the following equality and inequality constraints
˙ x = v (2)
˙ v = f(n, v, V
T
) −r(v) (3)
˙
V
c
= −I
c
(u)/C (4)
P
T
(n, v, V
T
) = P
s
(x, V
T
) + P
c
(u, V
c
) (5)
x (0) = 0 v(0) = 0 V
c
(0) = V
c first
(6)
x (T) = L v(T) = 0 V
c
(T) = V
c final
(7)
−1 ≤ n ≤ 1 (8)
−1 ≤ u ≤ 1 (9)
V
T min
≤ V
T
≤ V
T max
(10)
V
c min
≤ V
c
≤ V
c max
(11)
0 ≤ x ≤ L (12)
v ≥ 0 (13)
where
I
s
sum of load currents supplied by substations
I
c
current through the capacitor
f,r acceleration/deceleration force and running resistance per kg
P
T
electric power supplied to motorinverters of the train
P
s
, P
c
power from substations and the capacitor
V
T min
, V
T max
lower and upper limitations of the voltage at the pantograph
V
c min
, V
c max
lower and upper limitations of the capacitor voltage
V
c first
, V
c final
ﬁrst and ﬁnal values of the capacitor voltage
L, T distance and running time between the departure and arrival
station.
The objective function is sum of supplied energy by two substations given as
eqn. (1). Equality constraints are given as eqns. (27). Eqns. (2),(3) are motion
equations of the train. The capacitor voltage is given as the eqn. (4). As mentioned
above, we must solve the circuit eqn. (5) because we deﬁned V
T
as an auxiliary
variable. Eqns. (6),(7) describe the initial and ﬁnal conditions of state variables.
Inequality constraints of control inputs, state and auxiliary variables are shown in
eqns. (813). Especially, we did not consider speed limitations in eqn. (13).
48 Power Supply, Energy Management and Catenary Problems
We deﬁned functions as below.
P
T
(n, v, V
T
) =
Mvf(n, v, V
T
)m
e
(n ≥ 0)
Mvf(n, v, V
T
)/g
e
(n ≤ 0)
(14)
P
s
(x, V
T
) =
V
s
−V
T
R
0
+ R
1
(x)
+
V
s
− V
T
R
0
+ R
2
(x)
(15)
R
1
(x) = (L
a
+ x)r
0
R
2
(x) = (L −x + L
b
)r
0
(16)
P
c
(u, V
c
) =
V
c
I
c
(u)c
e
(u ≥ 0)
V
c
I
c
(u)/c
e
(u ≤ 0)
(17)
I
c
(u) = uI
c max
(18)
Here, m
e
and g
e
are motor/generator efﬁciency. Wire resistances R
1
and R
2
are
given in eqn. (16) when the position of the departure station is deﬁned as x = 0.
The constants L
a
and L
b
indicate the distance fromthe departure and arrival station
to the substation1 and substation2 shown in ﬁg. 1. The constant r
0
is the wire
resistance per meter. The constant c
e
is the chopper efﬁciency. The constant I
c max
is the rated value of the current from the capacitor.
Additionally, maximum acceleration/deceleration characteristics, such as the
control input n is 1 or 1, and running resistance are given in ﬁg. 3. these
characteristics are inﬂuenced by the voltage at the pantograph V
T
. Especially, we
assume that the braking system is the air supplement control. In short, the use of
electrical and mechanical blended braking system is considered if the regenerative
braking force is not enough for the speciﬁc braking force. Moreover, we are not
concerned with the characteristic of the squeezing control because we also assume
the regenerative power can be absorbed at substations.
The absorbed power can be accounted in eqn. (1).
0 5 10 15 20 25 30
0
0.2
0.4
0.6
0.8
1
1.2
Velocity[m/s]
A
c
c
e
l
e
r
a
t
i
o
n
F
o
r
c
e
[
N
/
k
g
]
Running Resistance
1300[V]
1500[V]
1700[V]
0 5 10 15 20 25 30
0
0.2
0.4
0.6
0.8
1
1.2
Velocity[m/s]
D
e
c
e
l
e
r
a
t
i
o
n
F
o
r
c
e
[
N
/
k
g
]
Maximum Braking Force
Electrical Braking Force
1300[V]
1500[V]
1700[V]
Figure 3: Acceleration/deceleration characteristics and running resistance.
Energy Management in the Train Operation 49
4 Optimization method
We show the optimization method for solving the optimal control problem. The
optimal control problem is solved by Sequential Quadratic Programming (SQP).
SQP is an optimization method to solve a general nonlinear programming problem.
A general optimal control problem with equality and inequality constraints can be
written as
Minimize : J (ω)
Subject to : g
i
(ω) = 0 (i = 1, · · · , n
g
) (19)
h
j
(ω) ≤ 0 (j = 1, · · · , n
h
)
where ω = (n, u, V
T
, x, v, V
c
)
is the vector of variables, J is the objective
function, g and h are equality and inequality constraints, n
g
and n
h
are the number
of equality and inequality constraints.
Next, the objective function is expressed using the second order approximation
around the feasible point ω
(k)
. Similarly, equality and inequality constraints are
also expressed as the ﬁrst order approximation around the feasible point ω
(k)
in
problem (19). The transformed problem is shown as
Minimize :
1
2
d
k
B
k
d
k
+∇J(ω
(k)
)d
k
Subject to : g
i
(ω
(k)
) +∇g
i
(ω
(k)
)d
k
= 0 (i = 1, · · · , n
g
) (20)
h
j
(ω
(k)
) +∇h
j
(ω
(k)
)d
k
≤ 0 (j = 1, · · · , n
h
)
where d
k
= ω−ω
(k)
, B
k
is positive deﬁnite matrix. In general, the problem (20)
can be solved by the interior point method [3, 4]. The optimization result of the
problem (20) is a search direction d
k
.
Here, we deﬁne merit function as
ψ(ω) = J(ω) + µ
n
g
i=1
g
i
(ω) +
n
h
i=1
max(h
j
(ω), 0)
(21)
where µ is a large positive constant. Finally, we have to ﬁnd α to minimize
ψ(ω
(k+1)
) where ω
(k+1)
= ω
(k)
+ αd
k
. The vector ω
(k+1)
is the next feasible
point. Consequently, The optimization problem (19) can be solved by iterating the
following procedure.
Step 1) give the initial feasible point ω
(0)
, and set k = 0
Step 2) solve problem (20), and obtain a search direction d
k
Step 3) stop iteration if the norm d
k
 is less than 10
−6
.
Step 4) ﬁnd α minimizing the merit function, and obtain the next feasible point
ω
(k+1)
Step 5) set k = k + 1, and return step2
50 Power Supply, Energy Management and Catenary Problems
5 Optimization result
Optimization results are presented in this section. Speciﬁc parameters are
presented as table 2. Here, the continuous time formulation must be transformed
to a discrete time program in order to apply SQP to the optimal control problem.
Therefore, ∆t is deﬁned as the sampling interval. In addition, the ﬁnal capacitor
voltage is given to equal the initial one. So, the capacitor does not have to be
charged when the train arrives at the station.
We show three optimization results when the train run on straight line without
speed limitations and gradients. Case 1 is the optimization result of the train
without the capacitor. Case 2 and Case 3 show results of the sensitive analysis in
case of the constant R
c
is 0.3[Ω] or 0.03[Ω] respectively. These three optimization
results are shown in table 3.
Case 1: The optimization result is shown in ﬁg. 4. The optimal train operation
consists of the maximum acceleration, reduced acceleration by degrees, coasting
and maximum deceleration. In previous works [1, 2], we obtained the similar
optimization result. Therefore, these results indicate the reliability of the proposed
method.
Case 2: ﬁg. 5(a) shows the optimization result in case that R
c
is 0.3[Ω]. The
optimal control input n do not differ from one in case 1. Next, we examine
charge/discharge characteristics of the capacitor. As supplied power to the train
is higher, discharge current is higher. Similarly, this pattern of the charge
characteristic is also represented when the train decelerates. Qualitatively, this
Table 2: Speciﬁc parameters.
Operating condition capacitor parameters
T, ∆t 130[s],1[s] C 32.3[F]
L 2000[m] V
c max
560[V]
R
0
0.03[Ω] V
c min
300[V]
V
s
1500[V] V
c first
560[V]
r
0
0.04 ×10
−3
[Ω/m] V
c final
560[V]
L
a
, L
b
5000[m] I
c max
500[A]
M 250 ×10
3
[kg] R
c
0.3[Ω] or 0.03[Ω]
Table 3: Optimization results.
Total energy consumption [MJ] Energysaving effect[%]
Case 1 27.55 
Case 2 27.45 0.35%
Case 3 26.74 2.92%
Energy Management in the Train Operation 51
characteristic is proper, because it is the most effective control to compensate the
voltage drop at the pantograph and prevent the regenerative failure. In Case 2,
despite the lower limit value of the capacitor voltage V
c min
set to 300[V], the
capacitor stops discharging when the value of the capacitor voltage drops to about
480[V]. This result is attributed to the higher interior resistance of the capacitor.
As a result, the efﬁciency of the capacitor itself is severely down according to the
voltage drop of the capacitor.
Case 3: The optimization result is shown as ﬁg. 5(b). In this case, there is
also little variation in control input n. However, the optimal charge/discharge
command includes two signiﬁcant difference compared to Case 2. Firstly, the
capacitor voltage reaches the value of the lower limitation when control input
n changes from acceleration to coast at about time 50[s]. Secondly, substations
supply the power with a small current for charging the capacitor when the train
coasts. It is found from the result that the capacitor is utilized effectively. Finally,
the total energy consumption is 26.74[MJ]. Compared with Case 1, the total energy
consumption is reduced about 2.92%. The result indicates that energysaving effect
is higher if the interior resistance of the capacitor is lower in the future.
0 20 40 60 80 100 120
0
10
20
v
[
m
/
s
]
0 20 40 60 80 100 120
1
0
1
n
0 20 40 60 80 100 120
1200
1400
1600
1800
V
T
[
V
]
0 20 40 60 80 100 120
4
2
0
2
4
P
T
[
M
W
]
time[s]
Figure 4: Optimization result (case 1).
52 Power Supply, Energy Management and Catenary Problems
0 50 100
0
10
20
v
[
m
/
s
]
0 50 100
0
10
20
0 50 100
1
0
1
n
0 50 100
1
0
1
0 50 100
1
0
1
u
0 50 100
1
0
1
0 50 100
1200
1400
1600
1800
V
T
[
V
]
0 50 100
1200
1400
1600
1800
0 50 100
300
400
500
600
V
c
[
V
]
0 50 100
300
400
500
600
0 50 100
4
2
0
2
4
time[s]
P
T
[
M
W
]
0 50 100
4
2
0
2
4
time[s]
(a)Case 2 (b)Case 3
Figure 5: Optimization results.
Energy Management in the Train Operation 53
6 Conclusion
This paper presents the optimal train operation with EDLC minimizing energy
consumption. SQP can be applied to the formulated optimal control problem with
discretetime transformation. As a result, it is found that the energy consumption
supplied from substations can be reduced by using onboard energy storage device
effectively. Compared with the train without EDLC, the total energy consumption
is reduced by 0.35% and 2.92% in Case 1 and Case 2 respectively. It is also
clariﬁed that the EDLC should not fully discharge when the internal resistance
is high.
A further direction of this study will be to optimize the train operation problem
which has more complicated running conditions, for example, speed limitations
and gradients.
References
[1] H. Ko, T. Koseki and M. Miyatake: “Numerical Study on Dynamic
Programming Applied to Optimization of Running Proﬁle of a Train”, IEEJ
Transactions on Industry Applications, Vol.125D, No.12, pp. 10841092,
2005 (in Japanese).
[2] H. Ko, T. Koseki and M. Miyatake, “Application of Dynamic Programming
to Optimization of Running Proﬁle of A Train”, Computers in Railways, WIT
Press, pp. 103112, 2004.
[3] Imad M. Nejdawi, Kevin A. Clements and Paul W. Davis, “an efﬁcient
interior point method for sequential quadratic programming based optimal
power ﬂow”, IEEE Trans. on Power Systems, vol.15, NO.4, November 2000,
pp. 11791183.
[4] G. Irisarri, L. M. Kimball, K. A. Clements, A. Bagchi and P. W. Davis,
“Economic Dispatch with Network and Ramping Constraints via Interior Point
Methods” IEEE Trans. on Power Systems, Vol.13, No.1, February 1998
54 Power Supply, Energy Management and Catenary Problems
Computeraided design of ATO
speed commands according to
energy consumption criteria
M. Domínguez
1
, A. Fernández
1
, A. P. Cucala
1
& L. P. Cayuela
2
1
Instituto de Investigación Tecnológica, Escuela Técnica Superior de
Ingeniería (ICAI), Universidad Pontificia Comillas, Madrid, Spain
2
Dirección de Ingeniería, Metro de Madrid, Spain
Abstract
Traffic regulation systems of metro lines equipped with Automatic Train
Operation (ATO) use a set of preprogrammed speed commands selecting
coasting points and brake deceleration. Different speed commands provides
different travel times between stations and the regulation system online selects
and sends to the train one of these commands. Nowadays, speed commands are
designed based on time and comfort criteria. In this paper a new approach of
speed commands design, which takes into account not only present operational
criteria but also energetic ones, is proposed in order to obtain energy efficient
ATO commands. Firstly, the travel time and energy consumption of every
command is calculated using a simulator that combines all the possible discrete
values of the ATO configuration parameters. A set of systematic rules has been
defined to include the consumption, operative and comfort criteria in the
selection of the speed commands applying decision theory techniques. A
software tool has been implemented for a computeraided design of the speed
commands. This tool includes a thorough simulation module of the train
movement (ATO, motor and train dynamics), an automatic generator of every
possible command and a graphical assistant for the speed commands selection
according to the mentioned rules. The methodology described in this paper has
been used to redesign the current ATO commands (4 for each station) of Line 1
of Madrid Underground. The results are presented in this paper. According to the
simulation results, about 10% of energy savings are expected to be achieved with
these new speed commands.
Keywords: energy consumption, speed commands design, train simulation.
Energy Management in the Train Operation 55
1 Introduction
Metro lines equipped with Automatic Train Operation systems (ATO) use pre
programmed speed commands to control the circulation of trains, providing a set
of alternative ATO speed profiles per interstation. As a result, driving is not
influenced by human factors and runtimes and energy consumptions are quite
stable when signalling systems do not affect the circulation of trains. Traffic
regulation systems performance and total energy consumption strongly depend
on the offline design of the ATO speed commands.
The ATO speed profile to be executed between two stations is online
selected by the regulation system according to the required runtime. When a
train must be held up, from the user’s point of view a longer runtime is
preferred rather than a longer station waiting time. In addition, this regulation
strategy involves energy saving because longer runtimes are obtained with
slower speed profiles. However, these ATO speed profiles are usually designed
according to runtime and comfort criteria, but not to energy consumption.
There are two main approaches to optimise train driving speed profiles:
mathematical optimisation models and computeraided design based on detailed
simulation and direct search techniques.
The first approach includes in a mathematical model those restrictions that
govern the train movement, such as track geometry, traction equipment, speed
restrictions and driving rules. For instance Howlett [1] and Khmelnitsky [2]
apply optimal control techniques for determining the optimal switching times in
manual driving. The formulation and resolution of these analytical models
require very complex techniques as well as important simplifications of the train
dynamics or the driving strategy.
On the other hand, approaches based on simulation do not require
simplifications and enable an accurate calculation of runtimes and energy
consumption, as Lukaszewicz manual driving modelling [3]. To explore the
solution space and select alternative driving, different direct search methods are
used, for instance heuristic search for ATO speed commands design [4] or
genetic algorithms and fuzzy logic for manual and automatic driving
optimisation [5], or genetics for optimisation of coasting points [6, 7]. Wong and
Ho [8] compare different search methods for the online control of a train using
an accurate simulator, determining the coasting points. This work stresses the
importance of an accurate train movement modelling for practical applications.
The work presented in this paper is focussed on the computeraided design of
the ATO speed commands between two metro stations, to be preprogrammed in
the ATO equipment. The variables to optimise will be the configuration data of
each particular ATO system, four parameters in the case study of Line 1 of
Madrid Underground: coasting, remotoring and regulation velocities, and
braking deceleration rate. The considered ATO system provides only certain
discrete values for each parameter, resulting a solution space of 220 alternative
speed profiles per interstation in Madrid Underground , and this fact allows the
exhaustive and accurate simulation of the whole feasible ATO speed profiles.
Instead of search techniques, decision theory techniques can be directly applied
56 Power Supply, Energy Management and Catenary Problems
to select a set of solutions per interstations (4 in the case study) including
operational and energy consumption criteria. This way, the obtained driving
solutions will be fully adjusted to real features and capabilities of the ATO
equipment in service.
A software tool for computeraided design of ATO speed profiles has been
developed, to support the design procedure defined. Next sections will describe
briefly the simulator, the proposed design procedure and the simulation results of
its application to Line 1 of Madrid Underground. The obtained solutions are
compared with the current driving profiles in service in this line in order to value
the expected energy savings.
2 The simulator
The proposed design method is based on the accurate simulation of energy
consumption and runtime of all the possible combinations of ATO speed
commands for each interstation. Train velocity, acceleration, traction or brake
force and energy consumption are computed at each simulation step. The
speed/distance profile between stations is plotted to assist the design process.
The simulator is composed of three modules: ATO equipment simulator, train
dynamics model and train consumption model. This modular architecture allows
the validation of each module separately and an easy adjustment for specific
features of a particular ATO equipment. To this end, the simulator input
interfaces are designed to enable the definition of track layout, train
characteristics, and ATO system configuration.
2.1 ATO equipment simulator
This module calculates the type of motion at each simulation step: motoring,
coasting or braking. The particular ATO system modelled in the case study (Line
1 of Madrid Underground) supplies a unique traction command (the maximum).
In contrast, there are four values for deceleration braking command which can be
selected. The regulation speed order is implemented by means of maximum
traction and coasting cycles in ramps (or braking and coasting cycles in slopes).
This control is also applied when the train reaches the maximum speed of the
track.
The ATO simulator is configured with the real fixed and variable parameters
needed for an accurate simulation. The fixed parameters are:
 Safety distance: to be observed when the train has to brake due to a
maximum speed reduction.
 ATP safety offset: to be observed under the maximum speed
 Positive and negative regulation offsets: The regulation speed cycles
previously mentioned are applied between the speed limits defined by these
offsets.
On the other hand, the variable parameters (speed commands to be designed)
are:
 Braking command: deceleration rate during the braking process.
Energy Management in the Train Operation 57
 Regulation speed. The target speed of the train is the minimum between the
maximum velocity and the regulation one.
 Coasting speed: When reached, traction is turned off up to remotoring
speed.
 Remotoring speed: When reached, traction is turned on up to coasting
speed.
The ATO simulator model also includes the necessary rules and algorithms to
emulate the real operation of the equipment calculating:
 Coasting and remotoring points: where the traction is turned off or turned
on.
 Starting braking points. The actual ATO braking algorithms are replicated to
decide at each simulation step if the train has to brake to observe a speed
reduction or a stopping point.
2.2 Train dynamic model
I
a g
M
F F F
a
∑
− −
=
) (
(1)
This module recalculates train speed and position at each simulation step.
They are obtained from the train acceleration a, which is the result of the traction
force F minus the rolling and aerodynamic resistance to train movement F
a
and
track gradient and curvature resistances F
g
as shown in eq
(1). M
I
stands for train mass plus the rotational inertial effect.
The traction motor curve (maximum force/speed) is given as input data in
order to calculate F. Power and grip limitations are taken into account in this
curve. The resistance to train movement F
a
is modelled as a quadratic function of
the velocity with nonnegative parameters depending on each particular train. The
track gradient resistance F
g
is the resistive force due to gravity, positive for
ramps and negative for slopes.
While the train is braking, the ATO system regulates the force continuously
in order to obtain the braking command reference. The simulator applies
precisely this deceleration rate, assuming that the train is able to supply the
needed braking force.
2.3 Consumption model
The energy consumption E is recalculated according to the time increment ∆t
and the current I at each simulation step. The current is obtained from the line
current/speed curve. A constant line voltage U is assumed.
t U I E ∆ = · · (2)
2.4 Operational and comfort restrictions
After simulating every ATO speed profile, a validation model checks the
fulfilment of the operational and comfort restrictions specified. In the study case
58 Power Supply, Energy Management and Catenary Problems
of Madrid Underground, the current restrictions are the minimum speed along
curves (avoiding to wear the track) and the jerk restriction for a comfortable trip.
3 ATO speed commands design and case study
Once interstation runtimes and energy consumption have been simulated for all
the possible ATO speed profiles, a set of them are selected to be programmed in
the ATO equipment.
Decision theory techniques have been used to solve this multicriteria problem
finding an appropriate tradeoff between costs (energy consumption) and run
times because the longer the runtime is, the lower the energy consumption is.
According to that, the optimal solutions are over the Pareto curve which
represents the minimum consumption for each runtime. An example is given in
Figure 1.
The proposed procedure follows three criteria: domination, sensitivity and
uniform distribution of runtimes. They have been applied to optimize the speed
commands design of the Line 1 of Madrid Underground. The following
description of the procedure will be illustrated with this realistic application. The
energy consumption and runtimes of the current speed profiles and the proposed
ones will be compared in order to value the achievable energy saving.
In Madrid Underground there are four alternative speed profiles per inter
station to be designed, with increasing runtimes from the first (flat out, the
fastest) to the fourth (slowest). The first speed profile is obtained applying
maximum speed and deceleration conditions.
3.1 Domination criterion
According to this criterion, optimal solutions are over the Pareto curve which is
formed by the solutions with less energy consumption and approximately the
same runtime of all the possible ones. Solutions not located on the curve are
said to be dominated and are discarded.
Energy Consumption (kWh)
Plaza de Castilla (Platform 1)
0,00
2,00
4,00
6,00
8,00
10,00
12,00
60,0 65,0 70,0 75,0 80,0 85,0 90,0 95,0 100,0 105,0
Run time (s)
C
o
n
s
u
m
p
t
i
o
n
(
k
W
h
)
Possible profiles
Current profiles
Pareto Curve
Figure 1: Current speed profiles dominated by others with less consumption
and Pareto curve.
Energy Management in the Train Operation 59
An example of the case study is given in Figure 1 (from Plaza de Castilla
station to Valdeacederas station). The marked 4
th
speed profile currently in
service, is dominated by two others over the Pareto curve with almost the same
runtime. The proposed solution consumes 28% less energy than the current one
(Table 1).
Table 1: 28% energy savings with an alternative speed profile.
Braking Coasting Motoring Time Consumption
Profile m/s2 km/h km/h s kWh
Current 0.65 32 10 75.0 4.99
Alternative 0.7 35 20 74.4 3.6
Saving 27.89%
3.2 Consumption sensitivity criterion
This is the criterion to be applied for the slowest speed profile selection.
In the time/consumption graph, the slope of the Pareto curve progressively
decreases from the fastest speed profile as the runtime increases. That is to say,
solutions near the minimum runtime speed profile have high marginal energetic
cost associated per second. These marginal costs go down reaching almost 0 at
the end of the curve. The proposed criterion places the slowest speed profile
where the Pareto slope becomes almost flat. This strategy guarantees energy
savings when trains are held up for traffic regulation purposes.
The maximum runtime gap between the fastest and the slowest speed profile
is limited in practice by operational requirements, so the slowest profile must be
moved and placed before the flat slope of the curve if it is necessary to observe
this restriction. Figure 2 shows an example of the case study (from Tetuán to
Estrecho stations). The Pareto curve becomes flat for runtimes greater than 140
s, and the fastest one takes 70 s. However, the maximum runtime gap defined
for Madrid Underground is 20 s. Therefore, the fourth speed profile must take
less than 140 s and it is placed 20 s slower than the fastest one, observing the
operational requirements. This selection involves that the fourth profile can be 7
s slower than the current one, and this implies 22% of energy savings (Table 2).
Energy Consumption (kWh)
Tetuán (Platform 1)
0,00
2,00
4,00
6,00
8,00
10,00
12,00
68,0 88,0 108,0 128,0 148,0 168,0 188,0
Run time (s)
C
o
n
s
u
m
p
t
i
o
n
(
k
W
h
)
Possible profiles
Current profiles
Alternative 4th profile
Figure 2: Alternative 4
th
speed profile with less consumption.
60 Power Supply, Energy Management and Catenary Problems
Table 2: An alternative speed profile 7 s slower achieves 21.7% of savings.
Braking Coasting Motoring Time Consumption
Profile m/s2 km/h km/h s kWh
Current 0.65 42 10 83.1 5,98
Alternative 0.75 37 30 90.5 4.68
Saving 21.7%
Time 7
3.3 Temporal uniform distribution criterion
Once the fastest and the slowest speed profiles have been selected in accordance
with the previous criteria, the remaining profiles are designed applying the
uniform distribution criterion. The speed commands must be selected in order to
obtain a design with a uniform distribution of the runtimes over the Pareto
curve. This design allows a proper operation of the traffic regulation system.
Energy Consumption (kWh)
Sol (Platform 1)
9,00
9,50
10,00
10,50
11,00
11,50
12,00
12,50
13,00
13,50
72,0 77,0 82,0 87,0 92,0 97,0 102,0 107,0 112,0 117,0
Run time (s)
C
o
n
s
u
m
p
t
i
o
n
(
k
W
h
)
Possible profiles
Current profiles
Figure 3: Current speed profiles.
Energy Consumption (kWh)
Sol (Platform 1)
9,00
9,50
10,00
10,50
11,00
11,50
12,00
12,50
13,00
13,50
72,0 77,0 82,0 87,0 92,0 97,0 102,0 107,0 112,0 117,0
Run time (s)
C
o
n
s
u
m
p
t
i
o
n
(
k
W
h
)
Possible profiles
Proposed profiles
Figure 4: Proposed speed profiles designed.
In the study case, the uniform distribution criterion is applied to design two
profiles (the second and the third) between the two profiles previously selected.
Figure 3 shows the current profiles in service between Sol and Tirso de Molina
stations. In the proposed design shown in Figure 4, second and third profiles are
selected to obtain time differences of 5 s approximately.
Energy Management in the Train Operation 61
3.4 Redesign of the current ATO speed commands of the Line 1 of
Madrid Underground
The design procedure has been applied to redesign the ATO speed commands of
all the trips between the 26 stations of Line1 of Madrid Underground in order to
value the expected energy savings. The discrete possible values of the ATO
equipment parameters in service are shown in Table 3. The combination of them
supplies 220 possible speed profiles to be simulated between each two stations.
The average simulation time for these 220 profiles, for one interstation, is 4s.
Table 3: Configuration parameters of the ATO in service.
Braking command
(m/s2) 0,8 0,75 0,7 0,65
Regulation Speed (km/h) 62 60 57 55 52 50 47 45 42 40 37 35 32 30 27 25
Coasting speed
(km/h) 57 55 52 50 47 45 42 40 37 35 32 30 27 25 22 20
Remotor speed
(km/h) 30 20 10
Table 4: Summary table. Average energy savings and runtime increase in
the proposed speed profiles in Line 1.
Designed Profiles
Consumption Time Consumption Time Consumption
Profile kWh s Km/h % %
1 225.4 1665.5 252.4 0.5 10.7
2 176.3 1794.6 196.5 0.4 10.3
3 147.0 1915.8 159.1 0.6 7.6
4 128.3 2032.5 145.9 1.4 12.1
677.0 7408.5 753.9 0.5 10.2
Current Profiles Diferences
The simulation results show that 70% of the redesigned speed profiles save
more than 10% of energy, while 26% of them save more than 20%. Overall, an
average of 10% energy savings is expected increasing the runtimes only 0.5% as
Table 4 shows. Moreover, not only an energy benefit is obtained, but also speed
profiles are timeuniformly distributed.
4 New ATO configuration parameters
Although the new speed profiles for Line 1 have been designed according to the
current possibilities of the ATO equipment, the simulator allows researching and
justifying further modifications and improvements for future ATO systems. This
is another advantage of a computeraided design based on accurate simulation.
In Madrid Underground the ATO system applies a unique set of speed
commands for each speed profile between two stations that is executed during
the whole trip. A set is composed of a brake deceleration and a pair of coasting
and motoring or a regulation speed command. However, being possible to define
different sets for a single trip so that they could be on active in particular sections
of the trip, more alternative speed profiles would be obtained. In particular,
defining the activation of a coasting command set at the end of the trip, just
before the final breaking, it is possible to replace braking by coasting. For
62 Power Supply, Energy Management and Catenary Problems
instance, the feasible profiles considering the ATO in service in Line 1 between
Sol and Tirso de Molina stations are plotted in Figure 3. In Figure 5, new speed
profiles are obtained defining different activation sections for the coasting
commands sets. The number of solutions below the previous Pareto curve rises
notably. In particular, the third profile could be redesigned, saving 17% of
energy with the same runtime. These speed profiles are shown in Figure 6.
Energy Consumption (kWh)
Sol (Platform 1)
8,00
9,00
10,00
11,00
12,00
13,00
14,00
72,0 82,0 92,0 102,0 112,0 122,0 132,0 142,0 152,0 162,0 172,0
Run time (s)
C
o
n
s
u
m
p
t
i
o
n
(
k
W
h
)
Possible profiles
Current profiles
Figure 5: Simulated speed profiles with a starting point of commands set.
0,0
5,0
10,0
15,0
20,0
25,0
30,0
35,0
40,0
45,0
50,0
6
6
0
5
,
0
6
6
5
5
,
0
6
7
0
5
,
0
6
7
5
5
,
0
6
8
0
5
,
0
6
8
5
5
,
0
6
9
0
5
,
0
6
9
5
5
,
0
7
0
0
5
,
0
7
0
5
5
,
0
7
1
0
5
,
0
7
1
5
5
,
0
7
2
0
5
,
0
7
2
5
5
,
0
Position (m)
S
p
e
e
d
(
k
m
/
h
)
Alternative speed profile
Maximum speed
Current speed profile
Figure 6: Comparison between the current speed profile and an alternative
one with a starting point of commands set.
5 Conclusions
The study shows the importance of a detailed model of the particular ATO
system in order to obtain a realistic simulation that allows calculating slight
differences between alternative speed profiles. In the case study of Madrid
Underground, these differences can be a few seconds.
In addition, current ATO equipments provide only a certain range of discrete
configuration values. Thus, there is a finite and relatively short number of
possible speed profiles that can be exhaustively simulated. Decision theory
Energy Management in the Train Operation 63
techniques can be directly applied to select solutions finding a tradeoff between
runtimes and energy costs.
From the redesign of the ATO speed commands of Line 1 of Madrid
Underground it is possible to conclude that an improvement of about 10% of
energy consumption is expected without degrading runtimes. Even 20% of
savings are expected with 26% of the redesigned speed profiles. In high
frequency metro lines the programmed speed profiles are repeated systematically
many times, so the energy savings become a relevant aspect.
Finally, the proposed design procedure based on simulation allows the study
of new features and configuration parameters to be implemented in future ATO
equipments, with the aim of obtaining more energy efficient speed profiles, as it
has been shown in the case study.
References
[1] P. Howlett, “The Optimal Control of a Train,” Annals of Operations
Research, vol. 98, pp. 65, 2000.
[2] E. Khmelnitsky, “On an Optimal Control Problem of Train Operation,”
IEEE Transactions on Automatic Control, vol. 45, pp. 1257, 2000.
[3] P. Lukaszewicz, “Energy Consumption and Running Time for Trains,” in
KTH, Department of Vehicle Engineering. Royal Institute of Technology,
Stockholm, 2001, pp. 153.
[4] F. de Cuadra, A. Fernandez, J. de Juan, and M. A. Herrero, “Energysaving
automatic optimisation of train speed commands using direct search
techniques,” in Computers in Railways V  Vol.1 Railway Systems and
Management, 1996, pp. 337–346.
[5] Y. V. Bocharnikov, A. M. Tobias, C. Roberts, S. Hillmansen, and C. J.
Goodman, “Optimal driving strategy for traction energy saving on DC
suburban railways,” IET Electric Power Applications, vol. 1, pp. 675, 2007.
[6] H. Seong Ho, B. Yun Sub, B. Jong Hyen, A. Tae Ki, L. Su Gil, and P. Hyun
Jun, “An optimal automatic train operation (ATO) control using genetic
algorithms (GA),” in Proceedings of IEEE. IEEE Region 10 Conference.
TENCON 99. `Multimedia Technology for AsiaPacific Information
Infrastructure’ (Cat. No.99CH37030), vol.1.
[7] C. S. Chang and S. S. Sim, “Optimising train movements through coast
control using genetic algorithms,” IEE ProceedingsElectric Power
Applications, vol. 144, pp. 65, 1997.
[8] K. K. Wong and T. K. Ho, “Coast control for mass rapid transit railways
with searching methods,” IEE ProceedingsElectric Power Applications,
vol. 151, pp. 365, 2004.
64 Power Supply, Energy Management and Catenary Problems
Charge/discharge control of a train
with onboard energy storage devices
for energy minimization and consideration
of catenary free operation
M. Miyatake, K. Matsuda & H. Haga
Sophia University, Japan
Abstract
The optimal operation of rail vehicle with onboard energy storage device mini
mizing total energy consumption is discussed in this paper. Until now, not enough
research deals with the optimal control of the devices. The authors have developed
the mathematical model based on a general optimization technique. In our study,
the electric double layer capacitor (EDLC) is assumed as an energy storage
device, because of its high power density etc. The proposed method can determine
the optimal acceleration/deceleration and current commands at every sampling
point under ﬁxed conditions of transfer time and distance. The authors have also
modiﬁed it for applying to catenary free operation. Using the proposed methods,
simulations were implemented in some cases. The trend of optimal solutions such
as values of control inputs and energy consumption is ﬁnally discussed.
Keywords: power management, onboard energy storage, optimization, energy
saving operation, supercapacitor, catenary free operation.
1 Introduction
Electrical regenerative braking has reduced total energy consumption in electric
railway systems. However, if the energy is not absorbed by another train, catenary
voltage rises and regenerative failure is occurred under DC power feeding system.
One of the way for absorbing regenerative energy is to use energy storage. Regen
erative energy is stored in the energy storage and reused in the next acceleration.
The energy storage decreases the loss of circuit resistance by compensating voltage
drop. It also prevents regenerative failure even if substations cannot absorb energy.
Energy saving effect as well as preventing regenerative failure is expected.
Energy Management in the Train Operation 65
C
R
1
Chopper
R
0
R
0
V
S
V
S
R
C
R
2
Train
V
T
o
V
c I
Substation1
Substation2 (a) Circuit model
SS1 AS DS SS2
L
L
a
L
b
AS:arrival station
DP:departure station
SS:substation
5km 5km 2km
(b) Location of stations and substations
Figure 1: Modeling of a feeding circuit with one train between substations.
Some research projects on the application of the energy storage devices to
railway systems have been reported in [1–6]. Most of them discussed reasonable
circuit conﬁguration and sizing of energy storage system, however, very few
papers that deal with optimal charging/discharging control of the energy storage
can be found. The charging/discharging command of energy storage affects the
energy consumption and may inﬂuences the optimal speed proﬁle, the trajectory
of a train in the velocityposition state space. The authors pointed out that the
charge/discharge command and vehicle speed proﬁle should be optimized together.
There are a few papers that deal with the energysaving vehicle operation with
a kind of optimization in [7–10]. However, they did not consider the control of
energy storage.
When onboard energy storage is used, catenary free operation technique is
sometimes used. Energy management control is signiﬁcant in this operation
because the train have to run with very limited onboard energy. However, no papers
can be found that optimize train speed control.
The authors have developed the mathematical model composed of DC power
feeder and energy storage that was already reported in [11]. In this paper, the
authors introduce the simulation results under different condition from [11], add a
few discussion to [11]. The authors also modify the model for applying to catenary
free operation.
2 Modeling of energy storage and DC feeding circuit
The EDLC is assumed in the modeling of energy storage in this study. It has
the characteristics of maintenancefree, long lifetime, quick charge/discharge,
lower energy density than that of batteries at present, and wide range of terminal
voltage regulation. The fact shows the difﬁculty in using EDLC as a main power
source of high speed vehicles. However, if it is used with other main energy
sources, the EDLC is expected as one of the most promising auxiliary devices for
transportation systems.
66 Power Supply, Energy Management and Catenary Problems
INV.
Chopper EDLC
IM
DC link
INV.
EDLC
IM
(a) model under catenary (b) model of catenary free operation
DC link
Chopper
Figure 2: Circuit model of an onboard EDLC train.
A DC feeding circuit is modeled with one train between substations. The model
circuit appears in Figure 2(a). In this ﬁgure, V
s
and R
0
are the supply voltage
and the internal resistance at a substation respectively. The values of R
1
and R
2
are equivalent resistances of feeder and return circuits. These resistance values are
proportional to the distance between the train and substation. The constants C and
R
c
are the capacitance and internal resistance in the capacitor respectively. It is
necessary to convert voltage by using a bidirectional chopper because the voltage
difference between the DC link and EDLC is high. The motorinverters of the train
were modeled as a current load that helps solving circuit equations simply.
If battery is applied instead of EDLC, slight change of the model enables us to
deal with the battery system. We have only to change the capacitor in Figure 1 to
a voltage source and modify some equations.
3 Formulation of the operation under catenary
3.1 Deﬁnition of variables
The optimal control problem is formulated from the circuit model. Variables are
deﬁned as follows. Control inputs n and u determine the acceleration/deceleration
force and charging/dischrging current, respectively. State variables x, v and V
c
indicate the train position, speed and capacitor voltage, respectively. The variable
V
T
is the catenary voltage at the train. It is treated as an auxiliary state variable to
avoid complexity in solving circuit equations analytically, although it is derived
by solving circuit equations. It is derived by adding circuit equations in the
optimization problem as the constraints.
3.2 Optimal control problem
The optimal control problem is described as the following mathematical formula
tion.
Minimizing the objective function
J =
T
0
V
s
I
s
(x, V
T
)dt (1)
Energy Management in the Train Operation 67
Subject to the following equality and inequality constraints
˙ x = v, ˙ v = nf
max
(n, v, V
T
) −r(v) (2)
˙
V
c
= −I
c
(u)/C (3)
P
T
(n, v, V
T
) = P
S
(x, V
T
) +P
C
(u, V
c
) (4)
x(0) = 0, v(0) = 0, V
c
(0) = V
c init
(5)
x(T ) = L, v(T ) = 0, V
c
(T ) = V
c f inal
(6)
−1 ≤ (n, u) ≤ 1 (7)
V
T min
≤ V
T
≤ V
T max
(8)
V
c min
≤ V
c
≤ V
c max
(9)
0 ≤ x ≤ L, v ≥ 0 (10)
where
I
s
, I
c
currents supplied from substations and EDLC current;
f
max
maximum acceleration/deceleration force;
r running resistance per unit weight of the train;
P
T
electric power supplied to motorinverters of the train;
P
s
, P
c
power from substations and EDLC;
V
T min
, V
T max
lower and upper limitation of the catenary voltage;
V
c min
, V
c max
lower and upper limitation of the capacitor voltage;
V
c init
, V
c ﬁnal
ﬁrst and ﬁnal values of the capacitor voltage;
L, T distance and running time between the departure and arrival
stations.
The objective function is sum of supplied energy from two substations given
as (1). Equality constraints are given as (2)–(6). Equation (2) is a motion equation
of the train. Gradient can be considered as including the inﬂuence to the running
resistance r. The capacitor voltage is given as the (3). Equations (5) and (6)
describe the initial and ﬁnal conditions of state variables. The constraint (9) gives
the terminal EDLC voltage as well as the initial one. Inequality constraints of
control inputs, state and auxiliary variables are shown in (7)–(10).
The functions related with circuit equations are the following equations.
P
T
(n, v, V
T
) =
Mv · nf (n, v, V
T
)η
m
(n ≥ 0)
Mv · nf (n, v, V
T
)/η
g
(v) (n ≤ 0)
(11)
P
s
(x, V
T
) =
V
s
−V
T
R
0
+R
1
(x)
+
V
s
−V
T
R
0
+R
2
(x)
(12)
R
1
(x) = (L
a
+x)r
0
R
2
(x) = (L −x +L
b
)r
0
(13)
68 Power Supply, Energy Management and Catenary Problems
P
c
(u, V
c
) =
V
c
I
c
(u)η
ch
(u ≥ 0)
V
c
I
c
(u)/η
ch
(u ≤ 0)
(14)
I
c
(u) = uI
c max
(15)
Here, η
m
and η
g
(v) are motorinverter efﬁciency in accelerating and braking
respectively. The constant M is the total weight of the train including onboard
energy storage. The regenerative efﬁciency η
g
must be treated as the function of
speed v for considering electropneumatic blended braking. The constant η
ch
is the
chopper efﬁciency assumed as constant and I
c max
is the rated value of the EDLC
current.
The optimal control problem is discretized using sampling time t and solved
by the Sequential Quadratic Programming (SQP). SQP is an optimization method
to solve general nonlinear programming problems. Please see [11] for the detailed
procedure.
4 Consideration of catenary free operation
If the train with onboard energy storage runs off the catenary, so called catenary
free operation, the mathematical formulation should be modiﬁed because the
circuit topology is changed to Figure 2(b). The objective function (1) should be
changed to (16) because no power is supplied from the catenary.
J =
T
0
V
c
I
c
dt (16)
In this study, bidirectional chopper is stopped to avoid switching loss. Then,
V
T
is determined by (17). The equations (11), (14) and (15) should be changed
to (18), (19) and (20), respectively.
V
T
= V
c
−R
C
I
c
(17)
P
s
= I
s
= 0 (18)
P
c
= V
c
I
c
(19)
I
c
= P
T
/V
T
(20)
Some other equations such as (8) and (13) must be eliminated for topological
change of the circuit.
In this operation, the control input u is ﬁnally eliminated and only the notch
control input n is left. Therefore, the Dynamic Programming (DP) that was already
proposed in [8] can be applied to the modiﬁed problem. Please refer [8] for details.
In the application of DP in this case, the ﬁnal EDLC voltage V
c ﬁnal
is given
to implement backward search from t = T to 0. The initial EDLC voltage V
c init
cannot be given together with V
c ﬁnal
because the value is decided by the trajectory
of n. When V
c init
is initially given instead of V
c ﬁnal
, V
c
(T ) should be adjusted so
as to satisfy V
c
(0) = V
c init
.
Energy Management in the Train Operation 69
Table 1: Speciﬁc parameters.
f eed in g c ir cu it an tr a in o p e r a tio n
t 1[s] R
0
0.03[] η
g
≤ 90 % η
ch
95%
V
s
1500[V] r
0
0.04 η
m
90 % M 250[Ton]
[m/m]
EDLC
C 32.3[F] V
c max
560[V] V
c init
560[V] weight 500[kg]
I
c max
500[A] V
c min
300[V] V
c f inal
560[V] R
c
0.3[]
Another problem in using DP is that the EDLC voltage much affect accel
erating/decelerating ability. The proposed method in [8] assumes this ability
as constant, however, the accelerating/decelerating ability must be calculated in
ﬁnding the optimal control input at each lattice point in state space.
5 Simulations of optimal operation under catenary
5.1 Condition of simulation
Speciﬁc parameters are tabulated in Table 1. In the simulations, a train runs on a
straight line without speed limitations and gradients for simple analyses. The ﬁnal
capacitor voltage is given to equal the initial one.
In the acceleration/deceleration characteristics, electropneumatic blended
braking system with the air supplement control is assumed. Only if the
regenerative braking force is not enough for the speciﬁc braking force, air brake
works. The detailed value of characteristics is shown in [11]. It is also assumed
to use receptive substations that are now in the initial state of practical application
for relaxation of constraints.
Two cases are prepared as tabulated in Table 2 for evaluation under various
conditions. Cases A and B are the optimization of the train without and with the
capacitor, respectively.
Table 2: Conditions and evaluated energy consumption in each case.
T [s] EDLC minimum total energy energy saving
value of V
c
[V ] consumption [MJ] in %
case A 120 without – 37.56 –
case B 120 with 500 35.43 5.67
case 1 130 without – 27.55 –
case 2 130 with 460 27.45 0.35
Cases 1 and 2 were already reported in [11].
70 Power Supply, Energy Management and Catenary Problems
Figure 3: Graphs of optimal control inputs and state variables.
5.2 Optimization results
Optimization results are shown in Table 2 and Figure 3. In Figure 3, the graphs
of control inputs n and u, catenary voltage at train pantograph, capacitor voltage,
train speed, and train power at inverter input are drawn.
The optimal control input n in both cases consists of the maximumacceleration,
reduced acceleration by degrees, coasting and maximum deceleration. The results
are consistent with the results of previous paper [8]. Very little difference of the
optimal control input n can be seen in cases A and B. Regarding the control input
u, the higher the absolute value of power to the train is, the larger the absolute value
of current is. Qualitatively, this trend is proper, because the energy loss by current
through the feeder reduced. Substations supply the power with a small current for
charging the EDLC when the train coasts.
Despite the lower limit value of the capacitor voltage V
c min
is set to 300[V], the
EDLC stops discharging when the EDLC voltage drops to about 460[V]. The efﬁ
ciency of the capacitor itself is reduced according to the voltage drop of the EDLC.
These results can be compared with those of previously reported paper [11]. The
only difference between cases A and B and cases 1 and 2 in [11] is the running
time T . In [11], T is 130 [s], 10 second longer than this study. The EDLC voltage
V
c
, in case B, drops much lower than that in case 2. The stored energy in EDLC
Energy Management in the Train Operation 71
Table 3: Speciﬁc parameters.
general conditions
t 1[s] η
m
70 % η
g
≤ 70 % M 30[Ton]
EDLC
C 40[F] V
c max
600[V] weight 300[kg]
I
c max
500[A] V
c init
600[V] R
c
0.1[]
x
0m 500m 1000m
T
1
T
2
T
s
section A section B
Figure 4: Running condition of tramcar.
is more effectively used in case B. Regarding energy consumption in Table 2,
the energysaving effect by introducing EDLC is 5.67%, much improved than that
in [11]. Fromthe comparison, it is derived that the shorter the margin time between
stations is, the more effectively used the capacity of the EDLC.
6 Simulations of optimal catenary free operation
6.1 Condition of simulation
In this study for catenary free operation, a tramcar with EDLC is assumed. Speciﬁc
parameters are tabulated in Table 3. The tramcar runs 1 km without suppled power
from substations as shown in Figure 4. It runs within T
1
and T
2
at the sections A
and B respectively and stops once between sections at x = 500[m] for T
s
= 10[s].
In the simulation, the ratio of T
1
and T
2
is changed while total time T
1
+ T
2
+
T
s
= T
1
+ T
2
+ 10 is ﬁxed at 200[s] in order to ﬁnd the optimal distribution of
margin time for energysaving operation by sensitivity analysis.
6.2 Optimization results
The simulation results are shown in Figure 5. Figure 5 (a) shows a sample of the
optimal control input u in section A in case of T
1
= 100[s]. Figure 5 (b) is a graph
that indicate relation between T
1
and energy consumption.
In this assumption of simulation, T
1
= T
2
= 95[s] is the condition of equal
division of margin time. From Figure 5 (b), it is observed that the minimum point
appears when T
1
is 95[s]. However, energy consumption in T
1
> 95[s] is larger
than that in T
1
< 95[s]. The result is mainly caused by that the EDLC voltage at
x = 500[m] is lower than that at x = 0[m]. If the tramcar must run faster with
72 Power Supply, Energy Management and Catenary Problems
c
o
n
t
r
o
l
i
n
p
u
t
n
1
0.5
0
0.5
1
40 60 100 20 0
time [s]
80
(a) A sample of control input.
95 100 90 85
0
10
20
30
40
T
1
[s]
e
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
[
M
J
]
95 90 100 105 T
2
[s]
(b) Sensitivity analysis in changing T
1
.
Figure 5: Simulation results.
lower voltage of the EDLC, larger current ﬂowing from the EDLC increase loss
by internal resistance.
7 Conclusion
This paper presents the optimal train operation with EDLC minimizing energy
consumption. As a result, it is found that the energysaving effect by using
the EDLC is strongly inﬂuenced by the margin time of train schedule. On the
other hand, regarding the acceleration/deceleration command, very few difference
between with and without EDLC is observed.
Optimization of train speed proﬁle in catenary free mode is also mentioned in
this paper. It indicates that the distribution of margin time in each section can
be optimized with the modiﬁed optimization model. The model can be used for
planning of train schedule.
The knowledge extracted from the trend of optimization results will be applied
to the design and parameter tuning of future charge/discharge controllers for
energy storage.
References
[1] M. Ogasa: “Energy Saving and Environmental Measures in Railway Tech
nologies: Example with Hybrid Electric Railway Vehicles” IEEJ Transac
tions on Electrical and Electronic Engineering, Vol.3, No.1, pp. 15–20, 2008.
[2] Y. Taguchi, H. Hata, S. Ohtsuyama, T. Funaki, H. Iijima and M. Ogasa:
“Simulation results of Novel Energy Storage Equipment SeriesConnected
to the Traction Inverter” in Proc. of EPE2007, No. 554, Aalborg, Denmark,
2007.
[3] M. Steiner and M. Klohr: “Energy Storage System with UltraCaps on Board
Energy Management in the Train Operation 73
of Railway Vehicles” in Proc. of EPE2007, Aalborg, Denmark, 2007.
[4] J. Tauﬁq: “Power Electronics Technologies for Railway Vehicles” in Proc. of
PCCNagoya 2007, No. LS555, Nagoya, Japan, 2007.
[5] S. Sone, T. Sato and J. Kouyama:“Proposal and Discussion of HighSpeed
Regenerative Braking for Realizing Genuine Pure Electric Braking” IEEJ
Technical Meeting on Transportation and Electric Railway, No. TER0526,
pp. 71–74, 2005 (in Japanese).
[6] Y. Sekijima, M. Inui, I. Aoyama, Y. Monden: “A trial of Regenerated Energy
storage with an Electric Double Layer Capacitor for Rolling Stock” in Proc.
of IEEJ JIASC2006, Vol.1, No. 1OS5, pp. 125–128, Nagoya, Japan, 2006
(in Japanese).
[7] E. Khmelnitsky:“ An Optimal Control Problem of Train Operation” IEEE
Trans. on Automatic Control, Vol.45, No.7, pp.1257–1266, 2000.
[8] H. Ko, T.Koseki and M. Miyatake: “Application of Dynamic Programming
to Optimization of Running Proﬁle of A Train”, Computers in Railways IV,
WIT Press, pp. 103–112, 2004.
systems by simultaneous train running time control” Computers in Railways
IX, pp. 885–894, 2004.
[10] M. Miyatake and H. Ko: “Numerical Optimization of Speed Proﬁles
of Inverter Trains Considering DC Feeding Circuit” PCCNagoya 2007,
No.DS839, Nagoya, Japan 2007.
[11] K. Matsuda, H. Ko, and M. Miyatake: “Train Operation Minimizing Energy
Consumption in DC Electric Railway with Onboard Energy Storage Device”
Computers in Railways X, pp. 767–776, 2006.
[9] T. Albrecht: “Reducin g p ower peaks and energy consumptio n in rail transit
74 Power Supply, Energy Management and Catenary Problems
Evaluation of energy saving strategies in
heavily used rail networks by implementing an
integrated realtime rescheduling system
M. Luethi
Swiss Federal Institute of Technology ETH, Zurich, Switzerland
Abstract
The Swiss Federal Railways in cooperation with the Swiss Federal Institute of
Technology ETH has developed an integrated realtime rescheduling system to
simultaneously improve rail network capacity and punctuality. The approach
combines realtime rescheduling (performed after a delay or incident) with very
precise train operation facilitated by providing dynamic schedule information to
train drivers. This information enables train drivers to change their driving
behaviour and adjust their speed based on the new schedule. This can
significantly reduce the number of unnecessary decelerations or stops due to
conflicts. Consequently, traffic flow is improved. In addition, also energy
consumption is reduced because unintended reaccelerations are minimised. This
paper describes results of an analysis performed to calculate the energy savings
possible using the integrated realtime rescheduling system.
Keywords: rail traffic management; train dispatching; energy optimisation
strategies.
1 Introduction
Railway operators have significantly increased passenger and freight service in
the last several years. Consequently, railway infrastructure is being used more
intensively and the system is becoming less stable. Under these conditions a
small initial disturbance can propagate causing substantial knockon delays
throughout the entire network. In order to reduce the impact of these delays, train
dispatchers must react quickly and make decisions based on current conditions.
Given the complexity of this task, research on realtime rescheduling systems for
rail networks has become an important research topic. Good overviews of the
Energy Management in the Train Operation 75
problem and potential solutions are described in D’Ariano [1] and Törnquist and
Persson [2]. The focus for realtime rescheduling systems is to minimise knock
on delays by detecting and solving train conflicts quickly and thereby optimising
usage of rail infrastructure. As realtime rescheduling systems are improved, the
information they provide to dispatchers will change from simply detecting
conflicts to proposing conflict solutions by reordering, rerouting and retiming
of trains based on train predictions and extended algorithms.
Another field of active research focuses on the optimal control of train
operation (driving), see Albrecht [3], Franke et al. [4] or Howlett and Pudney [5]
for more details. In this case, the optimisation objective is to minimise the train’s
energy consumption subject to physical (rolling stock and infrastructure) and
temporal (timetable) constraints. Today, especially on heavily used railway
networks, changing conditions and particularly delays cause trains to stop
unnecessarily which wastes energy. Current stateoftheart for energy optimal
driving solutions are only effective for simple systems (e.g. static timetables),
extending these systems to heavily utilised and heterogeneous rail systems will
require developing dynamically changeable schedules in real time.
The Swiss Federal Railways (SBB) in cooperation with the Swiss Federal
Institute of Technology (ETH) is implementing a new integrated realtime
rescheduling system. The new system combines realtime rescheduling with
continual provision of precise and uptodate information to all affected actors
(drivers, guard, infrastructure operators). The new system helps to improve
service quality and capacity by reducing unnecessary signal stops. Since it
reduces unnecessary stops it could also reduce energy consumption. This paper
evaluates the energy savings possible using the integrated realtime rescheduling
system with the support of microsimulation and describes the main influence
factors. Section 2 presents a short overview of the integrated realtime
rescheduling system’s structure and functionalities. Section 3 reviews results of
earlier train energy saving research. Section 4 evaluates the energy savings
possible using the integrated realtime rescheduling system and section 5
presents conclusions.
2 Integrated realtime rescheduling system
The main idea behind the SBB’s integrated realtime rescheduling process is to
continuously provide all actors with an uptodate and conflict free schedule for
all trains. This schedule would provide detailed information about time, speed
and route (with an accuracy of seconds). The new approach combines two
elements:
 Rescheduling trains in realtime after a disturbance, event, incident or
delay;
 Controlling trains and infrastructure so that the dynamically calculated
trajectories (new schedules) are followed with a predefined accuracy.
Figure 1 illustrates the proposed rescheduling system’s structure and data
flows.
76 Power Supply, Energy Management and Catenary Problems
Figure 1: Integrated realtime rescheduling structure and data exchange.
One of the main problems in developing a realtime rescheduling system is
that it must be both very fast and precise. The system must quickly identify any
disturbance, deviation or event. Then it must immediately provide the event
information (containing position, speed and state of train and infrastructure) to
the traffic management system (and finally the rescheduling algorithm). This
information must be as exact as possible to reduce the possibility that the system
will make bad predictions and generate suboptimal schedules. Next, the system
must quickly provide the new schedule information to all necessary actors.
The rescheduling system uses two ideas to improve its overall accuracy. First,
by communicating the new schedule information to train drivers it becomes
possible for drivers to very precisely control their trains; this means that the
rescheduling algorithm can safely predict when the train will arrive or pass
specific points, which greatly increases the accuracy of the entire process.
Second, by dividing the rail network into capacity bottleneck areas and links
connecting them the process of generating a new schedule is simplified and
speeded up [6].
One of the integrated realtime rescheduling system’s most distinctive
features is that it immediately communicates a dynamically changeable schedule
to all actors including the train driver, guard and passengers. When this realtime
schedule information is combined with a supporting assistant system such as a
drivermachineinterface (see [7] for an example developed in the Netherlands)
or used as input for an Automatic Train Operation (ATO) system, the schedule
can be followed very precisely. Therefore, the timetable information or target
trajectory, provided by the system to the drivers has an accuracy of seconds.
Recent improvements in communications technology have made it possible to
quickly and accurately transfer information in both directions between the traffic
management system and trains. Thus, train information such as state and position
can be made available for the traffic management system almost immediately
Energy Management in the Train Operation 77
and train trajectories can be transmitted to drivers without delay. While the
integrated realtime rescheduling system adds a significant level of control to
train operations, it relies on the existing signalling and interlocking principles
and infrastructure. A more detailed description of the integrated realtime
rescheduling system and its benefits is presented in Laube et al. [8], Luethi
et al. [9, 10] and Wuest [11].
3 Static energy optimisation
There is wide variation in the amount of energy a train needs to make a given
trip. In 1999 and 2000, the SBB completed a set of tests on the line Zurich – Zug
– Rotkreuz – Lucerne to identify key factors influencing the amount of energy
needed and to assess various strategies for reducing energy demand [12]. These
tests were carried out using IC2000 intercity trains operated with a Re 460
locomotive and 10 wagons. The distance between Lucerne and Zurich is 57
kilometres, runtime is 48 minutes and there is only one significant grade with 13
per mill. The tests showed a large variation in the energy consumption and
identified the following main influence factors:
 Runtime;
 Usage of electrical braking; and
 Number of unnecessary signal stops.
In addition to these main factors, weather conditions (adhesion) and train load
were also found to influence energy consumption.
Figure 2 illustrates the influence of runtime on theoretical energy
consumption for the track section between Lucerne and Zug. It shows that
energy consumption is highest for trips with the minimal runtime (16.6 minutes)
and that it decreases significantly with longer runtimes until runtime reaches a
certain point (17.25 minutes), and then it continues to decrease although only
slightly
Figure 2: Minimal theoretical energy consumption depending on the
runtime.
78 Power Supply, Energy Management and Catenary Problems
In addition to runtime, the train driver influences energy consumption through
brake usage and the choice of brake used. The Re 460 locomotive has a powerful
electric brake and a pneumatic disc brake where the entire kinetic energy is
transferred into heat. Drivers were asked to use mainly the electric brake, but
pneumatic brakes must be used when operating on minimal runtime schedules
and making signal stops. The test runs showed that brake type has a significant
impact on energy consumption. They also showed, using an elaborated
optimisation algorithm, that it is possible to achieve energy savings of between
10 and 30% for traffic situations without unnecessary signal stops [4].
Algorithms and consequently train drivers are only able to minimise energy
consumption when signals are open and the trains use the preplanned track.
Unexpected speed restrictions (e.g. due to changing a switch position) or closed
signals force drivers to use the pneumatic brake thereby increasing energy
consumption. The tests showed that energy consumption was approximately 10
15% higher for trains having an unnecessary signal stop than for unhindered
runs. On track sections with capacity bottlenecks many trains are forced to make
one or more unscheduled stops. For example over 50% of all IC2000 trains
from Zurich to Lucerne had to stop at Rotsee in the entrance area to Lucerne
station (a major bottleneck) due to delays.
To summarise, the driving strategy (especially reducing unnecessary stops)
significantly influences energy consumption and therefore offers a large saving
potential. However, as the number of trains increases, the number of unnecessary
stops due to delays also increases thus increasing energy use. As outlined in the
following section, integrated realtime rescheduling can provide drivers with
precise and uptodate schedules that reduce unscheduled stops and thereby
reduce energy consumption.
4 Energy optimisation with integrated realtime rescheduling
Railway traffic has grown significantly in the last several years; more trains are
being operated often with no additional infrastructure. This situation increases
the number of delays caused by train conflicts and also causes knockon delays
to propagate more quickly through the entire network. However, the new
integrated realtime rescheduling system makes it possible to effectively
implement measures such as rerouting, reordering and retiming of trains thus
minimising train conflicts and knockon delays.
As outlined in chapter 2, the integrated realtime rescheduling system
continuously provides train drivers with uptodate driving trajectories enabling
them to reduce unnecessary stops. The rescheduling system can use several
different strategies and optimisation criteria in developing these new trajectories,
including:
 Earliest arriving time;
 Minimal energy consumption; and
 Shortest blocking time.
In addition to these criteria, the topology, infrastructure, signalling and
operation rules have significant impact on the final trajectory. Albrecht described
Energy Management in the Train Operation 79
the influence of these trajectories that anticipate train driving [13]. As outlined
earlier, the precise trajectory strategy can be very effectively used before
entering a bottleneck area. For example, the SBB has evaluated the strategy of
slowing down and accelerating a train before reaching a capacity critical section
so that the train passes through the bottleneck area with the maximum allowable
speed. Figure 3 compares traditional train control to this type of dynamic train
control using integrated realtime rescheduling. Figure 3 shows that the train
with integrated realtime rescheduling arrives earlier in the station and uses less
energy. The results show that the time difference can vary between several
seconds up to several minutes depending on the train and infrastructure
characteristics. The energy evaluation is based on braking and reacceleration –
which provide more precise train control – rather than coasting to follow the
train trajectory. However coasting could be added to the algorithms to further
reduce energy use.
Figure 3: Comparison of traditional driving behaviour with speed control as
part of integrated realtime rescheduling.
The goal of this research was to more precisely evaluate the impact of the
integrated realtime rescheduling system (and in particular the point of time the
new schedule is transmitted to the train) on energy consumption. The research
used the OpenTrack train micro simulation application (see [14] for a description
of OpenTrack). The evaluation considered the 18 kilometre track section from
80 Power Supply, Energy Management and Catenary Problems
Rotkreuz to Lucerne. This section includes a single line section starting at Rotsee
(near Lucerne) which causes regular conflicts.
The research evaluated energy consumption based on when the train received
the revised schedule. In all cases it was assumed that the train could enter the
critical single line section exactly one minute later than originally scheduled (in
other words the signal turned green oneminute later than scheduled). Of course,
the earlier a new schedule is generated, the smoother or less deceleration and
reacceleration is needed. The closer a train is to the critical point (Rotsee) when
it receives the new schedule, the more energy it will use since the target speed
(i.e. speed the train should be going to be one minute later) is lower (see Figure
4). This means that train operators must use pneumatic brakes since immediate
and strong braking action is needed.
Figure 4: Possible rescheduling actions depending on the point of time
rescheduling is executed resulting in identical arriving times in
Lucerne.
Figure 5 illustrates the simulation results. It shows the minimal relative
energy consumption depending on the point of time the rescheduling is executed
(i.e. the new schedule is received and acted upon by the train driver). Figure 5
shows that for the first 100 seconds after departure from Rotkreuz, the relative
energy consumption for all alternatives is minimal and identical. This is due to
the fact that the train is accelerating to its target speed. When the rescheduling
information is received between 125 and 160 seconds, the relative energy
consumption increases, but remains approximately constant. During this period,
the reduced target speed varies only marginally and the speed when rescheduling
is executed is stable at 125 km/h.
When the rescheduling is executed between 160 and 200 seconds after
departure from Rotkreuz, the train is in the process of accelerating up to the
maximum line speed of 140 km/h and therefore relative energy consumption
increases more or less linearly. Finally, when the rescheduling is executed
between 200 and 310 seconds, the train must brake and slow down to a lower
and lower speed resulting in a maximal relative energy consumption of up to 50%
Energy Management in the Train Operation 81
higher than in the best case (early rescheduling). After 310 seconds, a full stop
can not be avoided anymore.
Figure 5: Influence of rescheduling point of time on relative energy
consumption for the section Rotkreuz – Lucerne.
To summarise, the rescheduling point of time is a decisive variable in
determining energy consumption. Providing precise train speed control
information early enough so that drivers can reduce their need for braking and
reacceleration to a minimum reduces energy consumption by up to 50%. The
nearer to the critical point where rescheduling information is executed, the more
energy will be consumed (since this requires strong braking and reacceleration).
However, in all cases where a full stop can be avoided, rescheduling will reduce
energy consumption.
Delays and blocked routes often occur in heavily used parts of the rail
network and in bottlenecks where trains interfere with each other. A significant
amount of energy could be saved in these areas by implementing integrated real
time rescheduling. As rail traffic increases, applying integrated realtime
rescheduling systems is becoming increasingly necessary, not only to reduce
knockon delays, but also to save energy. To be most effective realtime
rescheduling should be combined with energy efficient train control based on
elaborated algorithms. Only the combination of both approaches will make it
possible to fully achieve potential energy savings.
5 Conclusion
Railway companies are facing huge challenges, on the one hand demand is
growing as concerns over sustainability, energy use and climate change become
more significant, while on the other hand railways face increasing market
pressure to reduce expenditures and capital costs. In short they must increase
capacity and service quality at minimum cost.
82 Power Supply, Energy Management and Catenary Problems
Integrated realtime rescheduling systems are being developed to increase
infrastructure capacity without making significant investments or causing more
delays. An important element of these systems is providing train drivers with up
todate data and new schedules. This enables drivers to precisely control speeds
thus minimising unnecessary stops.
Unnecessary stops not only reduce network capacity and punctuality, but also
increase energy consumption. Therefore, integrated realtime rescheduling
systems that eliminate or reduce unnecessary stops can also reduce energy
consumption. The research completed for this study showed that on the rail
segment from Rotkreuz to Lucerne, energy consumption could be reduced by up
to 50% with speed adjustments based on decisions of the integrated realtime
rescheduling compared to traditional train control with a full stop and
reacceleration. The research also showed that the point of time when
rescheduling is executed is very important. Large savings are only achieved
when rescheduling is executed early.
Realtime rescheduling should be combined with energy efficient train
control to minimise energy consumption and achieve the largest possible energy
savings.
References
[1] D’Ariano, A. Improving RealTime Train Dispatching: Models, Algorithms
and Applications, Trail thesis series, Delft, 2008.
[2] Törnquist, J., Persson, J. Ntracked railway traffic rescheduling during
disturbances, Transportation Research Part B, 41(3), pp. 342–362, 2007.
[3] Albrecht, T. Ein Beitrag zur Nutzbarmachung Genetischer Algorithmen für
die optimale Planung und Steuerung eines flexiblen Stadtschnellbetriebes,
PhD Thesis, Dresden University of Technology, 2004.
[4] Franke, R., Meyer, M., Terwiesch, P. Optimal Control of the Driving of
Trains, Automatisierungstechnik, 50(12), pp 606–613, 2002.
[5] Howlett, P.G., Pudney, P.J. Energyefficient train control, Springer, Berlin,
1995.
[6] Caimi. G., Burkolter, D., Herrmann, T., Chudak, F., Laumanns, M. Design
of a new railway scheduling model for dense services, Proc. of the 2
nd
International Seminar on Railway Operation Research, Hannover, 2007.
[7] Albrecht, T., van Luipen, J., Hansen, I.A., Weeda, A. Bessere
Echtzeitinformationen für Triebfahrzeugführer und Fahrdienstleiter,
Eisenbahningenieur, 58(6), pp.73–79, 2007.
[8] Laube, F., Roos, S., Wuest, R., Luethi, M., Weidmann, U. PULS 90 – Ein
systemumfassender Ansatz zur Leistungssteigerung von Eisenbahnnetzen,
Eisenbahntechnische Rundschau, 56(3), pp.104–107, 2007
[9] Luethi, M., Laube, F., Medeossi, G. Rescheduling and Train Control: A
New Framework for Railroad Traffic Control in Heavily Used Networks,
Proc. of the 86
th
Transportation Research, Washington DC, 2007.
[10] Luethi, M., Nash, A., Weidmann, U., Laube, F., Wuest, R. Increasing
Railway Capacity and Reliability through Integrated RealTime
Energy Management in the Train Operation 83
Rescheduling, Proc. of the 11
th
World Conference of Transportation
Research, Berkeley, 2007.
[11] Wuest, R. Dynamic Rescheduling based on Predefined Track Slots, Proc. of
the 7
th
World Congress of Railroad Research, Montreal, 2006.
[12] Meyer, M., Roth, M., Schaller, B. Einfluss der Fahrweise und der
Betriebssituation auf den Energieverbrauch von Reisezügen, Schweizer
EisenbahnRevue, 22(8–9), pp 360–365, 2000.
[13] Albrecht, T. The influence of anticipating train driving on the dispatching
process in railway conflict situations, Proc. of the 2
nd
International Seminar
on Railway Operation Research, Hannover, 2007.
[14] Nash, A., Huerlimann, D. Railroad simulation using OpenTrack.
Computers in Railways IX, eds. J. Allan, R.J. Hill, C.A. Brebbia, G.
Sciutto, S. Sone, WIT Press, pp. 775–784, 2004.
84 Power Supply, Energy Management and Catenary Problems
Part B
Power Supply System Analysis,
Design and Planning
This page intentionally left blank
Online temperature monitoring of overhead
contact line at the new German highspeed rail
line CologneRhine/Main
N. Theune, T. Bosselmann, J. Kaiser, M. Willsch, H. Hertsch
& R. Puschmann
Siemens AG, Corporate Technology, Germany
Abstract
This paper reports on the first fiberoptic temperature measurement of overhead
contact line systems at a new ICE3 high speed line of the German Railway. The
installation of Fiber Bragg Grating (FBG) sensors, the online acquisition and
interrogation of data over a one year period is presented.
Keywords: Fiber Bragg Grating, temperature monitoring, overhead contact line,
catenary, power management.
1 Introduction
All railway companies try to achieve higher reliability and flexibility in train
service. As a consequence the quality of train control processes as well as
monitoring equipment for the next generation of high speed railway lines needs
to be improved. Next generation of high speed trains will consume more power,
because of an increased amount of electrical amenities inside the train and  of
course  higher speeds of up to 300 km/h and more. This goal can only be
achieved if the overhead contact lines that provide the energy for the trains are
protected reliably from thermal overload caused by electrical overcurrents and
hotspots at the catenary wire and contact wire. Up to now diagnostic systems for
online acquisition of real temperature data do not exist. As a state of the art,
conventional Digital Protection Devices (DPDs) in substations inhibit a simple
numerical model from which the socalled catenary temperature is estimated
with respect to current load and ambient air temperature [5]. If the estimated
temperature exceeds a certain limit of e.g. T=70°C the power supply of the
overhead contact line is switched off, regardless of the actual temperature on the
Power Supply System Analysis, Design and Planning 87
overhead contact line. If the actual temperature would be higher, thermal
overusage of the overhead contact line system could destrengthen parts of the
catenary construction. Furthermore elongation may cause a malfunction of a
passing train pantograph or lead to a destruction of the contact wire. If the actual
temperature would be lower than the limit, the power supply for the trains would
be switched off needlessly. Both cases are connected to costintensive down time
of the train line. In order to control reliably and adapt to the growing power
needs for high speed trains and to use up to date overhead line systems more
efficiently an accurate online determination of the overhead contact line and
contact wire temperature is essential. The appropriate temperature sensors have
to be compact, should be easily integrable in the catenary construction and must
be able to measure on high voltages of up to 25 kV. In general conventional
temperature monitoring techniques could be applied alternatively, but due to the
high voltage level of the overhead line an enormous amount of effort needs to be
put into the insulation of these sensors or into provision of independent power
supplies etc. [1]. On the other hand FBG sensors offer, because of their vitreous
nature, a simple way to measure on high voltages with a minimum of insulation
material needed. Multiplexed temperature measurements with the help of FBGs
could help to localize hot spots and provide a continuous measurement for the
protection of the power line [3]. If the temperature distribution of a complicated
network structure could be measured the required material effort of new
installations could be optimized which would lead to substantial savings.
Furthermore in peakperiods the power supply could be managed intelligently
with respect to the actual thermal usage of the catenary. The measuring system,
the multiplexed network of sensors and the sensor setup which were
implemented on the new high speed rail line between Cologne and Frankfurt will
be presented and discussed. In the future this technique can be easily adapted to
similar applications of railway transportation systems e.g. long distance railways
with overhead contact lines with still higher demands with respect to current
load.
2 Theory
2.1 Brief theory of fiber Bragg gratings
A change in fiber strain ε ∆ and temperature T ∆ is connected to a change in
center wavelength λ ∆ of the Bragg reflex via the two equations
ε
λ
λ
∆ ⋅ − =
∆
) 1 (
e
p ,
T p
e
∆ ⋅ + ⋅ − =
∆
] ) 1 [( ξ α
λ
λ
,
(1)
whereas
e
p represents the photoelastic constant, α the linear coefficient of
expansion and ξ the thermooptical coefficient of the fiber. Almost in any case
where FBGs are part of a composite material both effects, temperature and strain,
88 Power Supply, Energy Management and Catenary Problems
contribute to a center wavelength shift of the Bragg reflex. In case of the
catenary temperature sensors the first effect needed to be excluded. This was
done through a sensor design that decoupled any stress acting on the sensor
housing, e.g. elongating electric wires, from the sensing FBG element. The
thermal effect of current load on the overhead contact wires and the estimation of
the so called catenary temperature will be discussed in the next section.
2.2 Model for numerical estimation of the catenary temperature
Nowadays the DPDs inside transformer substations evaluate the temperature of
the catenary with the help of a simple numerical model. The electric current
signal I and the ambient air temperature
amb
T are processed in an exponential
function to compute the catenary temperature
cat
T
( )
2
max,
amb cat
) (
1 ) ( ) (
⋅ − ⋅ + =
∞
−
∞
I
t I
e T t T t T
t τ
, (2)
whereas
amb
T represents the ambient air temperature,
∞
T the maximum excess
temperature after ∞ = t , τ the thermal constant of the catenary type, t the time,
I the actual current and
∞ max,
I the maximum current after ∞ = t . In practice, if
the calculated catenary temperature
cat
T exceeds 70°C the relay causes a power
circuit breaker in the switch yard to switch off the current.
The numerical model is uncertain, because various parameters are not accounted
for, e.g. wind velocity, wind temperature, incident solar radiation, precipitation
etc.
3 Measuring setup
The setup for online measurement of the catenary temperature is based on a
polychromator interrogation unit, as in Figure 1:. The polychromator unit
consists of a diffraction grating, a CCD array, an ADC card and a PC for online
temperature evaluation of the spectral intensity information acquired from the
CCD array. As a light source a broadband LED with a center wavelength of
nm 840 = λ and a full width at half maximum of nm 45
FWHM
= λ ∆ was used.
As the whole interrogation unit was used in field, e.g. ambient temperatures were
ranging from C 10° − to C 40° + , an Argon calibration lamp was preferred to
eliminate the temperature dependent shift in wavelength over CCDpixel. The
Argon spectral peaks between 800 and 850 nm were used as a reference in order
to get a calibrated functional dependency between CCDpixel and wavelength. A
4x1 fiber switch operated continuously between the 4 sensor channels. The
whole setup consists of exchangeable card modules and was implemented into a
portable 19inch rack, as in Figure 3:.
Power Supply System Analysis, Design and Planning 89
Figure 1: Polychromator setup with Argon calibration lamp and 4x1 fiber
switch.
Figure 2: Location of FBG temperature sensors on the overhead wires at the
injection pylon – 300m apart from the Limburg transformer
substation.
4 Application of sensors
Figure 2: shows the locations on the catenary where FBG sensors were installed.
The sensors were manufactured in a single ended design that has been used
also in other fields of high voltage applications, e.g. power generators [2], as
redundancy for this first field demonstration in this delicate application was of
major importance. Every single FBG sensor was provided with an individual
optical link to the interrogation unit – located in the transformer substation 
90 Power Supply, Energy Management and Catenary Problems
m 300 apart from the injection pylon. Two FBG sensors are installed in close
thermal contact to the two injection wires, each of
2
mm 240 cross sectional
area. Further two FBG sensors are located at supporting current wires, each of
2
mm 120 cross sectional area. The signals of the four FBG sensors are then
transformed into a temperature value. Simultaneously the catenary temperature
signal provided by the DPD  also located inside the transformer substation – is
calculated. Both data are then submitted  via a GSM link  to a control PC, over
km 300 away, to our lab in Erlangen for analysis.
Figure 3: Left: Polychromator interrogation unit with online FBG
temperature data, Right: DPD with online calculated catenary
temperature data
5 Measurements
In Figure 4: the online catenary temperature data in 12/2002 and 01/2003
acquired from the DPD is shown. In Figure 4:(a) the overall current signal I
with values up to kA 2 at an effective voltage of kV 5 . 16 ≈ U @ Hz 16
3
2
can
be seen [4]. The measured ambient air temperatures
amb
T in (b) range from
C 10° − to C 10° . The calculated excess temperature T ∆ , i.e. the temperature
increase that is due to the current load, is shown in (c).
In Figure 5: the online FBG sensor temperature data of the sensors 1 to 4 is
shown. Overall the temperature data of the four FBG temperature sensors
corresponds very well to the calculated data of the DPD. But in detail the
differences between the fiberoptical and calculated data can be enormous. This
will be shown and discussed in the next section.
Power Supply System Analysis, Design and Planning 91
Figure 4: DPD data between 11/2002 – 01/2003.
Figure 5: Signals of 4 FBG temperature sensors, installed on catenary wires.
6 Discussion of results
In Figure 6: the difference between FBG sensor #2 and the calculated DPD
catenary temperature is plotted. It can be seen that the difference in temperature
between the fiberoptical sensor at the catenary and the calculated DPD data can
reach up to approx. C 8° . If we zoom into the rectangular dashed part in Figure
6:, then a typical behavior of the FBG catenary sensor can be seen, as in Figure
7:. Here the optical sensor takes into account the incident solar radiation, i.e. due
to the solar energy the catenary wire warms up over a period of several hours
only interrupted through periods of cloudiness. At that day  after the sunset at
around 16.30h  the influence of cold wind dominates and chills the catenary
wire.
92 Power Supply, Energy Management and Catenary Problems
Figure 6: Temperature difference between FBG sensor #2 and the DPD
temperature data.
Figure 7:
7 Conclusions
For the first time online temperature measurements with FBG temperature
sensors on railway overhead lines could be demonstrated. All sensors measured
successfully under outdoor conditions over a one year period.
As a first experimental result it was found out that the excess temperatures
due to current load are small compared to ambient sources of temperature
Zoom into the dashed rectangular part of Figure 6.
Power Supply System Analysis, Design and Planning 93
change. In the future this first result will be analyzed under different seasonal
and current load conditions.
In the future FBG sensors will be installed on other hotspot locations at the
overhead contact line. The measurements indicate that there might be a large cost
saving potential for similar future installations. FBG catenary sensors can
provide catenary designers with real data acquired over several years and help
optimize the construction system with respect to reduction of material expenses
and optimization of the power management.
Acknowledgements
The authors thank sincerely Mr. Walter Weisel from German Rail Energy for his
ongoing support during the whole project. Part of this work was financially
supported by the National FAMOS project.
References
[1] Kießling, F.; Puschmann, R.; Schmieder, A.; Schmidt, P.: “Contact Lines
for Electric Railways – Planning, Design, Implementation”. Siemens AG,
Publicis Corporate Publishing, Munich, 2001.
[2] Theune, N.M.; Müller, M.; Hertsch, H.; Kaiser, J.; Willsch, M.; Krämmer,
P.; Bosselmann, T.: “Investigation of Stator Coil and Lead Temperatures
on High Voltage inside Large Power Generators via use of Fiber Bragg
Gratings”. IEEE Sensors 2002 Conference, Conference Proceedings,
Orlando, USA.
[3] Bjerkan, L.: “Measurements of overhead transmission line loads with
Bragg gratings”. Conference Proceedings, SPIE 3746, OFS13, Kyongju,
Korea, P227, 1999.
[4] Kohlhaas, J.; Ortstädt, W.; Puschmann, R.; Schmidt, H.: ”Interoperable
overhead contact line SICATH1.0 for highspeed line Cologne
Rhine/Main”. Elektrische Bahnen 100 (2002), H. 7, S. 249257.
[5] Digital Overhead ContactLine Protection 7SA517, Instruction Manual,
Edition 08/2001, Siemens AG.
94 Power Supply, Energy Management and Catenary Problems
Electric traction energy metering on German
Railways and the impact of European
standardisation on the energy billing process
in Germany
K. Weiland
DB Energie GmbH, Germany
Abstract
With the onset of liberalisation of energy markets and the formation of DB
Energie Inc. as a wholly owned subsidiary of Deutsche Bahn AG (German
Railways), proper billing of electric energy to railwaybased consumers has
become an important issue. Hence, DB Energie had to develop and implement a
revenue metering scheme for traction energy consumed by railroad vehicles. The
impact of international interoperable traffic has been considered.
Keywords: energy meter, metering onboard trains, traction energy, railways.
1 Initial situation
In close cooperation with instrument manufacturers, DB Energie has devised
and developed an onboard metering device for electric railway vehicles. The
device is derived from a standard, utility grade load profile meter used widely in
the power distribution business, but complemented with a GSMmodem and
antenna for radio data transmission.
2 National requirements
The meters have to meet German and European calibration standards and
regulations, and require approval for use in commercial revenue metering
applications. Accordingly, the meters and the adjacent instrument transformers
need to be of Class 1 and 0.5, respectively. Both are officially certified for both
16.7 Hz and 50 Hz applications [1, 2, 4].
Power Supply System Analysis, Design and Planning 95
3 Technical concept
Though the device has fourquadrant capability, the vehicles’ power factor is not
of an immediate interest, because the majority of traction drives employ state of
the art power electronics and thus operate at unity power factor.
The recorded data of the meter is transmitted by a GSM radio to a central
control and data processing station, where the raw data for billing the respective
operators will be generated. The meters allow to store either a 5minute or a
15minute load profile in its builtin memory. The revenue data are downloaded
on a daily basis through a poll request transmitted by the central control station.
The data storage and processing scheme allows then for a variety of analysis.
One may – for instance – derive a graph of the power and energy consumption
for a single train journey
TEMA
compact box
Central
control station
Additional
information,
e.g. train number
Railway
disposition
systems
Information
adding
system
Load
profiles
Billing
system
Consumption
data
Energy
data
management
GSM
Accounting
(SAP)
Figure 1: Structure of the billing system for traction energy.
The energy consumption will be combined with the railroads operative data
like planned and actual schedules, train configuration data like weight, number
of cars etc. Based on this input, DB Energie will produce individual bills for
each train journey on its system. Some customers even use this information for
their own managementinformationsystems as well as for energy saving
programs [8].
As shown in the picture above, the optical interface located on the front panel
of the meter serves for local downloads. Using this data link, the locooperators
have a means to verify and double check billing data received from DB
Energie’s data processing system.
The communication interface allows data transmission in compliance with
Standards EN 6205621 [5] or IEC 87052 [6], and the central control station
96 Power Supply, Energy Management and Catenary Problems
will poll the meter. The specification of metering code follows the national
standard VDEW – Specifications 2.1 (see picture 3) [7].
The maximum power demand of the meter in the instrument transformer
circuit is 50 mVA.
LCD Display
Energy Impulse
LED (red)
Optical Infrared
Interface
(i. e. Notebooks)
GSM Receiver LED
(green)
Energiezähler
Antenne
Energy Meter
Modem
A
n
t
e
n
n
a
Figure 2: Meter used on board trains.
meter
GSMModem
antenna
F. F( 00000000)
0. 0. 1( 00000000)
0. 0. 2( 00000000)
0. 0. 3( 00000000)
0. 0. 4( 00000000)
0. 0. 9( 00136612)
0. 9. 1( 133743)
0. 9. 2( 000208)
1. 8. 1( 00024. 70015*kWh)
2. 8. 1( 00000. 02609*kWh)
P. 01( 991204124050) ( 00000080) ( 15) ( 2) ( 1. 5) ( kW) ( 2. 5) ( kW) ( 1121730) ( 0002001)
( 0. 0012) ( 0. 0000)
P. 01( 991205072451) ( 00000040) ( 15) ( 2) ( 1. 5) ( kW) ( 2. 5) ( kW) ( 1121730) ( 0002001)
( 0. 0000) ( 0. 0000)
( 0. 0005) ( 0. 0000)
( 0. 0011) ( 0. 0000)
( 0. 0011) ( 0. 0000)
( 0. 0010) ( 0. 0000)
( 0. 0010) ( 0. 0000)
( 0. 0038) ( 0. 0000)
P. 01( 991205084821) ( 00000000) ( 15) ( 2) ( 1. 5) ( kW) ( 2. 5) ( kW) ( 1121730) ( 0002418)
( 0. 0041) ( 0. 0000)
absolute meter
readings one time per
transmission
1/4 hvalues incl.
events
status information
meter: ok
Figure 3: Compact meter on board trains and load profile codes.
4 Experiences of DB Energie
As of 2002, electrically propelled vehicles such as locomotives and self
propelled trains (MTUs) operating on the German railway infrastructure have
Power Supply System Analysis, Design and Planning 97
been furnished with this device, allowing DB Energie to meter and analyse the
vehicles individual energy consumption profile including regenerative breaking.
Since January 2003, billing of electrical traction energy is entirely and
exclusively based on metered values. The on board meters have proven their
worthiness for this routine in daytoday business. It turned out, though, that the
availability of the radio communication system is an actual bottleneck in this
remote metering scheme. Once a vehicle moves into an industrial site with
narrow paths, the probability of loss of communications link increases
drastically. This may happen, for instance, when a locomotive is moved into a
service station or workshop. The following graph depicts the proportion of non
connected vehicles (missing radio data link) and the duration of this link outage
over a month’s period.
7,13
6,02
5,16
4,15
4,15
7,31
6,76
4,39
4,28 4,28
5,24
4,36
4,03
3,54
3,21
0
2
4
6
8
10
Quota of non stored metering data in the central database at
a month before
1 Tag
8 Tag
12 Tag
15 Tag
22 Tag
days
days
days
day
days
September 03 Oktober 03 November 03
Figure 4: Proportion of nonconnected vehicles and the duration of link
outage over a month’s period.
5 European standardisation process
Within the Europewide effort to furnish all electric trains with standardized
meters, several manufactures focus on the development of such devices. As a
next step, systems for DC traction power are expected to appear in the market
soon.
In December 2002, CENELEC TC9X called for experts for survey/working
group “Metering On Board Trains”. The target of this standardisation group is to
facilitate the interoperability of trains and the compatibility of technologies in the
various European countries. Under the guidance of Italian Railways (RFI) the
requirements stipulated in the different national regulations were collected.
Subsequently, in June 2003, the survey group specified the scope and content of
the standard to be drafted.
98 Power Supply, Energy Management and Catenary Problems
The paper describes the process of compiling and balancing the different
interests brought up by the participating countries, which in turn are due to
different states of implementation of the European Union’s guidelines
concerning the liberalisation of European Single Market (ESM) in the field of
energy and railway business.
Implementing the German energy market rules as agreed upon by utilities and
traders, private and foreign railway operators are entitled to be billed on metered
electrical energy consumption exclusively.
Denmark, Sweden and Germany are frontrunners in this effort, because their
respective national regulations already mandate onboard metering.
Therefore, the concerned railway infrastructure utilities bear the risk that a
future European standard does not match with today’s technology. The process
of European standardisation will provide a framework for future onboard meters
for trains allowing to employ stateoftheart metering technology. Also, it
follows the known national regulations. Optionally, a meter can be designed as a
modular system consisting of a separate modem, a location module, and an
antenna. Data transmission should comply with Standard EN 6205621 [5] or
IEC 87052 [6]. The protocols can be adopted to different digital telegrams.
This standard permits national system solutions like GSMR [3] for data
transmission, and also the correlation of an individual locomotive (via its meter)
with the momentary energy utility when operating in international, cross border
traffic or for the identification of different energy grid operators.
The present course of standardisation promises an roadmap towards a
common meter suitable for almost any train operating on European tracks.
Centralised control stations with standard databases store common data formats,
and railroad utilities would be able to meet commercial contract requirements for
billing energy to any European customer.
Though this process of international standardisation, a forum has been
established for European railroad operators, for railway infrastructure companies,
and not at least for equipment and instrument manufacturers to start a dialogue
about their individual needs and expectations. The aim is to eventually
harmonise technical and operational standards throughout Europe in order to
facilitate a reliable and dependable, crossborder proof revenue metering scheme
accepted by all participants. A first major step in this direction has been
accomplished by the foundation of CENELEC TC9X.
This ongoing standardisation process will eventually set the basis for an
effective and precise energy billing scheme within the liberalised Trans
European railroad community.
References
[1] Kahmann, M., Zayer, P., Handbuch Elektrizitätsmesstechnik, VWEW
Energieverlag: Frankfurt, Heidelberg and Berlin, pp. 2736, 2003.
[2] Eichordnung vom 12.08.1988 (BGBI. I pp. 1657 ff.).
[3] http://gsmr.uic.asso.fr/.
Power Supply System Analysis, Design and Planning 99
[4] Gesetz über das Mess und Eichwesen (Eichgesetz), Neufassung vom
23.03.1992, (BGBI. I, pp. 408 ff.).
[5] EN 6205621: Electricity Metering – Data exchange for meter reading,
tariff and load control – Part 21: direct local data exchange.
[6] IEC 87052 Data Link Transmission Services.
[7] Lastenheft Elektronische Energiezähler Version 2.1, VWEW
Energieverlag: Frankfurt, Heidelberg and Berlin, 2003.
[8] Proceedings of 2nd UIC Railway Energy Efficiency Conference, Paris, 4
5 February 2004. Energy efficient railways: next steps towards sustainable
mobility.
100 Power Supply, Energy Management and Catenary Problems
Development of feeder messenger catenary
with the auxiliary wire
K. Nishi, Y. Sato & T. Shimada
Railway Technical Research Institute, Japan
Abstract
Feeders are a standard installation of DC electrified railways in Japan in order to
complement the electric capacity of contact wires. Recently in metropolitan area
in Japan, the installed area of the feeder messenger catenary system that has the
function of feeder in messenger has been expanded. Then we developed a new
feeder messenger catenary system with the auxiliary wire that has the function in
which the contact wire is hard to break. For this purpose, we investigated the
best structure of this catenary system first and simulated electric performance.
Then we performed running tests by using a pantograph installed on the current
collection testing equipment of our institute and evaluated the current collecting
performance of the new system. From the results of these tests, it has been
proved that there are no problems in its current collecting performance up to
150km/h.
Keywords: feeder messenger, auxiliary wire, tension shift.
1 Introduction
Now in 1,500V DC electrified railways of metropolitan area in Japan, the
installed area of feeder messenger catenary system that has the function of feeder
in messenger has been expanded. In this catenary system, the number of parts
can be reduced because messenger has the function of feeder [1]. However, it is
apprehended the contact wires of this system may be broken like those of
conventional catenary system when their residual diameter becomes small [2].
Therefore, we devised a new catenary system that has an auxiliary wire. In the
overhead line equipment proposed here, the tension of contact wire shifts to the
connected auxiliary wire to prevent it from breaking when its residual diameter
has become small. We examined the electrical performance and performed
Power Supply System Analysis, Design and Planning 101
feeder messenger
hanger
pulloff ar m
auxiliar y wir e
cont act wir e
5m
feeder messenger
hanger
pulloff ar m
auxiliar y wir e
cont act wir e
5m
cont act wir e
auxiliar y wir e
feeder messenger
pulloff ar m
ear
50mm
hanger
5m
cont act wir e
auxiliar y wir e
feeder messenger
pulloff ar m
ear
50mm
hanger
5m
computer simulation to determine the optimum structure of this system that uses
a harddrawn copper stranded wire PH 590mm
2
as the messenger wire, PH
150mm
2
as the auxiliary wire and a bronze wire GTM170mm
2
as the contact
wire.
For this overhead contact line structure, we performed running tests by
using a pantograph installed on the current collection testing equipment of our
institute and evaluated its current collecting performance. Based on the test
results, we have confirmed that the contact loss rate is less than 1% up to
150km/h and proved that there are no problems in the quality of current
collection. We have also confirmed through measurement that the tension of the
contact wire shifts to the auxiliary wire when its residual diameter becomes
small. Then, it is proved that the contact wire does not easily break. As a result,
we found the possibility of this catenary system to be put into practical use.
2 Structure and feature of the proposed catenary system
The features of this catenary system are described below. An auxiliary wire is
installed between the feeder messenger and the contact wire that is connected to
the auxiliary wire with ears. Then, tension shifts from the contact wire to the
auxiliary wire when the residual diameter of the contact wire becomes small or
the contact wire has been softened by arcs and Joule heat. Therefore, the contact
wire does not easily break and improves the reliability of this catenary system.
This catenary system is classified into the following three types.
First we define “composite type” which makes the auxiliary wire contacts the
contact wire (Figure 1).
Figure 1: “Composite type” feeder messenger catenary.
Figure 2: “Hanger form compound type” feeder messenger catenary.
102 Power Supply, Energy Management and Catenary Problems
5m
cont act wir e
auxiliar y wir e
feeder messenger
dr opper
pulloff ar m
50mm
ear
5m
cont act wir e
auxiliar y wir e
feeder messenger
dr opper
pulloff ar m
50mm
ear
Second we define “compound type” which connects the auxiliary wire with
the contact wire by keeping a distance of 50mm in between. We classify the
structures of these two compound types into two. One is the structure that
connects the auxiliary wire with the messenger by hangers, which we define as
the “hanger form compound type (Figure 2)”. The other is the structure that
connects the auxiliary wire with dropper, which we define as the “dropper form
compound type (Figure 3).”
Figure 3: “Dropper form compound type” feeder messenger catenary.
3 Analysis of the quality of current collection
3.1 Catenary composition
By computer simulation, we calculated the contact loss rate in each feeder
messenger catenary form in Figures 1~3. Table 1(a) and (b) show the catenary
composition for which we performed simulation. We chose two combinations of
wires in order that current capacity and voltage drop between substations below
the specified value. A PH590mm
2
wire is used for the messenger and a
PH150mm
2
for the auxiliary wire to improve the quality of current collection in a
combination and a PH356mm
2
wire, which is generally used in Japan, for the
messenger wire and the auxiliary wire for to make construction easier in other
combination. Moreover, the simulation was performed under different
conditions for comparison.
Table 1 (a): Simulation condition for catenary composition.
PH590mm
2
Composite type
Feeder messenger PH356mm
2
×2 PH730mm
2
×1 PH590mm
2
Tension of feeder messenger 39.2kN 39.2kN 39.2kN
Auxiliary wire PH150mm
2
Tension of auxiliary wire 4.9kN
Contact wire GTMSn170mm
2
GTMSn170mm
2
GTMSn170mm
2
Tension of contact wire 14.7kN 14.7kN 9.8kN
System height 850mm 850mm 850mm
Interval of dropper
Interval of hanger 5m 5m 5m
Composition PH356mm
2
×2 PH730mm
2
Power Supply System Analysis, Design and Planning 103
Table 1 (b): Simulation condition for catenary composition.
PH356mm
2
PH590mm
2
PH356mm
2
Composite type Compound type Compound type
Feeder messenger PH356mm
2
PH590mm
2
PH356mm
2
Tension of feeder messenger 39.2kN 39.2kN 39.2kN
Auxiliary wire PH356mm
2
PH150mm
2
PH356mm
2
Tension of auxiliary wire 4.9kN 4.9kN 4.9kN
Contact wire GTMSn170mm
2
GTMSn170mm
2
GTMSn170mm
2
Tension of contact wire 9.8kN 9.8kN 9.8kN
System height 850mm 910mm 910mm
Interval of dropper 10m 10m
Interval of hanger 5m 5m 5m
Compositiion
3.2 Simulation results
(a) Contact loss rate
Figure 4 shows the simulation results of contact loss rate in each catenary
composition. The catenary composition whose contact loss rate exceeds the
allowable value of 5% in DC electrified railways is only the “composite type”
composed of a PH356mm
2
messenger under the condition where the maximum
speed is 160km/h on narrowgauge lines in Japan.
(b) Contact wire uplift at support
Figure 5 shows the simulation results of contact wire uplift at support in each
catenary composition. According to the simulation results, the contact wire
uplift at support of the compound type is larger than that of other catenary
system. However, the values are less than allowable value of 70mm up to
160km/h.
O
16
3O
46
7O 12O 17O 22O
6peed ´km/H)
C
o
n
L
a
c
L
l
o
c
c
r
a
L
e

ª
´²) 'T)
'Æ) 'N)
'N) 'V)
Figure 4: Simulation results of contact loss rate.
104 Power Supply, Energy Management and Catenary Problems
O
4O
8O
12O
7O 12O 17O 22O
6peed ´km/H)
U
p
l
l
f
L

m
m

´²) ´T)
´Æ) 'N)
´N) 'V)
Figure 5: Simulation results of contact wire uplift at support.
O
4OO
8OO
12OO
7O 12O 17O 22O
6peed km/H
6
L
r
a
l
n

·
1
O

6

´²) ´T)
´Æ) ´N)
´N) ´V)
Figure 6: Simulation results of contact wire strain at support.
Contact wire strain at support
Figure 6 shows the contact wire strain at support in each catenary composition.
The value of contact wire strain is less than the allowable value of a 1000×10
6
in
all catenary compositions up to 160km/h. From these simulation results, these
catenary compositions, except the one that uses PH356mm
2
wire for the
messenger wire, satisfy the allowable value up to 160km/h to prove that the
possibility for practical use is high.
4 Current collecting performance tests
4.1 Outline of test
We installed a feeder messenger catenary that provided satisfactory simulation
results on the current collection test equipment in our institute and performed
current collecting performance tests. Table 2 shows the catenary composition
and Table 3 shows the property of pantographs used in this tests. Table 4 shows
Power Supply System Analysis, Design and Planning 105
Feeder messenger PH590mm
2
×1
Auxiliar y wir e PH150mm
2
×1
Cont act wir e GTM170mm
2
×1
Span lengt h 50m
Hanger int er val 5m
Dr opper int er val 5m or 10m
Syst em height 960mm
the tension distribution of each wire. We ran pantograph at 80 to 150km/h at
intervals of 10km/h to perform current collecting performance tests.
Table 2: Catenary compositions.
Table 3: Property of the pantograph used in experiments.
Table 4: Tension distribution of each wire.
4.2 Test results of the “composite type”
4.2.1 Contact loss rate and uplift under different conditions
Figure 7 shows the contact loss rate measured in current collecting tests under
different conditions in Table 4. In the case of the pantograph PS21, the contact
loss rate was 1% or less up to 150km/h (Figure 7 (a)) to prove the satisfactory
quality of current collection. Contact breaks occurred at 140km/h or over under
the condition (1), but not under the condition (2) or (3).
When the tension distribution of contact wire increases, therefore the
contact loss rate decreases. This phenomenon appeared notably in the case
Pant ogr aph t ype PS21 PS32
For m Lozenge Single ar m
m
1
[kg] 9.3 9.8
k
1
[N/m] 35970 10740
m
2
[kg] 3.8 4.2
k
2
[N/m] 18720 21520
m
3
[kg] 9.4 8.7
c [Ns/m]  10
P
0
[N] 58.8 58.8
Aer odynamic upwar d for ce [N/(km/h)
2
] 2.0×10
3
4.7×10
4
m
1
m
2
m
3
P
0
c
k
1
k
2
Feeder Auxiliar y Cont act
messenger wir e wir e yoke
(1) 39.2kN 4.9kN 9.8kN 2:1
(2) 34.3kN 6.6kN 13.0kN 2:1
(3) 34.3kN 4.9kN 14.7kN 3:1
Rat io of
Tension
Condit ion
106 Power Supply, Energy Management and Catenary Problems
where PS32 was used (Figure 7 (b)). There were no problems in the contact wire
strain and uplift, since they were much less than the allowable values up to
150km/h.
O.O
2.6
6.O
7O 1OO 1JO 16O
6peed jkm/H
CoodlLloo,1)
CoodlLloo,2)
CoodlLloo,J)
C
o
o
L
a
c
L
l
o
c
c
r
a
L
e
j
ª

O.O
2.6
6.O
7O 1OO 1JO 16O
6peed jkm/H
C
o
o
L
a
c
L
l
o
c
c
r
a
L
e
j
ª

CoodlLloo,1)
CoodlLloo,J)
Figure 7: (a) Experiment results of contact loss rate (PS21) (b) Experiment
results of contact loss rate (PS32).
4.2.2 Contact loss rate when spring hangers were used
The current collection performance of pantograph was measured when one or
two hangers were used in place of spring hangers from the support in order to
make the spring constants of the catenary near the support smaller. Figure 8
shows the contact loss rate when the pantograph PS32 was used. Figure 8 shows
that the contact loss rate was almost the same whether normal hangers or spring
hangers were used up to 130km/h.
O.O
2.6
O
7O 1OO 13O 16O
6peed km/H
C
o
n
L
a
c
L
l
o
c
c
r
a
L
e

ª
Normal Hanger
6prlng Hanger
6.
Figure 8: Experiment results of contact loss rate (PS32).
At 140km/h or over, however, the contact loss rate decreased significantly
when spring hangers were used to show that the current collecting performance
improved.
The contact wire strain was smaller than that in the case where normal hangers
were used. There were no problems in the contact wire uplift at support, since it
was much smaller than the allowable value though it had became larger.
Power Supply System Analysis, Design and Planning 107
4.3 Test results of the “compound type”
4.3.1 Comparison of contact loss rate by suspended type
We constructed two structures of the “dropper form” whose dropper intervals
were 5m and 10m in our institute and performed current collecting tests for the
structures of Figure 9. Figure 10 (a) and (b) shows the results of current
collecting performance tests of the “hanger form” and the “dropper form”. The
contact loss rate was small with the “dropper form” of 5m dropper intervals in
both cases of PS21 and PS32. The contact loss rate of PS21 was 1% or less up to
150km/h, and the current collecting performance was extremely good. In the
case of PS32, the contact loss rate was 1% or less up to 130km/h, and nearly 2%
from 140km/h to 150km/h, and the results were allowable value.
Figure 9: “Dropper form compound type” feeder messenger catenary.
4.3.2 Comparison of contact loss rate by tension distribution
We performed current collecting tests on the “dropper form” whose dropper
interval was 5m to ensure the best current collecting performance under the
condition of wire tension distributions shown in Table 4. Figure 11 and 12 show
the measured contact loss rate and contact wire strain respectively, when the
pantograph PS21 was used. The contact loss rate was small when the tension
distribution of contact wire was large at the speed up to 130km/h (Figure 11).
The contact wire strain also became small when the tension distribution of the
contact wire was large (Figure 12).
At the speed of 150km/h, the contact wire strain to exceeded the allowable
value of 500×10
6
in the case of (1) contact wire tension 9.8kN, but was less than
the allowable value in the case of (3) contact wire tension 14.7kN. As mentioned
above, the current collection performance was also good to suit high speed
(b) Dropper interval with 5m
5m
50mm
5m
50mm
10m
Auxiliar y wir e
Feeder messenger
Dr opper
Pulloff ar m
50mm
Ear
Cont act wir e
10m
Auxiliar y wir e
Feeder messenger
Dr opper
Pulloff ar m
50mm
Ear
Cont act wir e
(a) Dropper interval with 10m
108 Power Supply, Energy Management and Catenary Problems
operation with the “compound form” when the tension distribution of contact
wire was increased.
O.O
6.O
7O 1OO 13O 16O
6peed km/H
C
o
n
L
a
c
L
l
o
c
c
r
a
L
e

ª

Hanger form
dropper form ´lnLerval 1Om)
dropper form ´lnLerval 6m)
O.O
6.O
7O 1OO 13O 16O
6peed km/H
C
o
n
L
a
c
L
l
o
c
c
r
a
L
e

ª

Hanger form
dropper form ´1Om)
dropper form ´ 6m)
1O.O 1O.O
Figure 10: (a) Test results of contact loss rate (PS21) (b) Test results of contact
loss rate (PS32).
6
O.O
2.6
7O 1OO 13O 16O
6peed km/H
CondlLlon ´1)
CondlLlon ´3)
6.O
C
o
n
L
a
c
L
l
o
c
c
r
a
L
e

ª

Figure 11: Test results of contact loss rate (different tension).
76O
26O
26O
76O
6
L
r
a
l
n

1
·
1
O

6

7O 1OO 13O 16O
6peed km/H
CondlLlon ´1)
CondlLlon ´3)
Figure 12: Test results of the contact wire strain.
5 Tension shift test
For the “compound type” that was constructed in our institute, we ground the
contact wire flat and measured changes in the tension auxiliary wire and contact
Power Supply System Analysis, Design and Planning 109
wire. Figure 13 shows the measured tension in the auxiliary wire and the contact
wire versus the residual diameter of the contact wire. From these measurement
results, we confirmed that the tension shifted from the contact wire to the
auxiliary wire with as the wear of the contact wire advanced. Although the limit
of the residual diameter of the contact wire was normally 8.5mm, we set the
residual diameter at 3.3mm. Furthermore we set it at 1.5mm locally in this
experiment. However the contact wire did not break. This proves that this
structure is strong against the local wear of the contact wire.
6
1O
16
2O
3 7 11 16
Pecldual dlameLer mm
6lmulaLlon reculLc
LxperlmenL reculLc
O
6
1O
16
3 6 9 12 16
Pecldual dlameLer mm
1
e
n
c
l
o
n

k
N

6lmulaLlon reculLc
LxperlmenLal recucLc
1
e
n
c
l
o
n

k
N

Figure 13: (a) The auxiliary wire tension (b) The contact wire tension.
6 Summary of the tests
We performed simulation, experiments of current collecting performance and
tension shift tests of the feeder messenger catenary with the auxiliary wire which
was composed of a small number of parts and strong against the wear of the
contact wire. The acquired results are shown below.
(1) Simulation results show that the quality of current collection satisfied
allowable value up to 160km/h with the “compound form” whose messenger
was PH590mm
2
, and the “compound form” whose messenger were
PH590mm
2
and PH356mm.
2
(2) From current collecting performance tests on the “compound form”, we
found that the current collecting performance improved when the total
tension in the contact wire and the auxiliary wire was increased (19.6kN).
(3) From the current collecting performance tests on the “compound form,” it
was found that the best current collecting performance was on the “dropper
form” whose dropper intervals was 5m or over, and we found that the
current collection performance improved when the total tension in the
contact wire and auxiliary wire was increased (19.6kN).
(4) In the measurements on the “compound form” when contact wire wear
advanced, we observed that the tension shifted from the contact wire to the
auxiliary wire as contact wire wear advanced.
110 Power Supply, Energy Management and Catenary Problems
7 Conclusion
Based on this research, we confirmed that there are no problems in the current
collecting performance up to the speed of 150km/h with the “composite form”
and the “compound form” whose dropper intervals are 5m, on the structure of
feeder messenger catenary that uses PH590mm
2
as the messenger. We also
observed the tension shifts to the auxiliary wire when the wear of the contact
wire advanced. We found that this structure was strong against the local wear of
the contact wire. These catenary systems are more stable than others to require
less maintenance.
References
[1] A. Iwainaka, A. Suzuki & Y. Shimodaira, Development of single copper
feeder messenger wire for overhead contact lines, Computers in Railways
VII, pp.703712, 2000.
[2] T. Hamada, A. Suzuki ＆T. Shimada, Current collecting characteristic of
catenary with nontension contact wires, Computers in Railways VIII,
pp.419428, 2002.
Power Supply System Analysis, Design and Planning 111
This page intentionally left blank
Catenary and autotransformer coupled
optimization for 2x25kV systems planning
E. Pilo, L. Rouco & A. Fern ndez
Instituto de Investigación Tecnológica,
E.T.S. de Ingeniería ICAI, Univ. P. Comillas, Madrid, Spain
Abstract
In the design process of a 2x25kV power supply, the location and sizing of
autotransformers and the choice of the catenary type are usually two aspects very
closely interrelated. As a result, the number of autotransformers can normally be
reduced or augmented if the catenary is upgraded or downgraded respectively. In
this paper, coupling equations between these two aspects are described in detail.
These equations are used to formulate a multiattribute optimization problem
in which the global efficiency of the investment is optimized. In this problem,
the considered attributes are: (i) investment cost and (ii) a performance index that
is defined in the paper. The optimization procedure gives a reduced set of
catenary and autotransformers combinations that maximize efficiency. This
optimal reduced set can be used as an input in the design process of power
supply. For the solution of this problem, different dominancy criteria are
evaluated. Furthermore, sensitivity studies are carried out during the
optimization process to help in the search of catenary and autotransformers
combinations.
As a study case, the proposed optimization procedure has been used to obtain
a reduced set of catenaries and autotransformers based in the catenary C350,
which has been used in the Madrid – Barcelona – French border new highspeed
line.
Keywords: power supply system, high speed railways, multiattribute
optimization, longterm infrastructure planning.
1 Introduction
In the design process of a 2x25kV power supply, the location and sizing of
autotransformers and the choice of the catenary type are usually two aspects very
á
Power Supply System Analysis, Design and Planning 113
closely interrelated. As a result, the number of autotransformers can normally be
reduced or augmented if the catenary is upgraded or downgraded respectively. In
this paper these relationships are analyzed and used to formulate a multiattribute
optimization problem in which the global efficiency of the investment is
optimized [1].
Section 2 describes the power supply system of ACelectrified railways. In
section 3 an equivalent model is used to represent bivoltage 2x25kV and to
analyze the relationships between catenary parameters and autotransformers
separation. In section 4 the optimization problem is described. In section 5 the
optimization procedure is applied to a real case. Finally, in section 6 the
conclusions of this work are presented.
2 The power supply of AC electrified railways
2.1 General structure
Figure 1 shows the general structure of powersupply systems of AC electrified
railways:
Sector 3R Sector 1R Sector 2L Sector 2R Sector 3L Sector 1L
Threephase highvoltage network
Traction
substation 1
Traction
substation 2
Traction
substation 3
Figure 1: Structure of the power supply system.
As shown, the electrical system is divided in electricallyisolated single
phased sectors, which are fed from the threephase network through a traction
substation. These substations are connected between two of the three phases of
the highvoltage network. Each of these sectors can use either monovoltage
system (1x25kV) or bivoltage system (2x25kV). In monovoltage systems [2],
the feeding conductors are set to the specified voltage level (see Figure 2).
V
transp V
feed POS
Figure 2: Monovoltage system configuration.
In bivoltage systems, a higher voltage is set between feeding conductors
[3, 4]. This voltage is reduced by using autotransformers distributed along the
114 Power Supply, Energy Management and Catenary Problems
catenary (see Figure 3). In these systems, the term cell very often refers to the
portion of catenary located between two consecutive autotransformers. Typical
values for cell lengths are 1015km.
V
transp
V
feed POS
V
feed NEG
Figure 3: Bivoltage system configuration.
As this paper is focused on this system, it is assumed that all the sectors are fed
using bivoltage system.
2.2 Catenary
The typical configuration of the catenary of an AC railway line is shown in
Figure 4. The catenary contains several physical conductors that can be grouped
into three groups: positive, negative and ground wires. In case of multiple tracks,
other conductor arrangements are possible.
Longitudinal section Transversal section
Negative
Feeder
Return
wire
Rail
Contact
wire
Sustainer
Positive
feeder
Positive
Negative
Neutral
Figure 4: Typical conductor distribution.
The positive wires are the positive feeder, the sustainer wire and the contact
wire. There is usually only one negative wire called negative feeder. The ground
wires are the rail, the collector wire and the return wire.
The conductors of each group are connected between them at regular intervals
(typically 300m). Additionally, ground conductors are frequently connected to
earth.
Power Supply System Analysis, Design and Planning 115
3 The effect of autotransformers and catenary upgrades
3.1 Base magnitudes
In order to improve numerical stability, normally per unit magnitudes are used to
carry out all the necessary calculations. Thus, the circuit can be divided into
three zones based on their nominal voltage. Figure 5 shows the considered zones:
(i) highvoltage zone, (ii) positive zone and (iii) negative zone.
Vthevenin
High voltage
Zthevenin Ztr1
K1:1
K2:1
Ztr2
Ztr3
Zcat Zcat
1:1 1:1
Zcat
Positive zone
Negative zone
Train 1
Figure 5: Zone division for base magnitudes selection.
A base power
base
S has to be chosen and is common to all the zones (a typical
value is 10MW). 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.
3.2 The 1x25kV equivalent model of 2x25kV systems
In [5] the behavior of bivoltage systems is analyzed and a equivalent model is
proposed to represent bivoltage 2x25kV systems as if they were monovoltage
1x25kV. This is the model used in the presented work.
Figure 6 shows the approximated behavior of the circuit with a train
consuming a current I , assuming: (i) that voltage drop along a cell in the positive
and in the negative side have the same value but different sign and (ii) that, as far
as autotransformers can be supposed ideal, it can be assumed that there are
current flows only in the autotransformers that are immediately adjacent to the
considered train.
In this figure
cell n
V is the voltage drop along the cell n,
p,trans
I and
n,trans
I are
respectively the positive and negative currents in the transmission cells,
p,train
I
and
n,train
I are respectively the positive and negative currents in the cell of the
train,
cell
L is the length of the cell of the train, x is the relative position of the
train, expressed as a fraction of
cell
L .
116 Power Supply, Energy Management and Catenary Problems
I
I0
In,trans
Ip,trans
VHV
»0 »0 »0 »0
AT1 AT2
»0 »0 »0 »0
Transmission
cell
Transmission
cell
Transmission
cell
Cell of the train Downwards cell Downwards cell
Vcell1
Vcell1
Vcell2
Vcell2
Vcell3
Vcell3
Vcell4
Vcell4
Vcell5
Vcell5
Vcell6
Vcell6
Ip,train
In,train
In,train
xLcell
Lcell
Figure 6: Approximated behavior of bivoltage system.
Based on these simplifications, the positive phase of the bivoltage system
can be represented as shown in Figure 7. In this model, two different
contributions have been identified: (i) the equivalent impedance of the catenary
, eqv cat
z that depends only on the configuration of physical conductors and (ii)
gap
z that is associated to the separation between autotransformers.
Vthevenin
High Voltage Network Transformer
Ground
Positive
Zeqv,SS
Catenary
Zeqv,cat
Zgap
Train
Figure 7: Monovoltage equivalent circuit of bivoltage system.
The parameters of this equivalent circuit are calculated as follows:
, , , eqv cat eqv cat ss train
D = ⋅ z z (1)
where the symbol ~ is used to refer per length unit magnitudes and
, ss train
D is the
distance between the substation and the train
The equivalent impedance
, eqv cat
z of the catenary can be obtained from the
elements of the equivalent conductors impedance matrix [6], where the sub
indexes
p
and
n
represents the positive and negative conductors respectively.
,
pp nn pn np
eqv cat
pp nn pn np
⋅ − ⋅
=
+ + +
z z z z
z
z z z z
(2)
The impedance gap
gap
z associated to the separation of autotransformers can be
obtained as follows:
Power Supply System Analysis, Design and Planning 117
( )
2
1
pp pp np pn pp pn np
gap cell
pp pn np nn
L x x
+ + +
= −
+ + +
z z z z z z z
z
z z z z
(3)
The voltage drop associated to the impedance
gap
z is referred as voltage
deviation from the equivalent model of the catenary. As shown in Figure 8, this
voltage deviation starts and ends in cell in which is located the train. In other
words, at the end of this cell all the voltage deviation is recovered and thus no
extra voltage drop has to be added to the trains that are located downwards.
Figure 8 summarizes the voltage drops in the sector in a scenario with only
one train:
Traction
Substation
A.T. 1 A.T. 2
(km)
Train
(V)
X
AT1
dV/dx=Z
eq,CAT
·I
train
V
SS
X
AT2
X
train
V
CAT
1
2
3
V
AT1
V
AT2 1
2
V V
AT1 train
3 Z ·i
gap train
Figure 8: Voltage drops in a bivoltage sector.
It can be seen that the deviation impedance is proportional to the distance
between autotransformers. Consequently, as far as the number of
autotransformers is increased, the relative weight of the deviation is reduced.
3.3 Influence of catenary type and autotransformers distance on voltages
Using the described model, sensitivity of the voltage drops to catenary upgrades
or to autotransformer additions can be analyzed. These are two common
investment decisions to be taken in the design process of the power supply in
bivoltage 2x25kV electrified railways.
Figure 9 shows the effect of upgrading the catenary in the voltage drops along
the catenary
cat
V ∆ . When upgrading the catenary, the most important effect is a
reduction of voltage drops all along the sector. Additionally, the voltage drop
due the separation autotransformers (corresponding to
gap
z ) can also be reduced.
Figure 10 shows the effect of shortening the distance between consecutive
autotransformers, which is typically achieved by adding extra autotransformers
to the sector. Unlike the catenary upgrade, the benefits of this enhancement are
limited to the cell whose distance has been reduced, due to the local influence of
voltage deviations VDESV .
118 Power Supply, Energy Management and Catenary Problems
Train 2 Train 1 SS
DVSE
DVcat
X (in m)
V (in kV)
Train 2 Train 1 SS
DVSE
DVcat
X (in m)
V (in kV)
Figure 9: Effect of upgrading the catenary.
SS Train
AT1 AT2
AT1 AT2
V
x
VDESV
VDESV
Figure 10:
As it has been described, both catenary upgrades and autotransformer
additions can be used to reduce voltage drops.
4 The optimization problem
In the design process of ACelectrified railways, voltage limitations are
commonly active, especially when evaluated in degraded situations (substation
out of order). Thus, determining the most efficient way of reducing voltage drops
is a key factor in the design of the power supply.
As it has been described, catenary upgrades and autotransformer additions are
often investment decisions that are exchangeable in order to reduce voltage
drops, from a technical point of view. Therefore, economical criteria have to be
considered to determine the most efficient combination of catenaries and
autotransformer distributions. For that reason, a multicriteria optimization
problem has been formulated and solved.
The goal of this optimization is to obtain a reduced repository of
combinations of catenary types and autotransformer distributions, in which the
efficiency of its elements is maximized. This set is to be used as an input in the
Effect of shortening length of cells.
Power Supply System Analysis, Design and Planning 119
design of the power supply system. To determine this repository, the considered
attributes are: (i) investment cost and (ii) a efficiency index that is to be defined.
In order to evaluate the efficiency of each {catenary,autotransformers}
combination the efficiency index ZEQ is defined in eqn. (4). This index
corresponds to the total equivalent impedance seen between the substation output
and a train located in the further cell of the corresponding sector. In order to
simplify the resulting expression, it has been assumed that autotransformers are
uniformly distributed.
,
1
eq cat gap
n L
L x
n n
−
= + +
ZEQ z z (4)
where n is the number of autotransformers of this sector and L is the length of
this sector. Expanding
gap
z in eqn. (4) becomes:
( ) ( )
, ,
1
1
eq cat pp eq cat
n L L
L x x x
n n n
−
= + + − −
ZEQ z z z (5)
As ZEQ depends on the relative position x of the train in its cell, one of the
following criteria for can be assumed:
a. The train is located in the middle of its cell ( 0.5 x = )
b. The train is located where the index ZEQ reaches its maximum.
It should be noted that ZEQ incorporates the effect of both catenary type and
autotransformers in an equilibrated manner, as it corresponds to their
contribution in the total voltage drops.
The considered cost function is:
, cat n cat AT
C C L C n = ⋅ + ⋅ (6)
where
, cat n
C is the cost associated to catenary cat with n autotransformers,
cat
C is the per length unit cost of the catenary cat and
AT
C is the unitary cost
of each autotransformer.
To get the optimal repository the following steps have to be accomplished:
a) Exploration of the possible catenary/autotransformer configurations. As
the sustainers and the contact wires are fixed, the number of
combinations is not excessive. Symmetry restrictions can also be used
to reduce the search space.
b) Eliminate dominated configurations (a configuration is dominated if it
is worse in cost and in efficiency than another configuration). Relaxed
dominancy criteria can also be applied in order the eliminate
configurations (i) a bit cheaper but much lower efficiency or (ii) a bit
more efficient but much more expensive.
c) Within the resulting repository, chose the final configurations trying to
cover uniformly the costs range.
120 Power Supply, Energy Management and Catenary Problems
5 Study case
The proposed optimization procedure has been used to get a repository of
catenaries and autotransformers based on the catenary C350 that has been used
in the Madrid – Barcelona – French border new highspeed line. The considered
costs are summarized in Table 1.
Table 1: Costs structure of the study case.
Description Cost Observations
Autotransformer 200 units Cost per autotransformer
Fixed cost of catenary 29 units/km Includes installation of
catenary towers, insulators,
contact wires and sustainers
Cost per added feeder 0.3 units/km/feeder Only in configurations with
positive or negative feeder
LA110 conductor 0.716 units/km To be included only if used
LA180 conductor 0.832 units/km To be included only if used
LA280 conductor 0.95 units/km To be included only if used
LA380 conductor 0.125 units/km To be included only if used
Figure 11 shows the 5 catenary/autotransformers combinations chosen
(marked with arrows), that have been evaluated for sector 30km, 40km and
50km long.
Figure 11: Effect of shortening length of cells.
Power Supply System Analysis, Design and Planning 121
Each curve represents the effect of adding up to 5 autotransformers to the
catenary specified, for the given sector length.
As shown in Figure 11, the efficiency gain of upgrading catenary (change to a
lower curve of the same family) is normally a more efficient option than adding
an extra autotransformer (go to the following point of the same curve).
It can also be observed that the efficiency of adding autotransformers
decreases very quickly and, thus, none of the chosen configurations uses more
than 2 autotransformers.
6 Conclusions
Using the monovoltage equivalent model of bivoltage systems, the existing
relationships between the location of autotransformers and the choice of the
catenary have been analyzed. Furthermore, a procedure has been proposed to
obtain a reduced repository of combinations of catenary types and number of
autotransformers in which efficiency and cost are optimized
As a study case, the proposed optimization procedure has been used to obtain
a reduced set of catenaries and autotransformers based in the catenary C350,
which has been used in the Madrid – Barcelona – French border new highspeed
line. The solution obtained in the study case suggests that upgrading catenary is
normally a much more efficient option than adding an extra autotransformer. The
reason for that is that the marginal cost of using larger conductors in the catenary
can be neglected compared to the marginal cost of adding an autotransformer.
References
[1] E. Pilo, “Optimización del Diseño de la Electrificación de Trenes de Alta
Velocidad,” in IIT, vol. PhD: Univ. Pont. Comillas de Madrid, 2003.
[2] J. D. Glover, A. Kusko, and S. M. Peeran, “Train voltage analysis for AC
railroad electrification,” IEEE Transactions on Industry Applications, vol.
IA20, pp. 925934, 1984.
[3] R. J. Hill and I. H. Cevik, “Online simulation of voltage regulation in
autotransformerfed AC electric railroad traction networks,” IEEE
Transactions on Vehicular Technology, vol. 42, pp. 365372, 1993.
[4] P. H. Hsi, S. L. Chen, and R. J. Li, “Simulating online dynamic voltages
of multiple trains under real operating conditions for AC railways, ” IEEE
Transactions on Power Systems, vol. 14, pp. 452459, 1999.
[5] E. Pilo, L. Rouco, and A. Fernández, “A reduced representation of 2x25
kV electrical systems for highspeed railways.,” presented at IEEE/ASME
Joint Rail Conference, Chicago, 2003.
[6] E. Pilo, L. Rouco, A. Fernández, and A. HernándezVelilla, “A simulation
tool for the design of the electrical supply system of highspeed railway
lines, ” presented at IEEE PES Summer meeting 2000, Seattle, 2000.
122 Power Supply, Energy Management and Catenary Problems
Investigation into the computational techniques
of power system modelling for a DC railway
A. Finlayson
1
, C. J. Goodman
2
& R. D. White
1
1
Atkins Rail, UK
2
University of Birmingham, UK
Abstract
The use of computer simulation techniques is now a fundamental part of the
design process for electrified railways and at the feasibility stage clients will
often request detailed calculations to be performed for the basic design. This will
establish a level of confidence in both the project and basic design parameters
that will justify the capital expenditure further on in the project lifecycle.
This paper will address how the total impedance of a railway network may be
represented, where the impedances of the traction return circuit and traction
power system are either combined together to form one impedance or studied
independent of one another. The accuracy of the modelling in this manner,
particularly how it affects the accuracy of rail voltage results shall be assessed. It
will also examine how much impedance is typically in the rails and what
proportion this needs to be before it has an unacceptable effect on the numerical
results. To assess this, a range of proportions will be studied, for example 70% of
the total impedance to be modelled in the conductor with 30% modelled in the
rails, 60%/40%, etc. In this way, a proper scientific assessment of the combined
or split calculation methods can be made via simplified case studies.
Keywords: DC railway, modelling, simulation, computational techniques,
accuracy, numerical method, rail voltage, stray current.
1 Introduction
DC electrified railway systems across the world are growing, both in passenger
and freight traffic, as an alternative solution to increasing road congestion. DC
electrified railways account for approximately 50% of electrified railway lines
throughout the world [1]. This growth may be in the form of new build railway
Power Supply System Analysis, Design and Planning 123
systems or the upgrade of existing railway systems and is inclusive of metro,
tramway, light rail and heavy rail systems. However this growth has to be
managed to ensure that the proposals for an improved infrastructure are designed
both safely and correctly and that the design delivers value for money. Computer
simulation is the tool that many consultants and manufacturers, associated with
the rail industry, are using to evaluate proposals and validate final designs prior
to the launch of the project and any major capital expenditure. The use of
computer simulation techniques is now a fundamental part of the design process
for electrified railways and at the tender stage clients will often request detailed
calculations to be performed for the basic design [2]. This will establish a level
of confidence in both the project and basic design parameters that will justify the
capital expenditure further on in the project lifecycle.
1.1 DC railway modelling and simulation
Modelling and simulation are the names we use for applying the laws of physics
and logic via the processing power of computers to predict the behaviour and
performance of railway infrastructure [2]. The various levels of engineering
modelling and simulation with respect to the DC railway may be considered as;
timetable and operational planning, train performance and signalling, traction
equipment and the power supply system. These engineering levels are not
considered to be exhaustive but what is important is the power system supply, as
this paper is primarily concerned with the modelling of the power system for the
DC railway.
2 The DC railway
2.1 Operational functions
The DC railway usually has the primary operational function of passenger
transportation, whereas the AC railway has the dual operational functions of
passenger and freight transportation. However the DC railway may also be used
within the mining industry, as well as freight transportation [3], so it may be
considered to have dual operational functions also. However, for the purpose of
this paper the DC railway shall be considered as predominantly suburban
railways often referred to as transit systems. Transit systems may be classified
based upon the demand for traction current from the power system. For example,
a metropolitan railway (metro) may be considered to be a mass transit system as
there will be more trains carrying more passengers with short headways between
trains and therefore a large demand for traction current from the power system.
Conversely, a tram railway may be considered to be a light rapid transit (LRT)
system as there will be less trams and also less demand for traction current from
the power system than the mass transit system.
Either of these systems may utilise an overhead line or third/fourth rail
feeding arrangement. The following DC traction supplies are commonly
used [4]:
124 Power Supply, Energy Management and Catenary Problems
• 3000 V (overhead line)
• 1500 V (overhead line)
• 750 V (third rail and overhead line)
• 630 V (third and fourth rail)
2.2 Components of the DC railway
The DC electrified railway comprises of many interacting variables or
components. These variables are usually contained within subsystems
such as; civil engineering, permanent way, electrification, signalling,
telecommunications, rolling stock, station services and third parties outside the
railway environment. Whilst this list is typical (nonexhaustive), all of these
variables must be considered both individually and interacting with one another
during the design process of a railway system. However, for the purpose of this
paper the components that are of most concern are;
• Electrification
• Rolling stock
• Signalling
2.3 Other considerations for modelling
In studying the DC railway power system, there are two further technical
problems that need to be included in the circuit model of the DC power system.
Firstly, there is the variation of rail voltage with respect to the train operational
timetable and secondly there is the magnitude of stray current with respect to the
train operational timetable.
33kV AC 33kV AC
11kV AC 11kV AC
750VDC 750VDC 750VDC
R1
R2
T1
T2
Figure 1: The DC railway represented in a single line diagram.
2.4 DC railway operation
Figure 1 shows a typical two track DC railway with one feeder station with a
primary voltage of 33 kV AC and a secondary voltage of 11 kV AC. The
Power Supply System Analysis, Design and Planning 125
11 kV AC is then rectified and an output voltage of 750 V DC is produced which
in turn will power the rolling stock on the rail network. The rolling stock will
draw traction current from T2 and will return traction current through R2.
However not all of the traction return current will return through the running
rails and this may be due to the earthing and track bonding arrangements
associated with the DC railway.
3 Choosing the model and method
3.1 The basic model
There are many methods available to us in solving the power system model of a
DC railway. Examples of these may include multiconductor modelling, finite
element analysis or transmission line theory to name but a few. However, given
the nature of the DC railway, the power system model may be considered to be
dynamic, i.e. the electrical load changes with demand and with multiple
electrical loads, i.e. trains. Whilst these parameters may appear complex, the DC
railway power system is easier to model than the alternative AC railway. This is
because the effects of capacitive and inductive reactance in the equivalent steady
state circuits may be ignored due to the general principles of DC circuit theory
[5]. Applying this to figure 1 will give rise to the equivalent circuit shown in
figure 2, where Z
c1
, Z
c2
and Z
c3
represent the impedance of the traction power
conductor, whilst Z
r1
, Z
r2
and Z
r3
represent the impedance of the traction return
conductor.
Train 1
Z
c1
I
Return Rail
Traction Power
Substation 1
Z
r1
Z
t
Train 2
Z
c2
I
Z
r2
Z
t
Substation 2
Train n
I
Z
t
Z
c3
Z
r3
Figure 2: Electrical circuit for DC railway.
Figure 2 may still not be considered to be the simplest model of the DC
railway. The electrical substations may be modelled in their Thévinin equivalent
model, a voltage source (V
s
) with a source impedance (Z
s
) and the train may be
modelled as an impedance (Z
t
) that is dependent upon whether the train is
motoring, coasting or braking.
Figure 3 now shows two cases of the simplified circuit, with the Thévénin
equivalent voltage source representing two electrical substations and a single
train represented by an impedance somewhere between the two electrical
substations. The circuit shown to the left of figure 3 represents the DC railway
where the traction power and traction return conductors are split from one
126 Power Supply, Energy Management and Catenary Problems
another, whilst the circuit to the right shows them combined. The circuit to the
right suggests that the DC railway power system model may be solved as a
simple DC circuit, using short transmission line theory [6].
Figure 3: Simplified electrical circuit of a DC railway.
Return Rails
Traction Power
Conductor
Z
r1u
Z
r1d
Z
c1
I
t1
50% I
t1
50% I
t1
Z
r1u
Z
r2d
Z
c1
I
t1
50% I
t1
50% I
t1
Z
c2
I
t2
50% I
t2
50% I
t2
Return Rails
Traction Power
Conductor
Figure 4: Developing the combined impedance model.
3.2 Representing the traction return conductor in the combined impedance
model
Figure 3 is representative of a one track railway. In practice, the DC railway,
under consideration for this paper, will usually be a minimum of two tracks,
namely up (u) and down (d). Considering figure 4, there are two trains present,
one on each track and there are four traction return conductors, or as is more
commonly known return paths. These return paths are considered to have
identical impedance characteristics and the percentage of traction return current
(It
1
and It
2
) in each return path is shown.
Considering the configuration shown to the left of figure 4, will give rise to
the equivalent impedance of all the return paths combined being expressed as:
Ω =
n
Zr
Zr
u
n
1
(1)
where ‘n’ is equal to the total number of traction return conductors in the circuit.
However, considering the configuration to the right of figure 4, the equivalent
impedance of all the return paths combined will be expressed as:
Ω =
n
Zr T
Zr
u
n
1
*
(2)
Power Supply System Analysis, Design and Planning 127
where T is the number of trains present. However, because the number of trains
present between electrical substations and on different tracks varies, this will be
difficult to program in a computer simulation tool. The equivalent impedance of
all the return paths combined may be expressed as an alternative to equation 1 by
Ω =
u n
Zr Zr
1
(3)
This paper will assess which the accuracy of modelling the equivalent return rail
impedance as either equations 1 or 3 for a number of tracks, as these are feasible
methods that may be used by computer simulation tools.
3.3 The method
Figure 3 may be solved via a number of well known network analysis techniques
such as mesh or nodal analysis [5]. For simple circuits such as that shown in
figure 3 where there is only 1 train present, the solution may be solved by
deriving linear equations and solving them as simultaneous equations. If the
circuit contains more electrical loads, it is better to solve the circuit via a series
of linear equations represented in matrices where the general equation may be
given as
[ ] [ ] [ ] I Z V * = (4)
The computational efficiency of the numerical procedures undertaken by both
mesh and nodal analysis may be measured by counting the number of numerical
operations (OC) i.e. multiplication, division, addition and subtraction for each
equation ‘N’ to be solved, i.e. ‘N’ is equal to the number of nodes in the circuit
or the number of loop currents in the circuit, dependent upon the network
analysis technique chosen. This may be defined as [10];
6
7
2
3
3
2
2 3
N N N
OC − + = (5)
Table 1 shows the difference in operational counts taking into account the
simplified models shown in figure 3. Table 1 illustrates that nodal analysis has a
much lower OC when 4 or more linear equations are required to be solved. This
can be attributed to the fact that in nodal analysis, each node represents an
equation and one node has to be set to zero.
Table 1: Summary of OC.
Number of Traction Return Paths Mesh Analysis Nodal Analysis
N OC N OC
1 2 9 2 9
2 4 62 3 28
4 8 428 5 115
6 12 1354 7 294
8 16 3096 9 527
128 Power Supply, Energy Management and Catenary Problems
If using mesh analysis, each loop current represents an equation and each and
every loop current must be used. This means that for an equivalent circuit that
requires 4 or more equations to be solved, mesh analysis will always have one
more equation to be solved than nodal analysis, thus meaning that nodal analysis
is the more efficient method.
3.4 Injecting train currents
If the combined impedance circuit of figure 3 is used to solve the power system
then it presents us with a problem as the technical issues cited in section 2.3 are
not represented in the model, whereas the rail voltage is represented and
calculated directly in the split impedance model. This problem may be overcome
by either including these parameters in the model for real time simulation or they
can be accounted for in a post processing program, i.e. a separate program that
operates on the results from the initial simulation [7]. This may be done by
injecting the train currents (I
t
) into a homogenous rail. The ground impedance
(Z
e
) is represented as a lumped parameter that ignores the effect of ground
capacitance, because any reactive component represented by the ground will be
so small that it can be ignored [4, 8, and 9] and in any case is only relevant for
transient conditions, as is shown in figure 5. The current flowing in the earth (I
e
)
represents the stray current and this may also be studied as part of the post
processing program.
Figure 5: Post process injecting train currents model.
4 Case studies
The theoretical analysis of railway modelling techniques discussed in section 3
may be applied to the practical railway, where numerical analysis can be applied
to ascertain the accuracy of modelling between the split impedance method (with
the traction return path impedance modelled as per equation 2) and the combined
impedance method with the equivalent traction return path impedance modelled
as per equations 1 and 3. The network analysis technique chosen was nodal
analysis as this calculates the train voltage directly and is considered to be more
efficient than mesh analysis, as defined in section 3.4. The following railway
types have been identified as case studies;
• 1500 V DC Overhead Line Railway (OHL)
• 750 V DC 3
rd
Rail Conductor Railway
• 750 V DC Overhead Line Railway
Power Supply System Analysis, Design and Planning 129
Practical traction return circuits, traction power system and voltage source
impedances have been used in the case studies. A range of proportions have
been studied, for example 60% of the total impedance to be modelled in the
traction conductor (α) with 40% modelled in the rails etc., where the summation
of the traction powers system and return circuit impedances is equivalent to the
loop impedance Z
L
. Each case study includes six traction return circuit with up
to six trains present, i.e. one on each track.
4.1 Calculating the error
The train voltage may be expressed as a percentage of the supply voltage derived
by the electrical substation. This may be expressed generally as;
% 100
1
× =
Vs
V
Vt
train
(6)
where V
train
is the train voltage and Vs
1
is the electrical substation voltage.
Equation 6 will satisfy both the split and combined impedance analyses. The
error in train voltage calculation may be expressed as;
% %
combined split
Vt Vt error − = (7)
where Vt
split
is the result of equation 6 for the split impedance model and
Vt
combined
is the result for the combined impedance method. A back calculation
may be performed using the result of equation 7 to calculate the error in terms of
voltage magnitude if required, as shown in equation 8.
V Vs
error
V
1
100
%
× = (8)
4.2 Summary of results
Table 2 shows the percentage error (%) in train voltage calculation, using
equation 7.
5 Discussion and conclusion
5.1 Discussion
The results in table 2 suggest that either equations 1 and 3 for the equivalent
traction return path impedance are satisfied when there is only one train present,
as in the case of equation 1 and six trains being present, as in the case of
equation 3. This does not definitively define which equivalent rail return
impedance should be used. Therefore an intermediate step was introduced by
assessing three trains being present. The choice for three trains is purely
130 Power Supply, Energy Management and Catenary Problems
Eqn. (3)
% error
0
0
0
0
Error for 6 Trains and
six return paths
Eqn. (1)
% error
1.668
2.705
17.086
17.112
Eqn. (3)
% error
1.001
1.623
10.252
10.267
Error for 3 Trains and
six return paths
Eqn. (1)
% error
0.667
1.082
6.834
6.845
Eqn. (3)
% error
1.668
2.705
17.086
17.112
Error for 1 Train and
six return paths
Eqn. (1)
% error
0
0
0
0
Substation
Spacing
km
2
1.5
3
3
Train
Current
A
4000
1000
5000
5000
α%
60
55
37
50
Z
L
Ω/m
0.0784
0.1472
0.0651
0.0828
Type
OHL
OHL
3rd
rail
3rd
rail
System
Voltage
V DC
1500
750
750
750
Table 2: Summary of case study results.
P
o
w
e
r
S
u
p
p
l
y
S
y
s
t
e
m
A
n
a
l
y
s
i
s
,
D
e
s
i
g
n
a
n
d
P
l
a
n
n
i
n
g
1
3
1
academic and is used to define a utilisation factor of the traction return circuits,
although in reality the number of trains may vary. This now shows that the error
is lower when using equation 1 to derive the equivalent traction return
impedance. However, whilst the error is acceptable for the overhead line
railways, it is not so for the 3
rd
rail railway. The impedance is too low in the
traction power conductor, allowing a large voltage drop to occur between the
train and the electrical substation, mainly due to the high values of current drawn
by the trains. Furthermore, the impedance in the traction return circuit is more
dominant and the inconsistency in the assumptions of equations 1 and 3 suggest
that this impedance must be modelled accurately, i.e. as per equation 2 and the
split impedance method.
Therefore this type of railway should be modelled as a split impedance
railway, such that the train and rail voltages are calculated correctly.
Furthermore, it is noticeable that the magnitude of percentage error increases as
the number of tracks occupied increases. This must also be considered when
deriving the power system model, with respect to split or combined impedance
method.
5.2 Conclusion
The case studies confirm that the most efficient method for solving the power
system model is to combine the traction power and return conductor impedances
such that the OC of the circuit is reduced to as low as reasonably possible. This
will allow post process simulation of the data obtained from the real time
solution and allow effective management of the computing overhead that is
always the fundamental driver with computational techniques and the desired
accuracy. This statement is true, with the exception of the 3
rd
rail traction
conductor railway, where the most appropriate method would be to model the
impedances of the railway separately.
Acknowledgement and disclaimer
Information and guidance given in this paper are views held by the authors. The
authors would like to thank colleagues within Atkins Rail and the University of
Birmingham for their support and advice in writing this paper and for their kind
permission in allowing this work to be published as part of the collaborative
agreement to develop a DC multi train simulator.
The authors, Atkins Rail and the University of Birmingham accept no liability
to anyone for any loss or damage caused by any error or omission in the work,
whether such error or omission is the result of negligence or any other cause.
References
[1] Profillidis, VA. (2000). Railway Engineering. 2
nd
Edition. Aldershot:
Ashgate Publishing Ltd. Chapter 13 pp252. ISBN 0 – 7546 – 1279 – 1.
132 Power Supply, Energy Management and Catenary Problems
[2] Goodman, CJ, (2005). Modelling and Simulation. The IEE Second
Residential Course on REIS 2005. Wrightsons. Chapter D3. ISBN 0 –
86341 – 511 – 3.
[3] BS EN501221:1998 Railway applications  Fixed installations Part 1:
Protective provisions relating to electrical safety and earthing.
[4] Kiessling, F, Puschmann, R, Scmeider, A. (2001). Contact Lines for
Electric Railways. 1
st
Edition. Munich Erlangen: Publicis Coproate
Publishing Ltd. Chapter 1 pp33.ISBN 3 – 89578 – 152 – 5.
[5] Bird, J. (2004). Electrical Circuit Theory and Technology. Revised 2
nd
Edition. Newnes. Chapters 30, 31, 33, 44 pp 531 – 559, 575 – 598 and
869 – 900. ISBN 0 – 7506 – 5784 – 7
[6] Stevenson, WD Jr. (1982). Elements of Power System Analysis. 4
th
Edition. McGraw Hill. Chapters 5, pp 88 – 126. ISBN 0 – 07 – 066584 –
2
[7] Yu, JG, (1992). Computer Analysis of Touch Voltages and Stray Currents
for DC Railways. PhD Thesis submitted to the University of Birmingham,
UK
[8] White, RD, Zhang, Z (2003) Atkins Rail internal report. Predictions of
50Hz induced voltage into lineside cables on 25kV railway. Report
Number: BF 5001270/RDW/001. Issue 4.0 August 2003
[9] Case, S, (2005). Earthing, Bonding and Stray Current Mitigation 
Principles. The IEE Second Residential Course on REIS 2005.
Wrightsons. Chapter D4. ISBN 0 – 86341 – 511 – 3.
[10] Gerald CF, Wheatley PO. (2004). Applied Numerical Analysis. 7
th
Edition.
Pearson/Addison Wesley. Chapter 2 pp 88 – 100. ISBN 0 – 321 – 1909 –
X
Power Supply System Analysis, Design and Planning 133
This page intentionally left blank
Optimal design of power supply systems using
genetic algorithms
J . R. J imenezOctavio & E. Pilo
Instituto de Investigación Tecnológica, E.T.S. de Ingeniería ICAI,
Univ. P. Comillas, Madrid, Spain
Abstract
This paper presents an optimization model based on genetic algorithms for
designing traction power supply systems. The proposed model is suitable both
for planning new lines and for expanding the old ones, resulting in a more
efficient operation as well as in lower investment costs. The minimization of
fixed installation represents the optimization criterion for searching innovative
designs that fulfil certain technical constraints: maximum voltage drops and
maximum power consumption in the substations. The variables involved in the
optimization problem are: number, type and location of railway overhead lines;
and number, size and location of traction substations. Finally, the evaluation of
possible designs involves simplified electrical modelling of the studied railway
stretch. Thus, electrical simulations and calculations have also been adapted for
their implementation in the genetic algorithm. A section of the MadridBarcelona
highspeed line has been considered as a study case in order to analyze the
performance of the proposed model. The results reveal the suitability of the new
designs obtained with the presented model and its goodness and robustness.
Keywords: railway, power supply system, optimization, genetic algorithms.
1 Introduction
The design of the electrification is a complex task that involves different
interdisciplinary analysis. Normally, the design process is done iteratively by
refining candidate designs based on its estimated performance. In this process
simulation tools are crucial to evaluate the performance in many different
situations. However, they do not usually include criteria to modify candidate
designs in order to obtain the final solution. Thus, the designer has to decide the
Power Supply System Analysis, Design and Planning 135
best changes to be done to the candidate design in order to fulfil technical
constraints and to reduce costs in a reasonable way. As this is not an obvious
task, the development of decisionmaking computeraided tools based on
optimization models can be very useful to ensure efficient uses of investments.
In [1] simplified models have been formulated to represent bivoltage 2x25kV
meshed topologies in a simplified way. Based in these models, an optimization
procedure has been presented in [2] to select the most efficient centenary and
autotransformer combinations. The results of this optimization can be used as an
input for the optimization of the powersupply system presented in [3], which
formulates a mixed integer programming (MIP) optimization problem. This
procedure is suitable for smallmedium network sizes (up to 200250 km), but
for bigger problems computation times are too long. In this paper, a genetic
algorithm based optimization procedure (GA) is presented and its performance is
analyzed. Furthermore, the solutions and computing times are compared to the
MIPbased optimization.
The optimization is formulated to minimize the total construction cost of the
power supply system while fulfilling technical restrictions. The optimization
variables include: (i) substations (number, location), (ii) location of neutral
zones, (iii) catenary for every sector (starting and ending point, feeding system,
conductors), and (iv) the locations of the required autotransformers. The total
construction cost includes: (i) substations, (ii) their connection to the high
voltage network, (iii) catenaries and (iv) autotransformers.
The second section of this paper describes the electrification of AC power
supply systems. The third section presents how the genetic algorithm is
formulated in order to perform the optimization. In the fourth section, the
performance of the proposed method is evaluated in a section of the new Madrid
Barcelona highspeed line. Finally, the fifth section summarizes the conclusions
of this work.
2 AC power supply
As shown in Figure 1, the railway electrical system is divided in electrically
isolated singlephased sectors, which are fed from the threephase network
through a traction substation [4, 5].
Sector 3R Sector 1R Sector 2L Sector 2R Sector 3L Sector 1L
Threephase highvoltage network
Traction
substation 1
Traction
substation 2
Traction
substation 3
Figure 1: General structure of the power supply system.
136 Power Supply, Energy Management and Catenary Problems
These substations are connected between two of the three phases of the high
voltage network. It should be noted that topology could be modified in case of
failures to guarantee the operation. For instance, if one of the transformers of a
substation fails, the other takes on the corresponding sector.
Each of these sectors can use either monovoltage system (such as 1x25kV)
or bivoltage (such as 2x25kV) system. In monovoltage systems, the feeding
conductors are set directly to the specified voltage level (see Figure 2). In bi
voltage systems, a higher voltage is set between feeding conductors [6, 7]. This
voltage is reduced by using autotransformers distributed along the catenary (see
Figure 3).
HighVoltage network
Positive
Train
Figure 2: Monovoltage system configuration.
HighVoltage network
Positive
Train
Negative
Figure 3: Dual system configuration.
In [1], the behaviour of dual systems is analyzed and a model is presented to
represent bivoltage systems as if they were monovoltage systems. This
representation is used in this paper to characterize monovoltage and bivoltage
systems in a similar manner.
3 The optimization problem
An optimization problem formulation has been carried out introducing several
criteria for improving classic designs. Particularly, total construction costs are
minimized and technical criteria are considered as restrictions. This optimization
problem has been tackled using a genetic algorithm.
While solving the optimization problem, the number of sectors (see Figure 2)
is determined and thus the number of substations and neutral zones, as well as
their location. Moreover, the catenary type (and number of autotransformers in
dual systems) of each subsector is obtained.
Power Supply System Analysis, Design and Planning 137
The formulation of this problem is organized in five steps. Firstly, the
constant parameters are presented. Then, the optimization variables are
established. Next, the objective function is defined. The considered restrictions
are formulated afterwards and finally, certain parameters and questions related to
the genetic algorithm implemented are pointed out.
3.1 Constant parameters
The input data used as constant terms are called constant parameters. In the
proposed design method, several calculations have to be done prior to the
optimization itself.
It has been supposed that the cost of installing a traction substation depends
mainly on the place where it has been located. After studying the railways to be
electrified and the possible connections to the available threephase networks,
the railway has to be partitioned into zones and the best feeding solution has to
be identified for each zone. For each zone, the following parameters have to be
specified:
 Position where each zone ends.
 Cost of installing a substation in each zone. Both the cost of the substation
and the cost of the line to connect it to the selected threephase network are
included.
 Maximum power that could be delivered by a substation depending of the
zone it was installed. This value has to be established to avoid (i)
excessive voltage drop or excessive unbalances in the threephase
network, and (ii) excessive power flow through the transformers of the
substation.
To perform this optimization, it has been assumed that a reduced set of
catenary types (with associated autotransformers in dual systems) has previously
been selected. A multicriteria optimization can be done to determine this set [2],
but actually it is not necessary for the purposes of this paper.
For each catenary type, c , the following parameters have to be specified:
 Impedance per length unit. In dual systems, the equivalent model [1] has to
be used to represent them as simple systems.
 The impedance per length unit (see [1]) due to the autotransformers
separation (for simple systems its value is 0). This parameter depends on the
catenary line parameters.
 The cost per length unit of installing each catenary type.
 The cost of installing each catenary type in a subsector that does not depend
on the length of catenary installed. It corresponds to the cost of the
autotransformers and thus is 0 for simple systems.
Several operation scenarios are used to evaluate technical constraints. A
scenario is a list of trains that are simultaneously consuming electrical power. To
138 Power Supply, Energy Management and Catenary Problems
define each scenario, sc , the following parameters have to be specified for each
train, t :
 Position of each train in the railway.
 Power that consumes each train:
t
S .
Other parameters to be defined are:
 Length of the railways to be electrified
 Maximum acceptable voltage drop
 Impedance of substations (all the substations are assumed to have the same
value).
 The current
t
I consumed by each train. It is estimated assuming an average
voltage:V
( )
( )
,
,
t
t
S sc t
I sc t
V
= (1)
3.2 Variables
The variables are quantities that are changed in order to minimize the total cost
of construction and to fulfil with technical constraints. The variables defined in
the proposed model are:
 Number of traction substations supplying the stretch considered for the
optimization.
 Location of each substation along the stretch.
 Type of catenary used in each subsector.
3.3 Objective function
The total cost of construction of the electrification to be minimized (see Eq.
(2)) includes the cost of building the substations,
Sub
C , the cost of the
catenaries,
Cat
C , and the cost of the autotransformers,
AT
C , (if dual systems
are used).
Sub Cat AT
C C C C = + + (2)
3.4 Restrictions
The following restrictions have been included in the genetic algorithm
formulation as penalty factors added to the fitness function.
 Maximum acceptable voltage drop for the operating conditions.
 Maximum power that can be delivered by the substations.
Power Supply System Analysis, Design and Planning 139
From the genetic algorithm point of view, the dynamic behaviour of these
restrictions is remarkable. That is, the restrictions are included in a penalty
function that: is active in case of not fulfilling each restriction; and
proportionally increases with the number of generations.
3.5 Genetic algorithm
Over the years, much work has been done in engineering optimization and the
current tendency is to deal with reallife applications which are multiobjective
by nature. Therefore, a standard genetic algorithm has been applied to the
computation of the aforementioned optimization problem, although some
specific tools and functions have been tailored too. Thus, its main characteristics
can be summarized as follows:
 The individuals have been codified in binary code, including each
chromosome the optimization variables described in section 3.2.
 The mechanisms of nature used for this optimization have been: a Roulette
based algorithm for the selection process; and monopoint schemes for
crossover and mutation processes, whose parameters of probability are
collected in Table 1.
 The fitness function that leads the genetic algorithm has been built by the
objective function, section 3.3, and the penalty functions derived from
section 3.4, as follows:
Sub Cat AT
F C C C t = + + + (3)
( ) ( )
1 1
d s
d
s
n
r v
t t t
t 
t
= ·
= + · ÷
= ·
(4)
where F is the fitness function which depends of the different costs involved in
optimization problem and a penalty factor. This penalty function is composed of
a static and a dynamic part as can be seen in Eq. (4), where: r is the static
penalty,  is the dynamic penalty, n is the number of generations computed and v
the degree of violation of each restriction.
Table 1 collects the numeric parameters of the genetic algorithm used in the
numerical simulation correspond to standard values of randomized operators and
another related to penalty functions.
Table 1: Genetic algorithm parameters.
Property Value
Static penalty, r 10
6
Dynamic penalty  0.2
Crossover probability 0.7
Mutation probability 0.03
Population size 200
140 Power Supply, Energy Management and Catenary Problems
4 Example
The proposed model has been used to obtain the power supply of a section of the
new Madrid – Barcelona highspeed line between the km 306 and km 441.6.
Table 2: Catenary repository to be used.
Catenary C5 C4 C3 C2 C1
Contact wire Bz 150mm2 Bz 150mm2
Bz
150mm2
Bz
150mm2
Bz
150mm2
Sustainer wire Bz 100mm2 Bz 100mm2
Bz
100mm2
Bz
100mm2
Bz
100mm2
Positive feeder Without feeder
Without
feeder
LA110 LA280 LA280
Negative feeder LA110 LA380 LA280 LA380 LA380
Return wire LA110 LA280 LA110 LA280 LA280
Number of AT per sector 1 1 1 1 2
ZCAT (ohm/km) 0,100530 0,085893 0,072287 0,066348 0,066348
CFIX (units) 200 200 200 200 400
CCAT (units/km) 26,136 28,140 30,900 33,840 33,840
ZDESV (ohm/km) 0,037155 0,036300 0,033548 0,028995 0,014497
Table 3: Zone partitioning for installing a substation.
From km To km CZONE PZONE Description
306 (Start) 341,5 1000 60 MVA Fed from 220kV network
341,5 346 ∞  Forbidden due to environmental protection
346 350 1500 100 MVA Fed from 220kV network
350 377 ∞  Forbidden due to environmental protection
377 397 2500 100 MVA Fed from 400kV network
397 410 ∞  Forbidden due to environmental protection
410 413 2000 60 MVA Fed from 220kV network
413 415 ∞  Forbidden due to environmental protection
415 441.6 (End) 2000 60 MVA Fed from 220kV network
The scenarios that have been considered in this study case correspond to
traffic meshes of trains consuming 24MVA (typically double compositions of
12MVA trains) separated by 24 km. These hypotheses correspond to highspeed
trains circulating at 350 km/h, with frequencies of 4 min.
Table 2 shows the considered catenary repository. It has been obtained by
using the multicriteria optimization procedure described in [3].
Table 3 shows the cost of installing a substation and the maximum power that
could be delivered depending on the location.
The results obtained can be discussed in two main aspects: on the one hand
the robustness of the genetic algorithm and; on the other hand, the accuracy by
comparing with another optimization procedure.
Figure 4 shows the fitness function evolution along generations of the genetic
algorithm. In different colours are plotted: the optimum evolution, which is
reached with only 11 generations; and the first and second quartiles, which can
give an idea of the robust behaviour of the algorithm. It seems clear that at the
Power Supply System Analysis, Design and Planning 141
tenth generation penalized individuals disappear of most of the solutions. Thus,
at least the 50% of possible solutions are feasible and their fitness values are
exactly the costs of the objective function.
Figure 5 shows the location of substations and subsectors; and the types of
catenaries (red lines refers to the solution obtained with the genetic algorithm
0 2 4 6 8 10 12 14 16 18 20
1
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
x 10
4
Generation
F
i
t
n
e
s
s
Q2
Q1
Optimum evolution
Optimum
Figure 4: Fitness evolution.
0 20 40 60 80 100 120
2
3
4
5
Railway stretch [km]
C
a
t
e
n
a
r
y
t
y
p
e
Limit of cost zone
Catenary per subsector−GA
Substation−GA
Neutral zone−GA
Catenary per subsector−MIP
Substation−MIP
Neutral zone−MIP
Figure 5: Location of substations and subsectors and types of catenaries.
142 Power Supply, Energy Management and Catenary Problems
and blue lines to the MIPbased one (colour refers to the online version)).
Marked with circles are pointed out the location of substations along the stretch
considered, between the km 306 and km 441.6 of the MadridBarcelona high
speed line. However, the horizontal axis shows the relative kilometre points,
from km 0 to km 135.6, to make clearer the understanding. Moreover, the
vertical axis means the type of catenary used in each subsector. As well as in
Figure 5, Table 4 collects the numeric catenary assignment that has been
obtained with the GA optimization, while Table 5 shows the assignment
obtained with the MIP optimization.
Table 6 compares the minimum total cost of the electrification obtained with
both methods as well their computation time. The proposed genetic algorithm
finds similar solutions in terms of total cost in significantly shorter time.
Table 4: Catenary assigned to each sector in the GA model.
From km To km Catenary
306.000 (substation) 323.15 (neutral zone) C4
323.15 (neutral zone) 340.3 (substation) C5
340.3 (substation) 367.55 (neutral zone) C4
367.55 (neutral zone) 394.8 (substation) C3
394.8 (substation) 417.9 (neutral zone) C5
417.9 (neutral zone) 441 (substation) C5
Table 5: Catenary assigned to each sector in the MIP model.
From km To km Catenary
306.000 (substation) 332.791 (neutral zone) C5
332.791 (neutral zone) 347.791 (substation) C5
347.791 (substation) 374.824 (neutral zone) C5
374.824 (neutral zone) 391.138 (substation) C5
391.138 (substation) 425.534 (neutral zone) C5
425.534 (neutral zone) 441.200 (substation) C4
Table 6: Compared results between MIP and GA models.
Model Global Cost CPU Time
MIP 11760 € 1:05:42 Pentium4, 1.7 GHz
GA 11813 € 0:00:58 Pentium4, 1.7 GHz
5 Conclusions
This paper presents a genetic algorithm based procedure whose main goal I to
minimize the total cost of construction of the power supply while satisfying
technical constraints such as voltage drops or power limitations.
For the sake of the evaluation of the performance of the model, a section of
the new Madrid – Barcelona highspeed line has been studied with a MIP
Power Supply System Analysis, Design and Planning 143
optimization together with the proposed genetic algorithm optimization. Both
procedures have found quite similar solutions in terms of costs, but the proposed
algorithm performs the optimization in a significantly shorter time.
As the geneticalgorithm approach has obtained very promising results, future
developments will be focused on designing particular evolutionary strategies to
enhance the convergence. Additionally, more complex technical restrictions will
be considered.
References
[1] E. Pilo, L. Rouco, and A. Fernández, "A reduced representation of 2x25 kV
electrical systems for highspeed railways." IEEE/ASME J oint Rail
Conference, Chicago, 2003.
[2] E. Pilo, L. Rouco, and A. Fernández, "Catenary and autotransformer coupled
optimization for 2x25kV systems planning" WIT Transactions on the Built
Environment, vol. 78, pp. pp.747756, 2006.
[3] E. Pilo, L. Rouco, A. Fernández, "An optimization procedure to determine
the topology of ac railways power supply networks" ASME/IEEE J oint Rail
Conference & Internal Combustion Engine Spring Technical Conference
(J RCICE2007). Pueblo, Colorado, USA, March 1316, 2007
[4] J . D. Glover, A. Kusko, and S. M. Peeran, "Train voltage analysis for AC
railroad electrification" IEEE Transactions on Industry Applications, vol.
IA20, pp. 925934, 1984.
[5] C. Gourdon and C. Herce, "The overhead system of the TGVAtlantique,"
presented at International Conference on Main Line Railway Electrification
(Conf. Publ. no.312), London, UK, 1989.
[6] H. Roussel, "Power supply for the atlantic TGV highspeed line," presented
at IEE International Conference on Main Railway Electrification, York,
1989.
[7] R. J . Hill and I. H. Cevik, "Parallel computer simulation of autotransformer
fed AC traction networks," presented at ASME/IEEE J oint Railways
Conference, 1990.
144 Power Supply, Energy Management and Catenary Problems
Application of linear analysis in railway power
system stability studies
S. Danielsen, T. Toftevaag & O. B. Fosso
Department of Electric Power Engineering, Norwegian University of
Science and Technology, Trondheim, Norway
Abstract
Dynamical phenomena, such as oscillations and instability in railway power
systems, have become of growing concern in the experts’ community in recent
years. On several occasions, modern advanced electric rail vehicles have been
the source for low frequency power oscillations leading to an unstable power
system due to the lack of damping, and as a consequence of operating problems.
A method to study these phenomena is needed. Well known linear techniques
based on smallsignal analysis provide valuable information about the inherent
characteristics of even nonlinear singlephase power systems. This paper
describes how a railway power system and its dynamical railwayrelated
components are modelled in a commercially available power system analysis
software and studied by linear analysis such as eigenvalues, participation factors
and parameter sensitivities. This is used to gain knowledge about the interaction
between the rail vehicles and the electric infrastructure. Linear analysis is found
to be a powerful tool in this respect, provided that adequate models of the
relevant components can be established in the RMS mode. The results reflect the
experienced poor interaction.
Keywords: AC railway power supply, traction power system, stability, advanced
electric rail vehicle, rotary frequency converter, lowfrequency oscillations,
eigenvalue analysis.
1 Introduction
The recent development of electric rail vehicles with utilization of power
electronics and complex control systems has introduced new phenomena of
dynamical interaction between the vehicles and the railway power supply. One
Power Supply System Analysis, Design and Planning 145
G M
Long distance 60 km
1~16 ⅔ Hz system
Rotary converter
3~50 Hz system
Advanced electrical rail vehicle
Figure 1: Sketch of the railway power system to be studied.
such phenomenon is lowfrequency power oscillations leading to system
instability, typically in the frequency range of 0.10.3 times the fundamental
frequency. The vehicles may oscillate together internally (Menth and Meyer [1])
or against the electric infrastructure (Danielsen and Toftevaag [2]). There is a
need for an integrated method for studying these oscillations.
Lowfrequency power oscillations are not new to power systems (Kundur
[3]). One method to study these inherent qualities of the power system is by
utilization of linear analysis. Several specialized power system analysis
computer programs include such tools, i.e. Simpow (Fankhauser, et al. [4]).
This paper introduces linear analysis to a railway power system study where
the system comprises both electric infrastructure and rolling stock as shown in
Figure 1. An alternative method is timedomain simulations as used by
Eitzmann, et al. [5].
2 Linear analysis theory
A dynamical system may be described by a number of characteristic differential
equations, normally based on the physics of the system to be studied. Based on
these equations and information about the initial conditions, the state of the
system can be determined and the response of a disturbance can be calculated.
If the system is nonlinear, commonly the system is linearized around an
operating point (Δvalues). In that way the mathematical tools that are used for
linear systems can be utilized for nonlinear systems as well, such as railway
power systems. This is formally only valid in vicinity of the linearization point.
A common way to describe a linear or linearized system is by a state space
model as in eqn. (1) where x is the state vector containing the state variables,
i.e. the variables of which the time derivative is to be considered, and u is the
vector containing disturbance variables. x is a vector containing the time
derivative of the state variables andy is called the output vector.
A = A + A
A = A + A
x A x B u
y C x D u
(1)
Matrix A is the state matrix and contains important information about the
inherent qualities of the linear/linearized system. The roots of the characteristic
eqn. (2), λ, are called the eigenvalues of the system.
det( ) 0 ì ÷ = A I (2)
146 Power Supply, Energy Management and Catenary Problems
The number of eigenvalues for a system is equal to the dimension of A and
the number of firstorder differential equations describing the system. An
eigenvalue can be a complex number λ = σ + jω describing a mode of the system
where the imaginary part describes the oscillation frequency and real part
describes the damping or time decay of the oscillation. A negative real part
identifies a stable mode. Complex eigenvalues appear as conjugate pairs, but
only the eigenvalue having positive imaginary part is shown in the figures in this
paper.
The concept of participation factors utilizes information from the state matrix
to find a measure for the relative participation of a specific state variable in a
specific mode. The factor itself is a phasor where its length reflects the relative
degree of participation in the mode compared to the other state variables (Kundur
[3]).
Linear analysis of large power systems requires a lot of computations, but it is
a simple task for a computer with suitable software (Slootweg, et al. [6]).
3 Concept of power system modelling
In an AC power system, both voltage and current vary as a sine with the
fundamental frequency f
1
(such as 16 ⅔ Hz or 50 Hz) plus additional harmonics.
The voltage drop Δu(t) given by a current i(t) over an inductive impedance is
described eqn. (3).
( ) ( )
( )
di t
u t R i t L
dt
A = · +
(3)
A common simplification for power system analysis is to represent voltages
and currents by phasors, i.e.
Re Im
U U j U A = A + A
. The calculated voltage
drop over the impedance will then depend on the resistance R and the
fundamental angular frequency ω
1
=2πf
1
times the inductance L. This common
simplification implies that the current is no longer a state variable where its time
derivative is considered. In traditional power systems analysis, the impact of this
simplification may be neglected in the view of stability (Kundur [3]). For
analysis of power systems with power electronic converters such as an advanced
electric rail vehicle, this simplification may not be valid any more (Danielsen,
et al. [7]). In the present studies the current is therefore kept as a state variable
giving the expression for the voltage drop of any series inductive impedance in
both singlephase and threephase networks as in eqn. (4). This includes the
synchronous machines’ stator inductances as well.
1
Re Re
Im Im
1
d
R L L
U I
dt
U I d
L R L
dt
e
e
(
+ ÷
(
A
( (
=
(
( (
A
¸ ¸ ¸ ¸ (
+
(
¸ ¸
(4)
Keeping the current as a state variable is often used in studies of power
electronic components in instantaneous value mode analysis of threephase
systems modelled in the rotating dq reference frame (Harnefors [8]), but is here
Power Supply System Analysis, Design and Planning 147
applied to an entire singlephase power system. The phasors are RMS values of
the respective voltages and currents, i.e. root mean squared values over one
fundamental frequency period, hence harmonic effects are neglected.
4 Rotary converter
4.1 Introduction
The rotary converter is the dominant solution of electric power conversion from
the threephase 50 Hz utility grid to the singlephase 16 ⅔ Hz decentralized fed
railway power system in Norway and Sweden (Banverket/J ernbaneverket [9]).
Such a converter consists of a threephase synchronous motor (M) mounted on
the same shaft as a singlephase synchronous generator (G), see Figure 1. Both
motor and generator are equipped with automatic voltage regulators and exciters.
4.2 Electromechanical eigenfrequency
A pronounced characteristic of these rotary converters is the poorly damped
electromechanical eigenfrequency around 1.6 Hz due to the lack of explicit
motor damper windings (Toftevaag and Pálsson [10]). These low frequency
oscillations are shown to be related to the basic swing equation (Biesenack [11]).
This equation is linearized in eqn. (5) and describes the electromechanical
behaviour (which is characteristic for all synchronous machines (Kundur [3])):
2
' 1
2
2 0
E
d d
H D K
dt dt
o o
e o
A A
+ + · · A =
. (5)
H Inertia constant in MWs/MVA
Δδ Change in power angle radians
D Damping constant in pu torque/pu speed
K
E’
Transient synchronizing torque coefficient in pu torque/rad
From the swing equation, an expression (eqn. (6)) for the eigenvalues λ
1,2
describing the electromechanical swing mode for a single machine connected to
stiff network can be derived (in this case the synchronous motor connected to the
50 Hz network):
2
' 1
1,2
4 4 2
E
K D D
H H H
e
ì
·
 
= ÷ ± ÷

\ .
. (6)
Eqns. (5) and (6) are valid for the converter in islanded operation, and need
some modifications to be valid in interconnected operation when taking the
damping and synchronizing torque of the generator into account as well.
4.3 Linear analysis
In this paper, a rotary converter in islanded operation is studied. There is no
connection to other converters in the railway power system. The converter is
148 Power Supply, Energy Management and Catenary Problems
connected to a typical 66 kV threephase utility grid having a shortcircuit ratio
of 250 MVA. The converter is feeding a 3.67 MW resistive load over 60 km
overhead contact line with the impedance of (0.19+j0.21) Ω/km. Including the
line’s transmission losses this results in a loading close to the converter’s rated
load. Both machines are represented by 5
th
order standard synchronous machine
models which are increased to 7
th
order models due to eqn. (4).
The lowfrequency modes found by linearization of the converter at this
operation point are shown in Table 1. Classification of the different modes is
done by use of participation factors that indicate which state variables that
participates mostly in each mode. The most participating state variables in the
electromechanical mode are the converter speed and motor power angle.
Damping ratio for the electromechanical mode is low but acceptable.
Table 1: Lowfrequency modes from linear analysis of rotary converter.
No. Eigenvalues Mode Damping ratio
8, 9
– 5.00 [1/s] ±j2.90 [Hz]
Generator exciter 26%
10, 11
– 4.83 [1/s] ±j3.19 [Hz]
Motor exciter 23%
13, 14
– 0.46 [1/s] ±j1.59 [Hz]
Electromechanical 5%
5 Advanced electric rail vehicle
5.1 Introduction
Almost all new electric rail vehicles today utilize the advantages of the induction
machine as the traction motor (Östlund [12]). The motor speed and torque are
controlled by use of a threephase power electronic pulsewidth modulated
inverter. The motor inverter takes the required power from the vehicle internal
DClink capacitor as shown in Figure 2. It is the task for the line inverter to
modulate the DClink voltage into an AC voltage at main transformer low
voltage side in amplitude and phase such that the resulting voltage drop over the
transformer leads to the needed current to keep the DC link voltage at reference.
3~M
=
3~
1~
=
Control system
Singlephase
main
transformer
Line
inverter
DC
link
Motor
inverter
Threephase
asynchronous
motor
Figure 2: Sketch of an electric advanced rail vehicle.
Power Supply System Analysis, Design and Planning 149
All these components are controlled by an advanced computer based control
system. As both line and DC link capacitor exchange energy and because the
power control cannot be faster than the fundamental frequency, this result in a
dynamical continuous feedback system (Menth and Meyer [1]).
5.2 The model
A model of such an advanced rail vehicle is made based on literature (Östlund
[12] and Steimel [13]) and is given a thorough description by Danielsen, et al.
[7] for use in RMS simulations. Such RMS models of rail vehicles are not
standard as most simulation studies in the rail vehicle industry are performed in
instantaneousvalue timedomain simulations. The model comprises two major
simplifications; the motor side is replaced by a resistor and the control system is
analogous and continuous. Also, the reactive power consumption at current
collector is controlled to be zero.
The line inverter control consists of the following controllers: Synchronizing
controller (phase locked loop, PLL) to track the phase of the line voltage, DC
link voltage controller (VC) for active power control and AC current controller
(CC). The controllers are implemented in a vehicleinternal rotating reference
frame, also known as vector control.
5.3 Linear analysis
The rail vehicle is operated at approximately half its rated power (3.67 MW) fed
from an ideal voltage source through 60 km of overhead contact line. This means
that in this subcase, there is no rotary converter in the system.
The linear analysis identifies two lowfrequency modes. Their figures are
shown in Table 2. The vehicle seems to be well damped. Eigenvalues no. 12 and
13 will be subject to further studies as it describes the power oscillations.
In order to check the sensitivity of the interesting low frequency mode
(eigenvalue 12 and 13), Figure 3 shows how the eigenvalue (positive imaginary
part only as the pair is a complex conjugate) moves when different parameters
for the locomotive are changed. The range of the change is from 0.5 to 2.0 times
the original value which is given in brackets in the legend. Arrows show the
movement direction of the eigenvalue when the most influencing parameters are
increased. The different parameters are explained in appendix.
It can be noticed that reducing either the voltage controller integration time
(TIVC) or the current controller gain (KPCC) will within the range alone move
Table 2: Lowfrequency modes from linear analysis of the vehicle.
No. Eigenvalues Mode Damping ratio
6, 7 – 12.9[1/s] ±j7.04 [Hz]
AC current measurement,
AC voltage measurement
28%
12, 13 – 5.67 [1/s] ±j3.31 [Hz]
VC, AC current meas.,
DClink voltage, PLL
26%
150 Power Supply, Energy Management and Catenary Problems
0
1
2
3
4
5
6
25 20 15 10 5 0 5
Real [1/s]
I
m
a
g
i
n
a
r
y
[
H
z
]
MP (2)
KUSOGI (0,8)
KISOGI (1)
TIPLL (0,079)
KPPLL (51)
TIVC (0,04)
KPVC (0,9)
TICC (0,5)
KPCC (1,347)
TUDC (0,008)
TIDC (0,008)
C (0,04)
KPVC
C
KISOGI
MP
TIVC
KPCC
KUSOGI
Figure 3: Root loci for the lowfrequency mode when parameters are
changed.
the eigenvalue into the right half plane and make the vehicle unstable. The
voltage controller gain (KPVC) has also influence on the mode. Changed DC
link capacitance (C) will move the eigenvalue in the same way as the inertia
constant H for synchronous machines in eqn. (6) as both express energy storages.
6 System interaction
6.1 Unstable system
The rotary converter and the rail vehicle are now combined into one system as
shown in Figure 1. For the vehicle this leads to two changes even though the
operating point is kept (as in section 4.3 and 5.3): First, the voltage source (both
amplitude and phase) is no longer constant but will change as the converter
oscillates. And second, the total inductance L that the vehicle current controller
has to change the line current in has increased due to the converter transformer
and the generator stator windings. The latter change may have an impact on the
vehicle lowfrequency mode.
Table 3: Lowfrequency modes from linear analysis of the railway power
system.
No. Eigenvalues Mode Damping ratio
16, 17
– 4.83 [1/s] ±j3.18 [Hz]
Motor exciter 24%
18, 19
– 5.14 [1/s] ±j2.94 [Hz]
Generator exciter 27%
23, 24
– 2.19 [1/s] ±j2.18 [Hz]
Vehicle DClink 16%
25, 26 +0.23 [1/s] ±j1.59 [Hz] Rotary conv. electromech. 2%
Power Supply System Analysis, Design and Planning 151
The lowfrequency modes are shown in Table 3. Both damping and frequency
of the vehicle DClink mode is decreased. The real part of the eigenvalue
describing the converter’s mode is positive and indicates an unstable system.
6.2 Stability improvement
By use of the linear analysis tool, again the influence of the vehicle control
parameters on the different modes can be studied. Figure 4 shows the converter’s
electromechanical root loci when the parameters for the voltage and current
controllers are changed. The arrows show the direction of movement when the
parameter value is increased.
1,4
1,5
1,6
1,7
0 0,1 0,2 0,3 0,4 0,5
Real [1/s]
I
m
a
g
i
n
a
r
y
[
H
z
]
KPVC
TIVC
KPCC
TICC
TICC KPCC
TIVC
KPVC
Figure 4: Root loci for the rotary converter mode when vehicle control
parameters are changed (0.5 to 2.0 times original value).
It can be observed that all the control variables influence on the damping of
the converter’s eigenfrequency. None of them, however, are alone able to
stabilize the system within the variation range studied.
The sensitivity to the control parameters for the converter’s mode may be
compared to the respective sensitivities of the vehicle DClink mode shown in
Figure 3. Increase of TIVC, KPVC and KPCC will increase damping of both
modes. These parameters should therefore be focused when improving the
stability in this case. Further studies are however needed to see if change of
vehicle control parameters only is sufficient to stabilize the system or not.
6.3 Interaction with a poorly damped rotary converter
To show how the rotary converter and the vehicle interact in this unstable
system, the absolute value of the participation factors for the unstable
electromechanical mode +0.23 1/s ±j1.59 Hz are shown in Figure 5. State
variables given by in eqn. (4) and the synchronous machines’ exciters are
152 Power Supply, Energy Management and Catenary Problems
0,00
0,02
0,04
0,06
0,08
0,10
0,12
D
s
t
a
t
o
r
w
i
n
d
.
Q
s
t
a
t
o
r
w
i
n
d
.
D
d
a
m
p
e
r
w
i
n
d
Q
d
a
m
p
e
r
w
i
n
d
P
o
w
e
r
a
n
g
l
e
S
p
e
e
d
F
i
e
l
d
w
i
n
d
i
n
g
D
s
t
a
t
o
r
w
i
n
d
.
Q
s
t
a
t
o
r
w
i
n
d
.
D
d
a
m
p
e
r
w
i
n
d
Q
d
a
m
p
e
r
w
i
n
d
P
o
w
e
r
a
n
g
l
e
S
p
e
e
d
F
i
e
l
d
w
i
n
d
i
n
g
A
C
v
o
l
t
.
m
e
a
s
.
A
C
v
o
l
t
.
m
e
a
s
.
P
L
L
P
L
L
D
C
v
o
l
t
.
m
e
a
s
.
D
C
c
u
r
r
.
m
e
a
s
.
D
C
v
o
l
t
.
c
o
n
t
r
.
A
C
c
u
r
r
.
m
e
a
s
.
A
C
c
u
r
r
.
m
e
a
s
.
A
C
c
u
r
r
.
c
o
n
t
r
.
A
C
c
u
r
r
.
c
o
n
t
r
.
D
C
l
i
n
k
v
o
l
t
.
Converter motor Converter generator Vehi cl e
0.48 0.24 0.24
Figure 5: Absolute value of participation factors to show how state variables
participate in the rotary converter electromechanical mode.
omitted. Participation of the converter speed and motor power angle is easily
observable. In the vehicle control system, the AC voltage measurements and the
DClink voltage controller integral part point out as the largest participants.
7 Concluding remarks
In this paper linear analysis tools for studying lowfrequency power system
oscillations have been applied to a railway power system to gain knowledge
about the interaction between power supply infrastructure and rolling stock. The
linear analysis clearly states the instability and provides information about which
parts of the advanced components that participate in the oscillation and which
parameters to change for improvement. The vehicle DClink voltage controller
points out to have an important role.
It is obvious that good results require adequate models that may be difficult to
obtain, at least for the very complex advanced electric rail vehicles. The concept
of keeping current as a state variable needs more investigation.
Acknowledgement
The work presented in this paper is a part of a PhD study funded by the
Norwegian National Rail Administration (Jernbaneverket).
References
[1] Menth, S. & Meyer, M., Low frequency power oscillations in electric
railway systems, Elektrische Bahnen. 104 (4), pp. 216221, 2006.
[2] Danielsen, S. & Toftevaag, T., Experiences with respect to low frequency
instability from operation of advanced electrical rail vehicles in a traction
Power Supply System Analysis, Design and Planning 153
power system with rotary converters. MET2007 8th International
Conference "Drives and Supply Systems for Modern Electric Traction".
Warsaw, Poland, pp. 5157, 2007.
[3] Kundur, P., Power System Stability and Control, McGrawHill California,
1994.
[4] Fankhauser, H.R., Aneros, K., Edris, A.A. & Torseng, S., Advanced
simulation techniques for the analysis of power system dynamics,
Computer Applications in Power, IEEE. 3 (4), pp. 3136, 1990.
[5] Eitzmann, M.A., Paserba, J .J ., Undrill, J .M., Amicarella, C., J ones, A.L.,
Khalafalla, E.B. & Liverant, W., Model development and stability
assessment of the Amtrak 25 Hz traction system from New York to
Washington DC. Railroad Conference. pp. 2128, 1997.
[6] Slootweg, H., Persson, J., van Voorden, A.M., Paap, G.C. & Kling, W., A
Study of the Eigenvalue Analysis Capabilities of Power System Dynamics
Simulation Software. 14th Power Systems Computation Conference 2002.
Sevilla, Spain, pp. 2002.
[7] Danielsen, S., Fosso, O.B., Molinas, M., Suul, J .A. & Toftevaag, T.,
Simplified models of a singlephase power electronic inverter for railway
power system stability analysis – development and evaluation, Accepted for
publication in Electric Power System Research September 1st 2009. pp.
2009.
[8] Harnefors, L., Modeling of ThreePhase Dynamic Systems Using Complex
Transfer Functions and Transfer Matrices, Industrial Electronics, IEEE
Transactions on. 54 (4), pp. 22392248, 2007.
[9] Banverket/J ernbaneverket, Requirements on rolling stock in Norway and
Sweden regarding EMC with the electrical infrastructure and coordination
with the power supply and other vehicles (BVS 543.19300/J D 590), 2007
[10] Toftevaag, T. & Pálsson, M.T., Low frequency oscillations in the
Norwegian electric traction power supply caused by interaction between the
supply system and propulsion machinery – analysis and consequences.
MET’05 7th International Conference “Modern Electric Traction in
Integrated XXI st Century Europe”. Warsaw, Poland, pp. 137142, 2005.
[11] Biesenack, H., Theorie und Betriebsverhalten von SynchronSynchron
Umformern, Doktor Ingenieur thesis, Fakultät für Technik und
Naturwissenschaft, Hochschule für Verkehrswesen "Friedrich List", 1972.
[12] Östlund, S., Electric Traction (original title in Swedish: Elektrisk Traktion),
School of Electrical Engineering, Electrical Machines and Power
Electronics, Royal Institute of Technology (KTH), Stockholm, Sweden,
2001.
[13] Steimel, A., Electric Traction  Motive Power and Energy Supply,
Oldenbourg Industrieverlag, München, 2008.
Appendix
MP is voltage dependency of motor load where value 2 is constant impedance
characteristic. KUSOGI and KISOGI describe the filtering of line voltage and
154 Power Supply, Energy Management and Catenary Problems
current, respectively. TIPLL and KPPLL are the integration time and gain of the
PLL, resp. TIVC and KPVC are the integration time and gain for the DClink
voltage controller, resp. TICC and KPCC are the integration time and gain for
the current controller, resp. TUDC and TIDC are the filter time constants for the
DC voltage and current measurements, resp. C is the DClink capacitor in Farad.
Power Supply System Analysis, Design and Planning 155
This page intentionally left blank
Fast estimation of aggregated results of many
load ﬂow solutions in electric traction systems
L. Abrahamsson & L. S¨ oder
Electric Power Systems, Royal Institute of Technology (KTH), Sweden
Abstract
Transports on rail are increasing and major railway infrastructure investments are
expected. An important part of this infrastructure is the railway power supply sys
tem. The future railway power demands are naturally not known for certain. This
means investment planning for an uncertain future. The more remote the uncertain
future, the greater the amount of scenarios that have to be considered. Large num
bers of scenarios make time demanding (some tens of minutes, each) simulations
less attractive and simpliﬁcations more so. The aim of this paper is to present a fast
approximator that uses aggregated traction system information as inputs and out
puts. This facilitates studies of many future railway power system loading scenar
ios, combined with different power system conﬁgurations, for investment planning
analysis. Since the electrical and mechanical relations governing an electric trac
tion system are quite intricate, an approximator based on neural networks (NN), is
applied. This paper presents a design suggestion for a NN estimating power system
caused limits on active and reactive power load, i.e., limits on the levels of train
trafﬁc.
Keywords: railway, neural networks, load ﬂow.
1 Introduction
For environmental and economic reasons, in Sweden and the rest of Europe, both
personal and goods transports on rail are increasing. Therefore major railway
infrastructure investments are expected. An important part of this infrastructure
is the railway power supply system.
One phase low frequency AC railway power supply systems are normally con
nected to the normal 50 Hz power system through frequency converters, see
Figure 1.
Power Supply System Analysis, Design and Planning 157
Z
1
Z
3
Z
2
S
1
50 Hz
S
2
16.7 Hz
3 phase
1 phase
Converter
50 Hz
16.7 Hz
3 phase
1 phase
Converter
Z Z
Previous power
system section
Next power
system section
This power
system section
Public grid
Railway grid
Figure 1: A section of the railway power supply system, illustrated as an electric
circuit.
The converters may be merely used as power sources, as in Sweden for example,
whereas in other countries the railway administrations produce electricity on their
own and sell it on the market when proﬁtable. The catenary (the overhead con
tact line from which the locomotive extract electric power) has a relatively high
impedance, i.e., the railway power grid normally can be considered weak. The
weakness can be somewhat reduced by strengthening the catenary system, plac
ing the converter stations closer to each other, or connecting a High Voltage (HV)
transmission line in parallel to the catenary system. One important consequence
of a weak railway power supply system is that when trafﬁc is dense, fast, and/or
the trains are heavy, the voltage will drop. For voltages lower than nominal, the
locomotives cannot extract as much tractive force as is normally the case. This
means that, depending on the strength of the power system, there is a limit on the
intensity of the train trafﬁc, caused by voltage drops in the catenary that reduces
the maximal tractive force.
Railway power supply systems are changing all the time. When a train moves,
the impedances between it and the feeding points changes, because both catenary
and HV line impedances are distance dependent. The magnitude of the apparent
power demands also change when the trains move. Both the active and reactive
power demands of the locomotives may vary with the slopes of the railway, the
train weights, the desired train velocity, etc.
When considering future possible railway power supply system investments, for
each possible expansion alternative, many different situations of railway operation
causing different loading on the railway power supply system have to be stud
ied. This is a kind of transmission expansion planning. Similar approaches can be
found in the references [1–3]. In the case of the railway, however, the locations of
feeding points are up to the railway power grid administrator, and not to the actors
on the market as in the case with the public grid.
For each possible investment alternative, the power grid calculations have to be
performed fast. There are several algorithms available, e.g., TPSS (Train Power
System Simulator, called TTS in [4]) and commercial programs such as TracFeed
Simulation [5]. These are however not fast enough, though, when several thou
sands of cases have to be studied. The intention of this paper is to propose a solu
158 Power Supply, Energy Management and Catenary Problems
tion of how to design an approximator that rapidly estimates the properties of an
electric railway power supply system which are considered important in the stud
ies of future expansion alternatives. In other words, to calculate aggregated results
from many load ﬂow calculations in a fast way.
The two most important consequences of train trafﬁc related to the state of the
power system, having the railway operation costs in mind, are:
• The power consumption of the grid, preferably divided up by individual
converter stations.
• The impact of catenary voltages on the minimal traveling times.
In this paper, the power system impact on travel time limits is studied, and there
fore these are modeled to be the outputs of the approximator. In order to keep
the approximator relatively simple, inputs as well as outputs must contain aggre
gated information. These aggregated ﬁgures can, from the power system point of
view, be divided into three groups: issues having inﬂuence on the power consumed
(inputs), issues having inﬂuence on the impedances (inputs), and issues related to
voltage drops (outputs). For more details, see sections 3.2.1 and 3.2.2.
Since the electrical and mechanical relations of an electric traction system are
quite intricate, an approximator of the black box kind is a tempting approach.
Therefore, Neural Networks (NN), which basically are a kind of nonlinear pre
dictors, were selected as a suitable approximator type. One reason to use aggre
gated inputs is that NNs with too many inputs and outputs may need a tremendous
amount of training cases in order to become reliable and general [6]. The exact
details of the approximation can be studied in section 3.
In a numerical example, an X km section of a traction power supply system is
modeled, see section 3.2. The operation is then simulated and the parameters and
the performance of the approximator are then estimated. The notation X signiﬁes
that the interconverter distance is varied.
1.1 Related work
Efforts constructing less complicated approximators, based on knowledge, intu
ition, and experience have been made in [7], an interesting German paper from
1998. That paper presents an obviously very fast and uncomplicated method of
approximating impacts of different levels of trafﬁc and different power system
setups, on the train speeds. The models in [7] do, however not consider the impacts
of: slopes, nonmaximal train velocities, train velocity and catenary voltage depen
dencies of the reactive power consumption (which implicitly is done in the model
proposed here, from now on called TPSA (Train Power System Approximator)
since the approximator is trained with simulation results based on load ﬂow calcu
lations), different train types at the same railway section, etc.
2 Basic power system modeling
2.1 Converters
Traction systems fed through converters, are normally either low frequency one
phase AC systems, or DC systems. A section of the low frequency railway power
Power Supply System Analysis, Design and Planning 159
supply systems used in Northern Europe can be modeled as in Figure 1. Because
of the relative weakness of such systems, converters can more or less be consid
ered as power injectors from strong grids. Thereby, converter stations need to be
distributed along the railway such that power does not have to be transmitted over
too great distances, and such that the converters do not get overloaded.
The frequency converter stations in Sweden control their terminal voltages,
although not stifﬂy. The voltage is controlled linearly with respect to reactive
power generation. Converter stations can produce reactive power of both posi
tive and negative signs. If no reactive power is generated, the catenary voltage is
controlled at a nominal level (16.5 kV). The voltage angles at the converter station
terminals are completely determined (although quite nonlinearly) by the voltage
levels, and the active and reactive power ﬂows in and out of the terminals of the
converter station.
In Figure 1, the power system sections are fed from both ends. In reality, how
ever, a railway power system section can also be supplied only from one side.
Singly fed sections are not studied in this paper, but can for example occur at cate
nary ends, or in cases of converter outages.
2.2 Power transmission
The impedances Z
j
, j ∈ {1, 2, 3}, in Figure 1, denote the total impedances as
seen by the locomotives. Included in these impedances one can ﬁnd catenary
impedances as well as (if present) impedances originating from the high voltage
lines and the transformers connecting them to the catenary.
The impedance values do not only depend on the train positions (impedances are
normally length dependent), they also depend on the power system conﬁguration.
Basically, there are two kinds of catenary technologies used: Booster Transformer
(BT) and Auto Transformer (AT). BT is the cheaper one, with higher impedances
than AT [8]. There is also an option of strengthening the power systemby installing
a High Voltage (HV) transmission line in parallel to the catenary. The impedances
between the locomotives and the converter stations are naturally smaller, the more
densely the transformers connecting catenary and HV line are distributed.
2.3 Locomotives and trains
For a train to move forward, a tractive force is needed, depending on its running
resistance (mainly air and mechanical resistance), its weight, the locomotive’s efﬁ
ciency, and track topography, different force levels are needed for acceleration or
speed maintenance. Moreover, depending on the train velocity, the catenary volt
age levels, and possible additional controls, the tractive force outtakes are bounded.
The following is important, especially from the power system dimensioning
viewpoint. For sinking catenary voltage levels, for a constant amount of consumed
power, the currents tend to increase. Therefore the tractive force is successively
reduced by onlocomotive controllers. This control is installed in order to pro
tect the engine from overloading. The connection between tractive force, F, and
160 Power Supply, Energy Management and Catenary Problems
the electric active power consumed, P, is, P = F · v, excluding electrical and
mechanical losses in the train, and where v denotes train velocity. Thus, in weaker
traction power systems, if the trafﬁc demand is too high, the trains will be forced
to travel either more sparsely, more slowly, or a combination of the two, com
pared to what would be the case if the power system had an unlimited capacity. An
important difference between locomotives and other kinds of power consumers is
that locomotives to a greater extent accept substantial voltage drops. Thus, even a
typical Swedish locomotive will still be in operation, at a reduced level of tractive
force, for voltage drops of about 40 %. The electrical and mechanical rules of train
movements are described in some detail in [4, 9].
Summarizing the preceding paragraph, trains are movable and heavily varying
electric power loads. In the exemplifying circuit of Figure 1, the apparent power
consumed by the two trains S
i
= P
i
+ jQ
i
, i ∈ {1, 2} can be found. As indicated
in the above, the apparent power is not constant with respect to time and position.
For some (older) locomotive types, the reactive power consumption, Q
i
, depends
on catenary voltage levels, P
i
, and v, see e.g. [4, 9] for details. For more modern
locomotive types, Q
i
can be controlled freely but such locomotives are, at least in
Sweden, normally operated purely resistively [10].
3 The approximator
3.1 Introduction
The aim of the approximator (TPSA), is to obtain a fast estimation of whether
a certain power system causes limitations in possible train power consumption
levels, i.e., in possible train trafﬁc.
Since the voltages are almost stiff at the catenary side of the converter station
terminals, the railway power supply system can be divided into separate sections
by using the converters as section borders. These sections are almost indepen
dent of each other. This means that train trafﬁc on one side of a converter station
inﬂuences the catenary voltage levels on the other side of that particular converter
station to a comparatively small extent. The power ﬂows between these power sys
tem sections are normally insigniﬁcant, but the smaller the impedances, the more
noticeable they are.
In [11], one modeling simpliﬁcation is that a train consumes power only from
the feeder units right in front of and right behind it. There, ”feeder unit” means
either a converter station or a transformer station connecting the catenary to the
transmission line.
With this assumption, the inputs to TPSA, as well as the outputs from it, only
have to describe the trafﬁc and power system states for one section at the time.
The power system sectioning assumption is an essential part of the TPSA model
presented here. The fact that the TPSS simulations for the creation of the TPSA
training set will treat smaller power systems, and thereby further reduce the simu
lation times, must simply be considered as a bonus. The main beneﬁt considering
future usage of TPSA is that it will now become a general module that can be
Power Supply System Analysis, Design and Planning 161
used for any traction power system section in a greater system. The part of TPSA
that determines the maximal onaverage velocities can be used as trafﬁcdependent
constraints in a suitable time table program. Such a constraint states the maximal
onaverage velocity for an additional train, given an already existing trafﬁc.
The main difference between the TPSA model presented in this paper, and the
so called TTSA in [4] is that inputs and outputs are aggregated differently. In this
paper the choices of aggregation are better adjusted for the kind of information that
one can expect to be available. In [4], the active power consumption of the railway
was also estimated. Since this paper is more detailed, all focus is here set on the
impacts of minimal train running times as a function of the strength and the load
of the power supply system.
3.2 The model
It is common in train time table planning programs to use models containing binary
or integer variables for train location. The models vary, but are similar. From now
on, assume we have a binary variable, say x
i,j,t
i
,t
j
,r
, that takes on the value 1 when
there is trafﬁc on that particular railway section in that particular time windowwith
that particular train, and equals 0 when not. The indices designate the following:
i is the node (a node in a trafﬁc planning model, i.e., a train stop) from which the
train leaves at time t
i
, j is the node at which the train will arrive at time t
j
, whereas
r is the train index.
The facts explained above regarding train timetable planning are crucial for the
modeling of TPSA, i.e., the choices of the aggregated inputs and outputs for the
TPSA. The inputs and outputs of TPSA must be such that they are possible to
aggregate both from TPSS as well as time tabling programs.
The points in time when the studied (additional) train starts, t
i
, and stops, t
j
, are
crucial when aggregating the inputs and outputs. For example they make it possible
to know when to start measuring the behavior of the other trains (connected to the
same power system section as the studied train during the time interval between t
i
and t
j
). Obviously, all trains are limited by the catenary voltage levels, but TPSA
is designed such that each train is studied at a time. Separate TPSA:s for different
numbers of trains on the power system sections would make the approximator less
general and probably demand a greater set of training data.
The choices of important aggregated TPSA inputs and outputs are in the bullet
lists of sections 3.2.1 and 3.2.2 described and motivated. The binary variables AT
and HV describing the power system technology are not aggregated, and as shown
in Figure 2 even not really inputs to the NN, and therefore not found in the bullet
list of section 3.2.1.
3.2.1 The aggregation of inputs
The TPSA inputs can be classiﬁed as either parameters giving rise to the con
sumption of active and reactive power, or parameters related to the power system
impedances. Some inputs may be connected to both consumption and impedance.
In this paper, the trains are allowed to stop only at the locations of converter sta
162 Power Supply, Energy Management and Catenary Problems
tions, a fact that reduces the number of input variables. Additionally, the model is
restricted to one type of trains traveling in one direction. The TPSA inputs of this
paper can be found in the following bullet list, including short motivations.
• The distance between i and j, denote it d(i, j). The distance gives informa
tion of the power system impedances.
• The average inclination, E(incl), between the power system sectionborders
i and j. The average inclination gives information of the net potential energy
consumed by the trains. The other trains will not experience the same incli
nations as the train studied in the time interval between t
i
and t
j
does.
• The standard deviation of the inclination, D(incl), between the power sys
tem sectionborders. The average inclination, E(incl), would for example
equal zero for both ﬂat ground as well as for a rail section with 20 per
mille uphill half the section and 20 per mille downhill the remaining half.
The standard deviation is a measure of how much the inclinations ﬂuctuate,
which will inﬂuence the consumed electric power of the trains.
• The trafﬁc of other trains on the studied power system section in the time
interval between t
i
and t
j
must also be measured, because within that time
interval, the power consumed by “other trains” affects the catenary voltages
experienced by the “studied” train. The average velocity, E(v), and the aver
age number of trains, E(t rains), on the section are calculated. The velocity
is used because higher train speeds means greater power consumption. The
average velocity is measured only when trains are in service. The number of
trains is important because the more trains, the greater need for electricity.
The average number of trains is calculated as total train trafﬁc time divided
by the length of the time window, t
j
− t
i
.
3.2.2 The aggregation of outputs
The TPSAoutput describes the running times inﬂuenced by catenary voltage levels
for the studied train.
• The fastest possible average velocity of the studied (added) train.
3.2.3 Neural network
The output of TPSA is the greatest possible average speed for an added train. The
inputs of TPSA describe the power system topology and parameters, the topogra
phy of the track, and the number of other trains (on the same power system section
as the studied train) as well their average speeds. TPSA is essentially a NN that
connects the inputs to the outputs.
Moreover, some of the TPSA inputs describing the power system are binary,
i.e., AT, which equals 1 for AT catenaries and equals 0 for BT catenaries, and
HV, which indicates the presence of HV transmission lines when it equals 1. Two
binary variables give four possible combinations. For each of the four cases, an
individual NN is constructed, see Figure 2. The separation into four different NNs
is motivated by the fact that backpropagation [12, 13] networks are not optimal for
such classiﬁcations [6]. Backpropagation networks are mainly approximators of
smooth functions. Thus, the NNs are only trained with inputs and outputs that can
Power Supply System Analysis, Design and Planning 163
TPSA
inputs
AT
HV
Remaining
inputs
Neural
Network #2
Neural
Network #1
Neural
Network #3
Neural
Network #4
Selector
TPSA
TPSA
outputs
Figure 2: For AT and BT catenary types, who can be either connected to or with
out an HV transmission line, there are four separate possible neural net
works.
f
f
f
f
input
g
g
g
h
h
h
h
output
hidden
layer
hidden
layer
output
layer
Figure 3: A sketch of an arbitrary neural network. The input vector is three ele
ments long, whereas the output vector is four elements long.
be regarded as continuous variables, i.e., “remaining inputs” in Figure 2.
Two different NN models are proposed, both based on the assumed outputs and
inputs presented in sections 3.2.1 and 3.2.2. An example of a multilayered neural
network can be found illustrated in Figure 3. In that network, there are three layers,
of which two are hidden. Normally, all neurons in one layer have the same type of
transfer functions, e.g., f in the ﬁrst layer in Figure 3.
The ﬁrst model, M1, is a nonlinear NN with two layers. In M1, the hidden layer
has tanh transfer functions,
out
1
n
= tanh
b
1
n
+
5
k=1
in
k
w
1
k,n
(1)
where subscript n denotes neuron number and superscript 1 denotes the ﬁrst layer.
The input vector is, as can be deduced from the bullet list, of length 5 and the
output is just a scalar. The second (output) layer has a linear transfer function
out
2
= b
2
+
N
n=1
out
1
n
w
2
n
(2)
where superscript 2 denotes the second layer, and N is the number of neurons in
the ﬁrst layer. The parameters b
1
n
, b
2
, w
1
k,n
, and w
2
n
are parameters whose values
164 Power Supply, Energy Management and Catenary Problems
are determined in the training step. According to the theory, page 78 in [14], and
in [13], this kind of network can be used as a general function approximator, given
a sufﬁcient amount of neurons in the hidden layer.
The second model, M2, assumes a linear dependency
out = b +
5
k=1
in
k
w
k
(3)
where k is input index, and w
k
and b are parameters whose values are determined
in the training process presented in section 3.2.3.1.
The motivation for testing a linear model at all is explained by the planned future
use of TPSA as a running time constraint in a train time tabling program. In such
a program, time consuming already as it is, a linear model would probably not
extend the computation times as much as a nonlinear one would. The accuracies
of the two models are compared and evaluated in section 4.2.
3.2.3.1 Training of the Neural Network In both models, i.e., M1 and M2, the
aggregated input and output data sets are normalized to lie in the interval [−1, 1]
before training and testing the approximators.
Training is essentially a method of determining the values of the parameters w
and b (in both M1 and M2), such that the mean square of the estimation error of
the network output is minimized.
Model M1 is trained by the trainbr (Bayesian regularization backpropaga
tion) algorithm (of Matlab’s NN Toolbox) with an error goal of 10
−5
. Thus, out
2
of equation 2 is the estimation of the maximal train velocity.
Model M2 is trained by the trainb (batch learning) algorithm (of Matlab’s
NN Toolbox). Thus, out of equation 3 is the estimation of the maximal train veloc
ity.
4 Numerical example
4.1 Problem setup
In order to be as clear a possible, in the numerical example presented in this paper,
just a few of all possible parameters are presented and used. That makes the prob
lem setup more surveyable. For example, only one kind of trains is studied (simi
larly as in [4]), only one kind of each respective catenary type is studied, unidirec
tional trafﬁc is assumed, and the converter stations are all assumed to be equipped
with six 10 MVA converters (Q48/Q49) each such that the installed power should
not be a limiting factor in the simulations studied. Furthermore, the train stops are
also assumed to coincide with the locations of the power system section borders,
i.e., the converter stations. The speed limit is set as high as 150 km/h all over the
simulated railway sections, i.e., for Rc4 locomotives a limiting constraint only in
very steep downhill situations. Moreover, the number of transformers connecting
Power Supply System Analysis, Design and Planning 165
0 50 100 150
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
km
v
U
P
D
Q
D
incl
incl = 0
− θ
Figure 4: An example of the mechanical and electrical states of a train traveling
through a section of the railway electric power supply system.
the transmission line (when present) to the catenary is set to three in all cases stud
ied in this paper. The model presented in section 3.2 is as simple as it is in order to
comply with these model restrictions. All trains also try to go as fast as possible in
the TPSS simulations, excluding a lot of reallife cases.
In Figure 4, a train from a TPSS simulation is studied. In that simulation, the
converter distance was 160 km on an AT system with HV line and the train depar
ture periodicity was 6 minutes. The velocity of the train, v, is scaled down by a
factor of 160 km/h. Additionally, the ﬁgure presents the consumed active power,
P
D
, reactive power, Q
D
, and the catenary voltage, U, all expressed in p.u., where
S
b
= 5 MVA and U
b
= 16.5 kV. The catenary phase angles of opposite sign, i.e,
−θ, seen by the locomotive of the train can also be found in Figure 4, expressed in
radians. Finally, the inclinations of the railway track are also plotted in Figure 4,
scaled such that 0 on the yaxis means −10 per mille, and 1 means 10 per mille.
In the numerical example, 400 different simulations have been done, 100 for
each neural network to train. There are four separate NNs, one for BT catenaries,
one for AT catenaries, one for BT with transmission line, and ﬁnally one for AT
with transmission line – as described in Figure 2.
These hundred cases consists of ten different inter converter distances (of length:
30, 57, 80, 98, 114, 127, 138, 146, 154, and 160 km) combined with ten different
train departure periodicities (of times: 6, 8, 10, 13, 17, 22, 28, 36, 46, and 60
minutes between each train departure).
In the numerical example of [4], the measuring of a train’s running time is not
started until the railway is ﬁlled up with trains according to a certain train departure
periodicity. In this study, in order to speed up the calculations, the measuring is
started immediately. In order not to have a completely empty railway when starting
measuring, in the initialization procedure, TPSS distributes trains along the track
almost as in an ideal situation, i.e., trains going at maximally allowed speed all
166 Power Supply, Energy Management and Catenary Problems
1 2 3 4 5 6 7 8 9 10
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0.11
Number of neurons in hidden layer
A
v
e
r
a
g
e
M
e
a
n
S
q
u
a
r
e
E
r
r
o
r
Training Set
Test Set
Figure 5: A study of the average mean square errors on the training and testing sets
as a function of the amount of hidden neurons.
the time. After the simulation has started, however, of course the trains distributed
along the track with predeﬁned initial velocities will followthe same electrical and
mechanical laws of nature as the other trains.
4.2 Numerical results
The greatest onaverage velocity, for an added train, is determined through TPSS
simulations with the problem setup presented in section 4.1.
Worth pointing out is that all simulations use the same railway proﬁle, e.g.,
the ﬁrst 30 km in all simulated cases have the same inclinations. Consequently, the
average values and standard deviations of the inclinations will be different for each
interconverter distance.
As rule of thumb [6], two thirds of the simulations may be used for training
the NN, whereas the remaining third may be used for testing. That is applied also
here. The amount of hidden neurons, i.e., N in equation 2 was set to 4. This was
not done arbitrarily, but by studying the average (ten different random choices of
training and testing sets) mean square approximation errors for one to ten hidden
neurons – see Figure 5.
In the case studies, the accuracies of one linear NN and one nonlinear NN are
compared for both the training and the testing sets. The approximation errors in
Table 1 can be found for network M1, and Table 1 for network M2. The NNs are
modeled with the design introduced in section 3. The computer time needed for
execution of minor NNs like these is negligible.
Power Supply System Analysis, Design and Planning 167
Table 1: The approximation errors for the four different neural networks.
(a) For model M1
M1 Training Set
HV BT AT
Yes 7.22 · 10
−3
8.05 · 10
−3
No 1.04 · 10
−2
9.05 · 10
−3
Testing Set
Yes 1.14 · 10
−2
2.39 · 10
−2
No 3.03 · 10
−2
1.14 · 10
−2
(b) For model M2
M2 Training Set
HV BT AT
Yes 5.67 · 10
−2
6.04 · 10
−2
No 7.93 · 10
−2
6.24 · 10
−2
Testing Set
Yes 6.52 · 10
−2
8.80 · 10
−2
No 1.05 · 10
−1
7.46 · 10
−2
5 Conclusion and summary
The paper can be concluded with the fact that a ﬁrst suggestion to a fast method
of estimating the fastest possible train speeds, for an additional train, as a function
of the railway power supply system and the trafﬁc level of other already added
trains. The function uses aggregated parametric values as both inputs and outputs.
In the numerical example, an example of the results of TPSS were presented in a
graph. That is followed up with a comparison between the two suggested approx
imator models M1 and M2. As one could expect, the nonlinear model performs
noticeably better.
Some recommendations for further work:
• Include different traveling directions in the model
• Include trains with different speed limits
• Allow train stops at arbitrary locations
• Include different train types
• The next step of TPSA: estimate the injected active power, and the dividing
up between the converter stations.
References
[1] CruzRodrgues, R.D. & LatorreBayona, G., HIPER: Interactive Tool for
MidTerm Transmission Planning in a Deregulated Environment. Power
Engineering Review, IEEE, 20(11), pp. 61–62, 2000.
[2] Buygi, M.O., Balzer, G., Shanechi, H.M. &Shahidehpour, M., MarketBased
Transmission Expansion Planning. Power Systems, IEEE Transactions on,
19(4), pp. 2060–2067, 2004.
[3] Chennapragada, B., Radhakrishna, C. & Vallampati, R., Riskbased approach
for Transmission Expansion Planning in Deregulated Environment. Prob
abilistic Methods Applied to Power Systems, 2006. PMAPS 2006. Interna
tional Conference on, Stockholm, Sweden, 1–4.
[4] Abrahamsson, L. & S¨ oder, L., Fast Calculation of the Dimensioning Factors
of the Railway Power Supply System. Computational Methods and Experi
168 Power Supply, Energy Management and Catenary Problems
mental Measurements, Prague, The Czech Republic, volume XIII, pp. 85–96,
2007.
[5] Stern, J., TracFeed Simulation, Reference Manual. Technical Report
BBSE951112BNA, Balfour Beatty Rail, 2006.
[6] Communication with Bj ¨ orn Levin, SICS  Swedish Institute of Computer
Science, 2007. Kista, Sweden.
[7] v Lingen, J. & Schmidt, P., Railway Electric Energy Supply and Operation of
Railways (original title in German). Elektrische Bahnen, 96(1–2), pp. 15–23,
1998.
[8] B¨ ulund, A., Deutschmann, P. & Lindahl, B., Circuit design of the Swedish
railway Banverket in catenary network (original title in German). Elektrische
Bahnen, 102(4), pp. 184–194, 2004.
[9] Abrahamsson, L. & S¨ oder, L., Basic Modeling for Electric Traction Systems
under Uncertainty. Universities Power Engineering Conference, 2006. UPEC
’06. Proceedings of the 41st International, Newcastle upon Tyne, UK, vol
ume 1, pp. 252–256, 2006.
[10] Communication with A. B¨ ulund, Swedish national railway administration
(Banverket), 2007. Borlnge, Sweden.
[11] Nyman, A., TTS/SIMON Power Log – a simulation tool for evaluating elec
trical train power supply systems. Computers in Railways VI, Lisbon, Portu
gal, 1998.
[12] Backpropagation. http://en.wikipedia.org/wiki/Backpropagation, 2007.
[13] Neural Network Toolbox (Matlab online help). www.mathworks.com/
access/helpdesk/help/toolbox/nnet/backpro4.html. Retrieved on March 9
2007.
[14] Gurney, K., An Introduction to Neural Networks. CRC Press, 2003.
Power Supply System Analysis, Design and Planning 169
This page intentionally left blank
DC protection calculations –
an innovative approach
R. Leach, D. Tregay & M. Berova
Parsons Brinckerhoff Ltd, UK
Abstract
As part of a recent major resignalling scheme for a DC electrified railway,
substantial changes were made to the track layout and electrical feeding/switching
arrangements. In addition, as part of this scheme, the opportunity was taken to
introduce a novel approach to the way negative bonding in the return circuit is
configured, made possible by the introduction of axle counters, in place of track
circuits, for train detection purposes. As a consequence, the overall impact on
electrical impedance was unclear. The need arose to determine firstly whether the
existing protection settings for the track feeder circuit breakers remained safe in
the light of these changes and secondly whether the settings could be improved,
thereby offering the potential to increase services and/or run longer trains.
This paper outlines the challenges that needed to be overcome in gathering,
processing and validating input data to arrive at a complete, coherent and
consistent set of data necessary for determining the maximum fault impedances
seen by each of the circuit breakers. The paper goes on to present the
development, implementation and application of the methodology for the
calculation of the protection settings, including validation of the results.
The results from the calculations were surprising and contrary to original
expectations.
Keywords: calculations, traction, power, DC, protection, relay, settings,
spreadsheet, fault, impedance, overcurrent, overhang, tee, data, hand,
modelling, schematic, equivalent and circuit.
1 Introduction
As the result of a major resignalling project in the UK, the permanent way
layout and the third rail DC traction power supply infrastructure underwent
substantial change, both in terms of the track layout, particularly at switch
Power Supply System Analysis, Design and Planning 171
and crossing (S&C) locations, and in the plain line sections of the rail
network. The installation of new types of running rails and the renewal of
the traction return (negative bonding cables) based on a new technique [1]
meant that existing circuit impedances needed to be revisited to determine
whether existing DC circuit breaker protection settings remained safe, or
needed to be modified in the light of the changes. This necessitated carrying
out protection setting calculations for 36 electrical sections, some standard
plain line and others, involving diverging routes, resulting in complex
feeding arrangements.
2 Issues relating to data
The initial approach was to update the existing calculations, relevant to the
affected electrical sections, to reflect any changes in the infrastructure.
However, when compiling the data, it was found that before proceeding with any
analysis there were a number of issues to resolve, namely:
Record drawings often up to 20 years out of date or incomplete
Recent changes arising from an earlier Traction Power Supply Upgrade
project needed to be included
Feeder cable lengths not typically included in earlier calculations where
impedance values of some such feeders were actually quite significant
Variety of protection relay and circuit breaker current tripping device
types needed consideration
Recent drawings based on metric units whereas original calculations
utilised imperial units (miles and feet) thereby warranting conversion
Mileages shown on positive (conductor rail) drawings and negative
bonding drawings not always in agreement
Details of running rail types not readily available, or unknown
Different running rail types in parallel within electrical sections
Multiple reference points e.g. record drawings based on miles and feet
from London, permanent way drawings at switch and crossing based on
local datum point, signalling drawings based on separate datum relating
to extent of resignalling project
Changes to conductor rail crosssections, feeder cables and return
circuits arising from changes to permanentway track layouts
3 Selection of approach
3.1 Previous practice/issues
The existing protection setting calculations supplied by the client relating to
previous stages of the traction power system development were based on simple
hand calculations, which typically excluded cable lengths and tended to assume
connection points were adjacent to the Traction Substations (TSS’s) and Track
Paralleling Huts (TPH’s). This presented several issues to overcome:
172 Power Supply, Energy Management and Catenary Problems
Hand calculations cannot readily cope with complex layouts, or
multiple changes in conductor rail and running rail types.
With hand calculations it is easy to make mistakes and difficult to check
the calculations.
Often the lengths of junctions are now typically much longer. This
meant that physical connection points to conductor rails and
feeder/return cables were rarely adjacent, but often a considerable
distance from the TSS or TPH resulting in greater cable lengths.
Therefore, it became apparent that cable impedances could not be
ignored.
Significant ‘Overhangs’ or ‘Tee’ feeds arising from extended junctions
were omitted from the original calculations.
Historically, conductor rails were of smaller cross section, hence
impedance was very similar to the feeder cables and actual connection
point was not considered particularly critical.
3.2 Typical arrangements
Figures 1  3 below illustrate the variety of configurations typically found on
most schemes of this kind and associated fault paths.
Note: A ‘Tee’ feed is particularly difficult to evaluate in terms of providing
suitable protection settings. This arises as a fault in the ‘Tee’ or ‘overhang’
results in fault current from the two feeding circuit breakers flowing along the
same path, thereby increasing the apparent impedance seen by the circuit
breakers. It is therefore necessary to determine whether a fault at the end of the
‘Tee’, or a fault at the remote end of the section represents the higher
S/S 2 S/S 1
¨
.
¨
±
Fault
Figure 1: Normal feeding (no ‘Tee’ feed or ‘Overhang’).
S/S 2 S/S 1
Fault
¨
¯·
¨
¯`
¨
¯W·¯`³
Figure 2: Branch in section forms a ‘Tee’ feed.
Power Supply System Analysis, Design and Planning 173
S/S 2 S/S 1
Fault
¨
¯
Figure 3: ‘Overhang’ with positive connections remote from the substation.
impedance, in order to determine the required settings (see Figs. 2 and 3). The
risk is that neither circuit breaker may trip under a fault condition. An
‘overhang’ is the extreme form of a ‘Tee’ feed in electrical circuit terms.
3.3 Approach adopted
As a consequence of the above, it was decided that hand calculations were not
appropriate for this project. Instead, the process was undertaken through use of a
calculation spreadsheet (PBProCalc) with data inputted via a customised front
end template, incorporating the relevant formulae contained within Sections 8.3
& 8.7 of the client’s process document [3]. The calculations were also modified
to take account of the client’s specific requirements within the guidance
document [4].
The main changes from the client’s process document are clearly defined in
his guidance note [4]. The designer is instructed to exclude any allowance for
rail joints and bonding resistances. A 7.5% tolerance is to be added to the
settings to compensate for impedance relay tolerances. Any cables over 15m in
length are to be separately identified and included in the calculations.
In order to validate the accuracy of the spreadsheet, a full check of the
formulae within the spreadsheet was carried out using an independent Engineer
within Parsons Brinckerhoff (PB). In addition, the spreadsheet was tested
against one of the worked examples in the clients process document [3] and
against the example ‘Tee’ feed calculation results contained in the clients
Guidance document [4]. As a further means of verifying the accuracy of the
spreadsheet, a second independent set of calculations were undertaken using
Network Rail ‘Tee’feed computer program [5] for all of the electrical sections
on the project.
The Proposed Approach was documented [6] and submitted to the client, and
formal acceptance received prior to implementation.
3.4 Modelling
An equivalent circuit was created for each electrical section based on
information derived from design drawings and record drawings (see section
headed ‘Detailed Process’ below and the Flow Chart in Appendix A for further
details).
174 Power Supply, Energy Management and Catenary Problems
In order to simplify the calculation process, and to arrive at worst case, the
following rules were applied:
Benefits of reinforcement cabling for running rails around S&C
locations and at single rail track circuited areas were excluded from
calculations as worst case, on assumption these could be disconnected
during maintenance, etc. and therefore not available for traction return
Running rails were assumed to be 109lb/yd (smallest cross section)
throughout as per original, unless confirmed otherwise
Conductor rail also assumed to have the smallest cross section
(100lb/yd), unless confirmed otherwise
In order to simplify calculations in multitrack areas, where different
types of running rail are in parallel, a particular rail size was selected
and the length adjusted in the spreadsheet to provide an equivalent
impedance.
Cable lengths to substations and TP Huts were typically scaled from the
conductor rail and negative layout drawings as exact cable lengths were
not typically available
‘Tee’ points and ‘Overhangs’ were worked out separately for positive
and negative circuits, making the calculations simpler to handle. This
produced exactly the same result as if the elements were combined.
3.5 Fault path
The preparation of equivalent circuits for the positive side (conductor rails) was
relatively straightforward, although determination of actual connection points was
in some cases difficult due to imprecise record data for some areas. The situation
with regard to the negative return circuits was, however, often quite complex.
In some instances it was far from clear which equivalent circuit should be
used for the return circuit, in particular relating to multitrack areas with
substation return connections at differing points for each track (sometimes
hundreds of metres apart) and potentially involving an ‘overhang’ situation. The
dilemma was how to determine the position/length of the ‘overhang’ in relation
to the equivalent circuit for the remainder of the main line in terms of whether a
single point for all the return connections should be assumed, or each track
worked out separately. If a single point of connection is assumed, the issue
arises of where it should be placed: at the electrical midpoint for instance; at the
closest connection point; or at the furthest point from the substation. All of these
options then impact on the length of the assumed ‘overhang’, so in the end it was
decided that all of these options should be tested and the worst case taken.
By way of illustration, figures 4 & 5 below show two possible options for
calculating fault path, with a fault occurring just beyond the substation return
connections. The diagrams illustrate a 4 track railway on the left converging to a
2 track railway on the right, with distributed return connections, and a ring of
cables around the junction linking all of the tracks together. In both cases the
assumption has been made that some of the running rails (and associated return
Power Supply System Analysis, Design and Planning 175
Next substation
Substation
Fault
Figure 4: Return connections at the left hand end.
Next substation
Substation
Fault
Figure 5: Return rail connections taken at the right hand end.
cabling) are unavailable for traction return (shown dashed), due to maintenance
activities for instance. The solid lines are assumed to be part of the fault path
and included in the calculations, whereas the dashed lines are excluded. The
overhang is shown in red for each of the options.
4 Detailed process
4.1 Process flow
The process of carrying out the protection setting calculations are discussed
below and presented in the Flow Chart contained in Appendix A.
4.2 Preparation of a single track schematic diagram
A single track schematic diagram (Fig 6) of the positive and negative circuits
was first prepared and marked up with all the information necessary for the
definition of the geometry, circuit elements (sections of rail(s), or cables) and
their dimensions and parameters, i.e. circuit references, boundaries, mileage of
important points, rail types, cable sizes, points of change of rail type and
identification of the circuit elements together with their respective lengths.
176 Power Supply, Energy Management and Catenary Problems
Positive Circuit

v
e
b
u
s
b
a
r
Negative (return) Circuit

v
e
b
u
s
b
a
r
UP MAIN
DOWN MAIN
STATION
E356
UP MAIN
DOWN MAIN
3
7
M
4
0
6
1
3
7
M
3
6
3
1
3
7
M
3
4
5
5
3
7
M
2
2
2
1
3
7
M
2
3
8
7
3
7
M
1
6
9
6
3
7
M
1
6
5
2
(a)
(b)
(c)
(d)
(f)
(j)
(g) 3
7
M
2
4
1
9
3
7
M
3
2
2
5
'
3
7
M
2
3
0
5
3
7
M
2
3
0
5
3
7
M
2
0
2
3
(n)
(l) (m) (o)
(p) (q)
(k)
3
7
M
4
1
1
4
3
7
M
1
3
9
5
3
7
M
4
0
1
2
3
7
M
3
9
7
1
3
7
M
1
6
8
9
3
7
M
1
4
4
6
(e)
3
7
M
1
7
9
5
3
7
M
2
0
2
3
Substation
STATION
Figure 6: Extract from a typical track schematic diagram of a traction power
supply system (the shades on the negative circuit indicating
different rail types).
L4
L3 NEG
()ve
L5
(+)ve
L1
L3 POS
L2
(+)ve
V
A
V
B
R
3NEG
R
3POS
R
1
R
4
R
2
R
5
Figure 7: Equivalent circuit.
The schematic diagram was then used to derive the equivalent electrical
circuit (Fig 7). The left hand diagram indicated the absence or presence of a
‘Tee’feed in the positive and/or negative circuits, lefthand or righthand ‘Tee’
feed, whilst the right hand diagram identified the equivalent circuit components,
used as input data to the calculation spreadsheet summarised in Table 1. For the
purpose of traceability and clarity the schematic diagram contained a list of all
the source drawings and documents from which the data has been extracted,
together with any assumptions made.
4.3 Calculation of fault impedances
The data from Table 1 is then input into the Calculation Spreadsheet (PB
ProCalc), the sheet automatically determining if a ‘Tee’ feed exists or otherwise
Power Supply System Analysis, Design and Planning 177
Table 1: Input data summary table.
Equiv.
Circuit
Comp.
Ref.
Type of circuit element
(cable/rail)
Length
(feet)
Total length
(feet)
(i) Cable type ‘z’ 75 75
(h) Conductor rail type 100lb/yd 6163 6163
(g) Conductor rail type 150lb/yd 889 889
(f) Conductor rail type 100lb/yd 806 806
(e) Conductor rail type 150lb/yd 396 396
(d) Conductor rail type 150lb/yd 228 228
(c) Cable type ‘z’ 100 100
(b) Conductor rail type 150lb/yd 79 79
(a) Cable type ‘z’ 500 500
(m) 2 x Running rail type CEN60E1 327 327
2 x Running rail type CEN60E1 7 7
2 x Running rail type BS113A 294 294
(m) 2 x Running rail type CEN60E1 327 327
2 x Running rail type BS113A 250 250
2 x Running rail type 109lb/yd 51 51
L5 (cab) (k) 4 x Cable type ‘a’ 150 150
282 282 L3 NEG
(l)
(l)
L3 POS
L5
(Up)
L5
(Down)
(n)
L1
L2
Conductor rail type 150lb/yd 282 282 (j)
2 x Running rail type CEN60E1
Equiv.
Circuit
Comp.
Ref.
Type of circuit element
(cable/rail)
Length
(feet)
Total length
(feet)
L4 (cab) (s) N/A
(r) 2 x Running rail type 109lb/yd 6216 6216
2 x Running rail type 109lb/yd 49 49
2 x Running rail type CEN60E1 340 340
2 x Running rail type BS113A 41
2 x Running rail type BS113A 176
(p) 2 x Running rail type BS113A 1068 1068
2 x Running rail type CEN60E1 82
2 x Running rail type CEN60E1 84
2 x Running rail type CEN60E1 41
2 x Running rail type CEN60E1 340
2 x Running rail type CEN60E1 176
(p) 2 x Running rail type BS113A 1068 1068
2 x Running rail type BS113A 82
2 x Running rail type BS113A 84
2 x Running rail type CEN60E1 198 198
2 x Running rail type 109lb/yd 6216 6216
2 x Running rail type CEN60E1 49
606
(r)
(q)
L4
(Down)
2 x Running rail type CEN60E1 198
(o)
(o)
L4
(Up)
166
217
364
(q)
and proceeds to compute the fault impedance seen by each circuit breaker
feeding each electrical section (A and B) of the two track railway (in the example
given). The result can be seen from Fig 8.
4.4 Derivation of protection settings
Impedance relays and Falling Voltage Overcurrent protection were the two types
of DC traction feeder protection that existed on the portion of infrastructure
involving this project.
178 Power Supply, Energy Management and Catenary Problems
SECTION LAYOUT (L
1,
L
2,
L
3POS,
L
3NEG,
L
4,
L
5
)
TEEFEED CALCULATION
SUBSTATION LIGHT LOAD VOLTAGE V
LL
(V)
69.71468077
2.423407145
RADIAL FEED CALCULATION (remote cb tripped)
MAXIMUM FAULT IMPEDANCE
ROUTE SETTING
MAX TRACTION LOAD CURRENT I
ROUTE
(A)
SUBSTN SOURCE IMPEDANCE MIN (R
S
Ohms)
SUBSTN LIGHT LOAD VOLTAGE MIN [VLL10%] (V)
'MAXIMUM SAFE' (halfkA from falling voltage curve table)
RECOMMENDED SETTING
CIRCUIT BREAKER B (kA) 6.5
0.0862535
IMPEDANCE CB B (Ohms) 0.0862535
Z
ROUTE
(Ohms) 0.1065
CIRCUIT BREAKER A (kA) 6.5
VE 'TEEFEED' LENGTH  L3
NEG
(miles) 0.053
I
B
/I
A
at Z
Tmax
K' MAX
I
B
/I
A
at K' MAX
Z
Tmax
(Ohms) 
MAX FAULT IMPEDANCE (B) VIA MAIN (Ohms)
CIRCUIT BREAKER A (kA) 7
CIRCUIT BREAKER B (kA) 7
0.012
711
IMPEDANCE CB A (Ohms) 0.0862535
6000
CROSSBONDING ALLOWANCE (Ohms) 0
MAX FAULT IMPEDANCE (A) VIA MAIN (Ohms) 0.0862535
MAX FAULT IMPEDANCE (B) VIA TEE (Ohms) 0.0111041
MAX FAULT IMPEDANCE (A) VIA TEE (Ohms) 0.0804262
Z
A
(Ohms) 
Z
B
(Ohms) 
I
B
/I
A
at K' MIN 69.714681
2.4234071
K' MIN 0.3403391
2.9382456
V
B
MINIMUM VALUE (V) 285
V
A
& V
B
MAXIMUM VALUE [V
LL
+6%] (V) 837.4
0.0380052
790
V
A
MINIMUM VALUE (V) 285
"EQUATION 23" VALUE 0.0076348
TeeFeed Calcs Required (Yes/No)? NO
TEEFEED CIRCUIT R3 [=R3
POS
+R3
NEG
] (Ohms) 0.0026384
VE SECTION LENGTH (B to 'T')  L5 (miles) 0.119
+VE 'TEEFEED' LENGTH  L3
POS
(miles) 0.053
+VE SECTION LENGTH (B to 'T')  L2 (miles) 0.058
VE SECTION LENGTH (A to 'T')  L4 (miles) 1.571
+VE SECTION LENGTH (A to 'T')  L1 (miles) 1.563
Figure 8: Extract from the protection setting calculations spreadsheet (PB
ProCalc) illustrating the calculation of fault impedances and
protection settings.
Power Supply System Analysis, Design and Planning 179
4.4.1 Impedance relays
For impedance type relays, 7.5% was added to the value derived in the
box headed ‘Maximum Fault Impedance’ to determine the actual figure
(‘Maximum Safe’) to be used for the settings.
Recommended Settings were then derived manually by rounding up to
the next 2.5mΏ, but in any event to be no more than 20% higher than
those given for ‘Maximum Safe’ setting to ensure proper discrimination
with the circuit breakers feeding the next electrical section.
In order to avoid unnecessarily changing existing settings, where the
existing setting is between ‘Maximum Safe’ and 20% above ‘Maximum
Safe’, then the recommendation has been to leave the setting unchanged
Additionally a Direct Acting Overcurrent Electromagnetic Trip is set at
a higher ’common’ setting to cater for close up faults.
4.4.2 Falling voltage overcurrent protection
In the case of Falling Voltage Overcurrent protection which is
principally an electromagnetic device mounted within and forming
part of the mechanism of a DC circuit breaker, the settings were
derived by converting the manufacturer’s standard protection curves
to an impedance table (Fig 9) showing settings (kA) against
impedance. The actual ‘Maximum Safe’ setting is determined by
taking the impedance value from the calculation spreadsheet i.e.
‘Maximum Fault Impedance’ and then reading off the required setting
in kA (rounded down to next kA value on the chart if the impedance is
mid way between settings)
4.5 Route setting
This is the setting required on a particular electrical section which will allow
train services to operate without resulting in ‘nuisance tripping’ and is
normally advised by the client. Any settings should normally be above this
value.
4.6 Summarising of results
The settings required for all of the electrical sections between adjacent
substations (or substation and TP Hut) were then summarised on A4 sheets, in
order that all the information relevant to those sections was available from one
sheet. In all cases ‘Category 1’ safe settings were achieved (based on worst case
data and feeding arrangements, hence not requiring any operational restrictions).
5 Further development
Since undertaking the original calculations, further work has taken place to
determine whether network modelling would help in determining worst case
fault scenarios in complex areas and possibly offer an alternative means of
180 Power Supply, Energy Management and Catenary Problems
S
e
t
t
i
n
g
(
A
)
10000
9500
9000
8500
8000
7500
7000
6500
6000
5500
5000
46 52 58 65.5 73 81 89 98 107 118 129
Impedance (milliΩ)
Figure 9: Chart of overcurrent settings versus impedance including half (kA)
settings (for circuit breaker type RJ R530L).
independently checking the spreadsheet results. Proprietary nodal analysis
modelling software (B2Spice) was used for this purpose, as it requires little
training and allows circuits to be built up easily from standard components. It
also allows components to incorporate formulae, such that the operator only has
to enter data in terms of along track length (feet, metres, miles, etc), rather than
having to work out each component separately.
Figure 10 above illustrates a typical circuit built using B2Spice, and has been
arranged so that it reflects the schematic diagram in fig 6, with the positive
circuit shown at the top and the return circuit (running rails) at the bottom.. Each
track has been represented separately (in this case a 2track railway), as it is quite
often the case that each track will have different rail types, hence all changes in
rail types, both positive and negative, can be fully incorporated into the model.
If required, crossbonding of running rails at regular intervals can be introduced,
although generally omitted in the case illustrated, other than at the tee point.
Once built, faults can be introduced where required and voltages at substations
‘A’ and ‘B’ adjusted to determine worst case fault conditions under radial or tee
feed conditions.
Power Supply System Analysis, Design and Planning 181
STATION
Cable (c)
228 ft (d)
R3 (pos +neg)
B
V A
R5
Substation
Cable (k)
Cable (a)
Cable (i)
V
R4
R2
889 ft (g) 6163 ft (h) 806 ft (f) 396 ft (e)
79 ft (b)
6216 ft (r) 606 ft (q) 1068 ft (p) 364 ft (o) 327 ft (m) 301 ft (l)
R1
B
A
282 ft (j+n)
Running Rails
Conductor Rail
Conductor Rail
Running Rails
Tee pos
Tee neg
750
{0*Cableneg}
750
933.07
{51*Runrail109}
{75*Cablepos}
{79*Conrail150}
{250*Runrail113}
{198*RunrailRT60}
{84*Runrail113}
{327*RunrailRT60}
{228*Conrail150} {396*Conrail150}
{82*Runrail113} {176*RunrailRT60}
{340*RunrailRT60}
{38*Cableneg}
24.30u
{6216*Runrail109}
{49*Runrail109}
{41*Runrail113}
{6216*Runrail109}
{49*RunrailRT60}
{41*RunrailRT60}
{6163*Conrail100} {889*Conrail150}
{327*RunrailRT60}
{7*RunrailRT60}
{294*Runrail113}
{500*Cablepos}
{282*Conrail150}
{282*RunrailRT60}
{84*RunrailRT60}
{198*RunrailRT60}
{82*RunrailRT60} {176*Runrail113}
{340*RunrailRT60} {1068*Runrail113}
{1068*Runrail113}
{806*Conrail100}
{100*Cablepos}
Figure 10: Detailed equivalent circuit (using B2Spice) showing elements
extracted from the schematic and summary tables.
The models were found to produce identical results to the spreadsheets, hence
proving to be a very effective way of checking the PBProCalc results. This
calibration then, in effect enhances confidence in the PBProCalc for use as a
reliable tool for this type of application.
6 Discussions and conclusions
It was originally thought that the enhancements and reinforcement to the
negative bonding arrangements as described in [1], together with changes to
conductor rail size, would logically lower the associated electrical impedances in
comparison to the existing infrastructure. Benefits may have been realised in the
form of allowing increases in (current based) protection settings, thus permitting
higher train currents in section. However, the overall impact was seen to be
minimal. This was due to a range of factors which included modifications to the
S&C layout and associated detailed electrical feeding arrangements, together
with the application of the methodology contained in the client’s process
document. This was further supplemented by the corresponding guidance note
[4], namely working within the parameters of 7.5% when undertaking the
calculations, thus allowing retention of many existing settings. If one was to stay
within parameters of 2.5%, as suggested by [3] then definite changes would have
been warranted.
Consequently, the resultant sensitivity of the calculations to the actual
negative bonding reinforcement was seen to be relatively low. Moreover, the
majority of settings calculated were noted as being close to existing and
therefore not necessitating any changes. Only in a few instances were
recommendations made to change settings (actually requiring settings to be
reduced). Although no major changes were perceived, this exercise was still
seen to be of great benefit in that it provides a solid basis for future applications
182 Power Supply, Energy Management and Catenary Problems
when ascertaining these types of protection settings. This approach was
considered pragmatic and gave a scientific means of calibration between the
spreadsheet mathematics and software. Moreover, the method allows for further
refinement and development as changes take place in the future.
The further development work undertaken has provided additional confidence
in the spreadsheet and methods developed, offering an alternative approach for
consideration when undertaking calculations in the future.
Acknowledgements
The authors would like to thank Network Rail for their permission to publish this
paper.
The support of Thomas Palfreyman, Head of Electrification, PB Ltd in
developing this calculation process and spreadsheet, and in reviewing this paper,
is also recognised.
Much of this paper has been reproduced from the paper entitled ‘DC
protection calculations – an acceptable approach’, authors R. Leach, D. Tregay
& M. Berova, first presented at the Eleventh International Conference on
‘Computer System Design and Operations in the Railway and other Transit
Systems’ (COMPUTERS IN RAILWAYS XI) and can be found in section 6,
page 425, of the book of conference papers (WIT Transactions on the Built
Environment, Volume 103), © 2008 and is reproduced with permission of WIT
Press, Southampton, UK.
References
[1] Development of an Improved Traction Return System, Eur Ing Raymond
Leach and Dennis Tregay, paper Railway Engineering – 2007, 9th
International Conference, London, UK, 20/21 J une 2007.
[2] NR/SP/ELP/21051 ‘Calculation of Protection Settings for DC Circuit
Breakers’, Issue 2 dated Dec.2005
[3] NR/GN/ELP/27006 ‘Calculation of Protection Settings for DC Track
Feeders, Issue 2 dated April 2006
[4] Network Rail Southern Region Power Supply Upgrade Project  Guidance
Note A43700DC32031 ‘DC Protection Setting Calculation’, Issue C1.0
dated 11 May 2004
[5] Network Rail Computer Programme “RUN FILE DC3.EXE”
[6] Parsons Brinckerhoff (PB) Methodology for Undertaking Protection Setting
Calculations associated with remodelling and negative bonding changes
(Issue 1 dated 07/09/06)
Power Supply System Analysis, Design and Planning 183
Appendix A: Flowchart of the protection setting calculation
process for the Tee feed case
Yes
No
Analyse the electrical
section layout to determine
worst case scenario
Establish the type of electrical section equivalent diagram
Using the existing and As Built drawings populate the schematic
diagrams with the coordinates (in Miles and Feet) of important points,
including mile posts, points of cable connections, points of change of
conductor / running rail type, point tips, etc.
Is the electrical
section over switches
or crossings?
Provide existing records drawings, including:
•Comprehensive Track Diagrams
•Arrangement of conductor rail, hook switches and jumper cables
•Arrangement of track circuits & negative bonding
•Others, e.g. Signalling plans
Provide survey drawings showing the existing arrangement of the
permanent way, conductor rail and negative bonding
Obtain all necessary input data
Provide relevant standards, guidance notes and other instructions
Produce a schematic diagram for each electrical section based on
the Comprehensive Track Diagram (CTD), including the conductor
rail and the negative bonding layout showing points, substations,
TPHuts, cable connections, types of conductor and running rails, etc.
Provide drawings showing the As Built arrangement of the
Permanent Way, Conductor rail and Negative Bonding
Provide existing records for protection settings calculations, including
type of protective devices and current settings
B
A
184 Power Supply, Energy Management and Catenary Problems
Note: In plain line feeding case latter comparison stage is omitted.
Yes
No
B
Carry out a check
& correct error(s)
Is the discrepancy due to alignment
remodelling, different contact rail or
negative bonding design or
repositioning of the equipment?
No Yes
No Yes
No
A
Using the information in the schematic diagram populate the Summary
table with the types and lengths of the equivalent circuit components.
Compare the data describing the equivalent components in the
schematic diagram with those in the existing protection settings
calculation records
Do the data agree?
Establish the reason for discrepancy
Independent check of the schematic
diagram and input data
Error(s) found?
Protection settings – Results (A)
Run NR Protection Setting
calculation software (Teefeed)
Protection settings – Results (B)
Running PB DC Protection
Setting calculation spreadsheet
Do Results (A) agree
with Results (B)?
Comparison of protection settings
results (A) and (B)
Overall independent check
Yes
Power Supply System Analysis, Design and Planning 185
This page intentionally left blank
Author Index
Abrahamsson L. ....................... 157
Albrecht T. ................................... 3
Bae C. H. ................................... 35
Berova M. ................................ 171
Bosselmann T. ........................... 87
Cayuela L. P. ............................. 55
Cucala A. P. ............................... 55
Danielsen S. ............................. 145
Domínguez M. ........................... 55
Fernández A. ...................... 55, 113
Finlayson A. ............................. 123
Fosso O. B. .............................. 145
Goodman C. J . ......................... 123
Haga H. ...................................... 65
Han M. S. ................................... 35
Hertsch H. .................................. 87
Hisatomi K. ................................ 13
J imenezOctavio J . R. .............. 135
Kaiser J . ..................................... 87
Kim Y. K. .................................. 35
Ko H. ......................................... 45
Koseki T. ................................... 13
Kwon S. Y. ................................ 35
Leach R. ................................... 171
Luethi M. ................................... 75
Lukaszewicz P. .......................... 25
Matsuda K. .......................... 45, 65
Miyatake M. ........................ 45, 65
Nishi K..................................... 101
Okada Y. .................................... 13
Park H. J . ................................... 35
Pilo E. .............................. 113, 135
Puschmann R. ............................ 87
Rouco L. .................................. 113
Sato Y. ..................................... 101
Shimada T. ............................... 101
Söder L. ................................... 157
Theune N. .................................. 87
Toftevaag T. ............................ 145
Tregay D. ................................. 171
Weiland K. ................................. 95
White R. D. .............................. 123
Willsch M. ................................. 87
Power Supply, Energy Management and Catenary Problems 187
...for scientists by scientists
WITPress
Ashurst Lodge, Ashurst, Southampton,
SO40 7AA, UK.
Tel: 44 (0) 238 029 3223
Fax: 44 (0) 238 029 2853
EMail: witpress@witpress.com
Urban Transport XV
Urban Transport and the Environment
in the 21st Century
The continuing requirement for better urban
transport systems in general and the need
for a healthier environment has led to an
increased level of research around the world.
This is reflected in the proceedings of this
wellestablished meeting which demonstrates
the steady growth and research into urban
transport systems. The variety of topics
covered by the conference are of primary
importance for analysing the complex
interaction of the urban transport
environment and for establishing action
strategies for transport and traffic problems.
The fifteenth conference topics are: Urban
Transport, Planning and Management;
Transportation Demand Analysis; Intelligent
Transport Systems; Land Use and Transport
Integration; Air and Noise Pollution;
Environmental and Ecological Aspects;
Traffic Integration and Control; Transport
Modelling and Simulation; Safety and
Security; PublicTransport Systems.
WIT Transactions on The Built
Environment, Vol 107
ISBN: 9781845641900
eISBN: 9781845643676
Published 2009 672pp
£255.00/US$510.00/€357.00
Edited by: C.A. BREBBIA,
Wessex Institute of Technology, UK
Timetable Planning and
Information Quality
Edited by: I. HANSEN,
Delft University of Technology, The
Netherlands
The book comprises a number of research
papers presented at several Computers in
Railways Conferences. It has been compiled
by Ingo A. Hansen, President of the
International Association of Railway
Operations Research (IAROR) and comprises
selected papers originating from different
countries, such as Denmark, France,
Germany, Japan, Italy, Netherlands, Sweden
and Switzerland. The papers give an
overview of the current stateoftheart
analytical approaches, methods and
simulation tools for the modelling and
analysis of network timetables, the
distribution of train delays and realtime
rescheduling of perturbed operations. The
topics include e.g. railway capacity
estimation according to the UIC norm 406,
train punctuality analysis based on standard
track occupation and clearance data, and
boarding, alighting and distribution of
passengers along suburban trains, as well as
fast recognition and resolution of conflicts
between train movements in case of
disturbances by means of realtime speed
adaptation, reordering or rerouting. The
book can serve as an introduction to the
theory of railway traffic, timetable design,
operations analysis, simulation, safety and
control for Master and PhD students from
engineering faculties and professionals
working in the railway industry.
ISBN: 9781845645007
eISBN: 9781845645014
Forthcoming / apx 208pp / apx £79.00/
US$158.00/€111.00
...for scientists by scientists
Fundamentals of Road
Design
M.K. JHA, Morgan State University,
USA and W. KUHN, University of Leipzig,
Germany
Currently there is no single textbook that
addresses the fundamental geometric
concepts of urban and rural road design.
The traffic behavior has significantly
changed over the last two decades and
warrants newer methods for road design.
Unlike the advances made in building, ship,
and aircraft design, road design is still
carried out in a traditional way which is
more than fifty years old.
Starting from the traditional design
process this book introduces a 3dimensional
geometric design process of urban and rural
road design. It will prove to be a valuable
textbook for undergraduate and graduate
university students as well as road planning
and design practitioners. Interested readers
may find this book a valuable resource in
conjunction with the authors’ recently
published book entitled “Intelligent Road
Design”, which is written at an advanced
level and addresses intelligent algorithmic
applications in road alignment optimization.
Series: Advances in Transport, Vol 20
ISBN: 9781845640972
Forthcoming apx 350pp
apx £120.00/US$240.00/€180.00
Find us at
http://www.witpress.com
Save 10% when you or der fr om our encr ypted
or der ing ser vice on the web using your cr edit
card.
All prices correct at time of going to press but
subject to change.
WIT Press books are available through your
bookseller or direct from the publisher.
WIT eLibrary
Home of the Transactions of the Wessex
Institute, the WIT electroniclibrary
provides the international scientific
community with immediate and
permanent access to individual papers
presented at WIT conferences. Visitors to
the WIT eLibrary can freely browse and
search abstracts of all papers in the
collection before progressing to download
their full text.
Visit the WIT eLibrary at
http://library.witpress.com
Hybrid Vehicle Propulsion
C.M. JEFFERSON, University of the West
of England, UK and R.H. BARNARD,
University of Hertfordshire, UK
In this book, the authors review recent
progress in the development of a range of
hybrid vehicles and describe the results of
field trials and operational experience.
Numerous tables, graphs and photographs
are included together with clear references.
The volume will be of great interest to
engineering and technical staff working in
the road and rail vehicle industries, and final
year undergraduates and postgraduates
studying mechanical and automotive
engineering.
Series: Advances in Transport, Vol 10
ISBN: 1853128872 2002 176pp
£69.00/US$107.00/€103.50
...for scientists by scientists
Find us at
http://www.witpress.com
Save 10% when you or der fr om our encr ypted
or der ing ser vice on the web using your cr edit
card.
Advances in City Transport
Case Studies
Edited by: S. BASBAS, Aristotle University
of Thessaloniki, Greece
Highlighting the highly topical subject of
transport and the environment and the
closely related field of town planning, this
book contains chapters concerning
developments in the transportation systems
of various cities all over the world. These
include Singapore, São Paulo, Santiago,
Bilbao, Eindhoven, Adelaide, Bangalore and
Thessaloniki.
The studies featured will be of interest
to postgraduate researchers in transport and
the environment, engineers and planners
working within transport and environment
ministries and local authorities, and
consultants.
Series: Advances in Transport, Vol 17
ISBN: 185312799X 2006 208pp
£66.00/US$120.00/€99.00
How to Make TwoLane
Rural Roads Safer
Scientific Background and Guide for
Practical Application
R. LAMM, A. BECK and T. RUSCHER,
University of Karlsruhe, Germany,
T. MAILAENDER, Mailaender Ingenieur
Consult GmbH, Germany and S. CAFISO
and G. LA CAVA, University of Catania,
Italy
In most countries twolane rural roads make
up about 90 percent of rural networks and
account for about 60 percent or more of
highway fatalities worldwide – 500,000
people per year. Based on new research and
the demands of many design professionals
this book provides an understandable
scientific framework for the application of
quantitative safety evaluation processes on
twolane rural roads.
The methodology described will support
the achievement of quantified measures of
1) design consistency, 2) operating speed
consistency, and 3) driving dynamic
consistency. All three criteria are evaluated
in three ranges described as “good”, “fair”
and “poor”. It has been proved that the
results of these criteria coincide with the
actual accident situation prevailing on two
lane rural roads. By using the “good” ranges
sound alignments in plan and profile, which
match the expected driving behaviour of
motorists, can be achieved.
The safety criteria are then combined
into an overall safety module for a simplified
general overview of the safety evaluation
process. The authors also encourage the
coordination of safety concerns with
important economic, environmental and
aesthetic considerations.
This book will be an invaluable aid to
educators, students, consultants, highway
engineers and administrators, as well as
scientists in the fields of highway design and
traffic safety engineering.
ISBN: 1845641566 2006 144pp
£48.00/US$85.00/€72.00
...for scientists by scientists
Innovations in Freight
Transport
Editors: E. TANIGUCHI, Kyoto
University, Japan and R.G. THOMPSON,
University of Melbourne, Australia
Highlighting new ideas and best practice, this
book examines innovations in modern
freight transport systems.
Par t ial Cont ent s: I ntelligent Transport
Systems; Vehicle Routing and Scheduling;
Logistics Terminals; I ntermodal Freight
Transport; Underground Freight Transport
Systems; ECommerce and the
Consequences for Freight Transport; Future
Perspectives.
Series: Advances in Transport, Vol 11
ISBN: 1853128945 2002 216pp
£76.00/US$118.00/€114.00
We ar e now able to supply you with details of new
WIT Pr ess titles via
EMail. To subscr ibe to this free ser vice, or for
infor mation on any of our titles, please contact
the Mar keting Depar tment, WIT Pr ess, Ashur st
Lodge, Ashur st, Southampton, SO40 7AA, UK
Tel: +44 (0) 238 029 3223
Fax: +44 (0) 238 029 2853
Email: mar keting@witpr ess.com
Computers in Railways XII
Computer System Design and
Operation in the Railway and Other
Transit Systems
This volume features the proceedings of the
Twelfth International Conference on
Computer System Design and Operation in
the Railway and other Transit Systems. This
book updates the use of computerbased
techniques, promoting their general
awareness throughout the business
management, design, manufacture and
operation of railways and other advanced
passenger, freight and transit systems. It will
be of interest to railway managers,
consultants, railway engineers (including
signal and control engineers), designers of
advanced train control systems and computer
specialists.
The COMPRAIL series has become the
world forum for all major advances in this
important field, and this latest conference
volume highlights themes of great current
interest. These are: Planning; Safety and
Security; Advanced Train Control; Drivers
Operations; Communications; Energy
Supply and Management; Operations
Quality; Timetable Planning; Level Crossing
and Obstacle Detection; Computer
Techniques; Dynamics and Wheel/Rail
Interface; Maintenance; Rolling Stocks;
Training Tools and Technology; Condition
Monitoring; Asset Management; Maglev
and High Speed Railway; Passenger
Information Systems; Train Regulations;
Metro and Other Transit Systems; Advanced
Train Control.
WIT Transactions on The Built
Environment, Vol 114
ISBN: 9781845644680
eISBN: 9781845644697
Forthcoming /apx1000pp /apx£380.00/
US$760.00/€532.00
Edited by: B. NING, Beijing Jiaotong
University, China and C.A. BREBBIA,
Wessex, Institute of Technology, UK