IEEE Transactions on Transportation Electrification 1

Reliability and life evaluation of a DC traction power supply system

considering load characteristics
Yilin Chen, Zhongbei Tian, Clive Roberts, Stuart Hillmansen, Minwu Chen

Abstract—The traction power supply system is one of the most meeting of ASME (American Society of Mechanical
important parts of a railway system, which is responsible for Engineers) and IEEE (Institute of Electronics Engineers) [8].
providing electricity to power the running trains and other The paper first proposed the concept of reliability of traction
operating equipment. The performance of the power supply has a power supply systems. The definition of the reliability of a
profound impact on the railway system. Therefore, it is necessary
to conduct research on the reliability of the power supply system
traction power supply system is clearly stated in this article, but
considering traction load and ensure it is highly reliable. This the paper does not give a method for evaluating the reliability.
paper introduces some reliability-related indexes based on the The paper also proposed the establishment of a power
Fault Tree Analysis (FTA) method and proposes a method to reliability organization similar to NERC (North American
evaluate the system reliability of a DC-electrified railway. The Electric Reliability Corporation) to ensure the development and
effect of traction load on power supply system reliability is studied improvement of North American power system reliability. The
based on IEEE standards. A dynamic reliability and life research on the reliability of traction power supply system is
evaluation model is built up to evaluate the reliability of traction still in the initial stage of exploration. Its analysis method
substation and to develop a plan to reduce the operational risk of mainly relies on the reliability theory of power system, and
the power system. Through analysis of a practical example of the
Singapore East West line, the system reliability and remaining
combined with the characteristics of traction power supply
useful life of each traction substation are obtained. The outcome system. There are a few reliability-related studies on electrified
from this study provides guidance for the maintenance and future railways and their power systems. Most of the existing
operation plan of the power system. reliability studies focus on AC-electrified railways, but few
focus on DC railway power supply systems. Wang developed
Index Terms—DC power supply; Fault Tree Analysis (FTA); an analytical simulation method to evaluate railway catenary
Reliability evaluation; Traction load; Traction power supply systems using credibility theory [9]. Feng et al. applied Failure
system (TPSS) mode and effects analysis (FMEA) to high-speed railways for
reliability evaluation considering relay protection [10], which
I. INTRODUCTION provides guidance for the maintenance of high-speed railways.
DC traction power systems play an important role in the Chen et al. used the Fault Tree Analysis method to evaluate the
reliable operation of urban railways. Outage of the power power supply system in AC railways [11]. Ku & Cha used a
supply will not only cause transportation paralysis, but also Minimal Cut Sets Algorithm to assess the reliability of railway
cause serious economic losses. It is necessary to conduct substations [12].
research on DC traction power supply system fault analysis and Much of the current literature on the reliability of traction
to evaluate the power supply system reliability. power systems pays particular attention to the power grid and
The issue of reliability engineering dates back to around its components. Hayashiya evaluated the reliability of each
1920. During this period, the concepts of reliability were first component in a DC power supply system based on 10 years of
proposed [1]. After 10 years of research, people gradually metro operation data in Japan and conducted quantitative
learned about reliability issues. Then, around 1940, reliability analysis on a redundant system configuration [13]. Li used an
issues were included in engineering disciplines, and they were improved sequential Monte Carlo method to evaluate the
developed from the losses caused by the unreliability of system reliability of a metro traction substation, so as to
products [2]. In 1957, the US Advisory Group on Reliability of determine the weakest links of the system [14]. These studies
Electronic Equipment (AGREE) issued a report in which they analyze the links and factors that affect the reliability of power
clearly pointed out the standard specifications and supply systems from the perspective of power electronics and
measurement methods for reliability [3]. From then on, systems. Constant failure rate was adopted when conducting
reliability science gradually entered research fields. After reliability assessment, which is not able to reflect the real-time
nearly six decades of development, reliability science has been operating conditions of the system. This is the deficiency of
well developed, and it has gradually become a comprehensive current reliability research on railway power supply system. In
discipline with a wide range of topics and applications [4], such the process of daily operation, the traction load is affected by
as infrastructure designs [5], computer and communication station, route, passenger flow and other aspects. Research [15]
systems, and rail transportation systems [6, 7]. shows that an impulsive load will not only affect the normal
Research on the reliability of traction power supply systems operation of the main output substation, but also have a great
began in 2004. Sagareli published a typical paper at the annual impact on adjacent power plant units, reducing their service life
This paper has been partially supported by funding of Chengdu Guojia Electrical Engineering Co., Ltd (No. NEEC-2019-B04). (Corresponding author:
Zhongbei Tian.)
Y. Chen, C. Roberts, S. Hillmansen are with Department of Electric, Electrical and Systems Engineering, University of Birmingham, Birmingham, U.K.
(e-mail:, [email protected], [email protected], [email protected],).
Z. Tian is with the Department of Electrical Engineering and Electronics, the University of Liverpool, Liverpool, U.K. (e-mail: [email protected]).
M. Chen is with National Rail Transportation Electrification and Automation Engineering Technology Research Centre, Chengdu 611756, China and School of
Electrical Engineering, Southwest Jiaotong University, Chengdu 611756, China (e-mail: [email protected]).
IEEE Transactions on Transportation Electrification 2

and level of quality. However, the current research on railway system. The power supply system shown in Fig. 1 uses a dual
power system reliability has not taken into consideration of the redundant design for the traction power system. All of the
real-time load. The impact of traction load and the actual electronics are doubled, and if one of the branches is
operation of the trains on system reliability has not been disconnected from the fault and the other is turned on, the entire
considered when designing a DC railway power supply system. system can still be powered. Most of the current industrial
Previous studies of railway power supply networks usually designs use redundant system design. However, there is a
focused on infrastructure design and capacity, rather than problem with redundant systems. Compared with simple
system operation. In many cases, due to the complexity and systems, the number of devices in a redundant system
cost of power supply system equipment, it is impossible to increases, and the installation and maintenance costs of the
carry out extensive upgrades. Under the condition that the devices also increase.
power supply system has been built and the infrastructures are Zone 1 101 102 103 104
not changed, optimization of the traction load and operation T1 T2 T3 T4

schedules can be considered to maximize the reliability of the 1001

1005 1002 1003
1006 1004

system. There is potential to study the influence of traction load 201 202 203 204
Zone 2
on traction power supply systems for better system operations. T5 T6 T7 T8
This paper presents an approach to evaluating the reliability 2001 2002 2003 2004

and life of DC traction power supply systems, considering Zone 3

3001 3003
301 302 303 304 3002 307 308
305 306
traction load characteristics. Based on the Fault Tree Analysis Upline

(FTA) method, a few reliability-related indexes are introduced Downline

to evaluate system reliability. A dynamic multi-train power
3006
simulator is developed to study the impact of traction load, 3004 3005

which is able to reflect the real-time condition of the power Transformer Isolation switch

system. The highlights of this paper are as follows: Circuit breaker Insulation
Train

• The FTA method is applied to the railway DC traction power

supply system. The reliability of a typical 750 V DC power
Fig. 1. Main electric wiring diagram of a DC traction power supply system
supply system is analyzed.
• A TPSS life evaluation model is developed to study the There are three main components in the whole power
influence of traction load. In this model, the remaining life system, rectifier transformers, circuit breakers, and isolators.
and failure rate of the device under the influence of dynamic To simplify the evaluation model, the rectifier and transformer
load can be obtained, thus establishing a relationship with the are considered as a whole, because rectifier transformers have
reliability evaluation. been widely used in the modern railway industry. Each
• A multi-train simulator is developed for system evaluation component is defined and labeled as the bottom event in the
and operation plan decision-making. A case study of FTA, and its corresponding event code is shown in the
Singapore East-West metro line is presented to intuitively following Table I. T1-T8 represent the rectifier transformers in
show the impact of traction load on TPSS reliability. The the power supply system, 3001-3006 are the isolation switches,
real-time operation conditions of TPSS is evaluated. and the rest represent the circuit breakers.
The paper is organized as follows: Section II studies the
structure of a DC substation and its main components, as well TABLE I. BOTTOM EVENT AND ITS CORRESPONDING EVENT CODE
as the failure mechanism of the substation. Section III Bottom Event Bottom Event Bottom Event Bottom Event
illustrates a reliability evaluation example of a typical railway event code event code event code event code
power supply system. On this basis, the impact of load T1 X1 103 X11 305 X21 2001 X31
characteristics on the reliability of the substation system is T2 X2 104 X12 306 X22 2002 X32
studied in Section IV and Section IV gives a case study. Section T3 X3 201 X13 307 X23 2003 X33
T4 X4 202 X14 308 X24 2004 X34
VI concludes.
T5 X5 203 X15 1001 X25 3001 X35
T6 X6 204 X16 1002 X26 3002 X36
II. OVERVIEW OF DC TRACTION POWER SUPPLY SYSTEM T7 X7 301 X17 1003 X27 3003 X37
A. Schematic layout of a 750V DC substation T8 X8 302 X18 1004 X28 3004 X38
101 X9 303 X19 1005 X29 3005 X39
In the UK’s DC electrified railway system [16], the National 102 X10 304 X20 1006 X30 3006 X40
Grid or Distribution Network Operator (DNO) first supplies Zone 1 is the high-voltage section which gets AC voltage
AC high voltage (HV), typically 132, 66 or 33kV to the main from the National Grid. Zone 2 is called a traction substation,
substation, then the voltage is stepped down to a medium which is used to convert electrical power as supplied by the
voltage of 33, 22 or 11kV. After that, the alternating current is National Grid to a certain level for providing power to a rail
stepped down and rectified to 1500 or 750V DC through the system. Zone 3 connects to the catenary system. There are four
traction substation to supply power to the train. The train takes core components in the traction substation system,
power via the pantograph or the third rail, and finally the transformers, rectifiers, circuit breakers, and isolators. All the
current is fed back to the traction substation by the return line. components are in series and function together to enable the
The 750V level is most widely used for metro and urban rail system to provide 750V DC power to the train, as shown in Fig.
transport. Fig. 1 shows the feeding arrangements and basic 2 [17, 18]. It is important to note and understand the major
components of a typical 750V DC traction power supply
IEEE Transactions on Transportation Electrification 3

components shown in the diagram and to get an overview of faulty diodes or faulty circuits. Human error and the external
how they work together for the entire system. Although the environment such as weather are a big factor in causing
study focuses on a 750V DC traction system, it can be applied failures, which is sometimes inevitable. Note that it is not
to various different installations where the same principles will always possible to completely prevent damage, but the ultimate
be used. goal is to have as much control as possible and minimize the
MV Network
risk of failure with existing conditions. In the following
sections, the failure mechanism of important components in the
AC circuit breaker
substation will be analyzed based on the existing IEEE
standards, and the main factors affecting the fault will be
Transformer obtained.

AC/DC rectifier III. IMPORTANCE ANALYSIS USING FAULT TREE

At present, the method of assessing the reliability of traction
DC circuit breaker
power supply systems in academia is not perfect. In order to
750V DC adapt to the development of railway power supply systems,
Traction circuit new quantitative evaluation methods and evaluation models
breakers
Upline catenary should be explored. H. A. Watson at Bell Laboratories first
or third rail
developed the concept of Fault Tree Analysis to evaluate a
Insulator
Downline catenary
or third rail control system [22]. FTA has since become a comprehensive
Fig. 2. Schematic layout of a typical DC traction substation evaluation method with a wide range of topics and applications
[23-25]. It draws a fault tree diagram to determine the various
B. Electronic components fault mechanism combinations of product failure causes and the probability of
The Failure Mode and Effects Analysis, or FMEA, can be their occurrence. In this research, a reliability-related method is
defined as “potential failure mode and consequence analysis”. introduced and applied to the railway system to identify the
In the product design stage and process design stage, FMEA factors affecting the reliability of the substation power supply
analyzes the various processes that constitute the product, the system for fault diagnosis and maintenance. The weakest links
parts, and the process of the composition, finds out all the of the power system will be determined.
potential failure modes, and analyzes the possible
consequences, to determine the pre-requisites to improve the A. FTA modeling
quality and reliability of the product. It is important to Based on the existing literature using the more mature FTA as
understand the function of the major components of a traction the theoretical basis, this paper analyzes the reliability of an
substation and their fault mechanisms. Table II shows Failure electrified railway traction power supply system, the causes of
Mode and Effects Analysis (FMEA) summarized from IEEE system failure, and proposes measures to improve system
standards and some papers [19-21]. reliability. Before analyzing the reliability of the system, two
To summarize, most transformer failures are related to the preconditions for the hypothesis are proposed based on the
deterioration of insulation. Most rectifier failures are caused by basic characteristics of the traction power supply system:
1. All the pieces of electric equipment involved in the power
TABLE II. FAILURE MODE AND EFFECTS OF MAJOR COMPONENTS

Equipment Failure modes Cause Effects of failure

Transformer Over-temperature Overload or over-current due to fault in the Protection will cause supply to be switched off. If not, there
(Stepping down the system will be major damage to the transformer.
supply voltage to Short circuit Isolation materials damaged High current densities and a rapid increase in winding
750V) temperature.
Open circuit Loose connection, burnt windings The fault will cause gas and heat to build up in the
transformer. Protection circuit will switch off the supply.
Low oil level Leakage at gaskets If protection circuit operates, the supply will be switched off.
Otherwise the transformer will blow up.
Rectifier Over-temperature Arcing inside rectifier, loose connection Over-temperature protection should switch off the supply. If
(AC input to DC between diode and heat sink, cooling fans not not, the rectifier may be damaged.
output) working
Short circuit Faulty diodes, circuit fault Protection circuit should isolate the rectifier.
Open circuit One or more diodes not working, loose No output from the rectifier.
connections, circuit fault
Circuit breaker Not opening Faulty opening relay, protection circuits not Substation may be damaged because circuit breaker does not
(Protection switch) working properly open.
Not closing Faulty closing relay, protection circuits not No power supply.
working properly
Current leakage to Dirty porcelain or breakdown of isolating Damage to the circuit breaker and possible trip of substation if
earth materials leakage is detected by AC earth leakage.
External damage Vandalism, lightning, water Current leakage and possible disintegration.
Isolator Broken insulators Vandalism, lightning, water in the base plates Isolator short-circuited.
(To isolate the causing cracking
substation from Opening mechanism Rust, broken operating handle No operation will be possible.
supply) failure
IEEE Transactions on Transportation Electrification 4

system can be statistically considered to be independent of each 3) Importance factors

other during reliability analysis. The failure of one device has Importance analysis is conducted to determine the weakest
limited impact on other devices. To simplify the simulation links/parts of the system from a quantitative point of view. The
analysis, the influence of devices on each other is not fault tree formula is used to calculate the probability
considered in this study. importance of the bottom event for the top event. By comparing
2. All the pieces of electric equipment are considered as the magnitude of the probability importance, the weakest links
repairable components in the FTA. can be obtained. Enhancing these weak links can improve
1) Fault tree construction system reliability. There are a few importance factors which
In the FTA, the basic event is connected to one or more top help evaluate the system:
events by logical symbols such as AND gates and OR gates 1. Marginal Importance Factor (MIF)
[26]. Top events generally refer to undesired system failures or MIF is often called the Birnbaum Importance Factor. MIF
events that endanger the system. Bottom events usually refer to represents the difference between the unavailability when event
a component failure or a human’s faulty operation. X occurs and the unavailability when event X does not occur.
The basic steps of the FTA method are as follows: MIF gives the increase in risk associated with the occurrence of
1. Define the system and system failures and determine the event X. The MIF is defined in (6), where 𝑄(𝑡) is system
bottom events; unavailability, 𝑞𝑥 (𝑡) is unavailability of event x, 𝑄(𝑡)|𝑞𝑥=1 is
2. Build the fault tree model; unavailability with 𝑞𝑥 = 1 , 𝑄(𝑡)|𝑞𝑥=0 is unavailability with
3. Conduct qualitative and quantitative analysis. 𝑞𝑥 = 0.
𝜕𝑄(𝑡)
𝑀𝐼𝐹(𝑡) = = 𝑄(𝑡)|𝑞𝑥=1 − 𝑄(𝑡)|𝑞𝑥=0 (6)
2) Reliability evaluation indexes 𝜕𝑞𝑥 (𝑡)
The reliability index is an important part of reliability theory.
In order to quantitatively evaluate the reliability of the power 2. Critical Importance Factor (CIF)
system, four indexes are introduced in this paper: CIF is often called the Fussell-Vesely Importance. CIF
1. Reliability R(t) indicates the risk associated with a given event, that is, how
The probability that the system will perform the specified much occurrence of the event is contributing to system failure.
function under the specified conditions and within the specified CIF can be calculated using MIF and system unavailability
time t is called the reliability of the system, and is defined as 𝑄(𝑡), as shown in (7).
R(t)=P(T>t), t≥0. The reliability of a system at t is the 𝑞𝑥 (𝑡)𝑀𝐼𝐹(𝑡) 𝑄(𝑡) − 𝑄(𝑡)|𝑞𝑥=0
probability that it experiences no failure during time interval 𝐶𝐼𝐹(𝑡) = = (7)
𝑄(𝑡) 𝑄(𝑡)
[0,t]. Reliability is normally a function of time and can be
calculated by the following formula: 3. Risk Reduction Worth (RRW)
𝑅(𝑇) = P(T > t) RRW represents the risk that would be reduced by reducing
𝑇
(1) the unavailability of event X to zero. RRW is defined in (8).
= 𝑒𝑥𝑝 [− ∫ 𝜆(𝑡)𝑑𝑡]
0 𝑄(𝑡)
Where 𝜆(𝑡) is the system failure rate. 𝑅𝑅𝑊(𝑡) = (8)
𝑄(𝑡)|𝑞𝑥=0

2. Failure rate/Occurrence 𝜆(𝑡)

The failure rate is the ratio of the number of components that B. DC power system analysis
fail per unit time to the total number of components. The After understanding the structure and basic components of
relationship between failure rate and reliability is as follows: the traction substation, the failure events of the system are
𝑑 determined according to the operation mode of the traction
𝜆(𝑡) = − ln 𝑅(𝑡) (2)
𝑑𝑡 substation. Fig. 3 shows the fault tree model of the system.
If the component failure rate obeys an exponential Traction Power Supply System Failure
distribution, the reliability R(t) is:
𝑅(𝑡) = 𝑒 −𝜆𝑡 (3)

3. Repair rate 𝜇(𝑡). Zone 1 failure Zone 2 failure Zone 3 failure

Repair rate is the probability that a component will be

1 2 3
repaired in the next unit time after it has been in operation for a
period of time, which is defined as 𝜇(𝑡). a) DC power system fault tree model
Zone 1 failure
4. Unavailability
The unavailability of a system at t is the probability that it is Incoming lines Left part incoming Right part incoming

not working at t. The calculation formula can be expressed as: failure lines failure lines failure

𝜆
𝑄(𝑡) = 𝑞𝑒 −(𝜆+𝜇)𝑡 + (1 − 𝑒 −(𝜆+𝜇)𝑡 ) (4)
𝜆+𝜇
Average unavailability 1005 1006

1 𝑇
𝑄𝐴𝑣𝑔 (𝑇) = ∫ 𝑄(𝑡)𝑑𝑡 (5) 101 T1 1001 103 T3 1003 101 T1 1001 103 T3 1003

𝑇 0
Where 𝜆(𝑡) is the failure rate, and 𝜇(𝑡) is the repair rate. 102 T2 1002 104 T4 1004 102 T2 1002 104 T4 1004
IEEE Transactions on Transportation Electrification 5

b) Zone 1 fault tree model for fault tree simulation and reliability evaluation.
Zone 2 failure
TABLE III. RELIABILITY PARAMETERS OF RELEVANT EQUIPMENT AS
REPAIRABLE COMPONENTS

Equipment Failure rate Repair rate

16.6
Transformer 𝜆(𝑡) = ( ) ∙ 𝑡 15.6 𝜇(𝑡) = 5.15 × 10−2
167.2516.6
201 T5 2001 203 T7 2003 7.75
Circuit breaker 𝜆(𝑡) = ( ) ∙ 𝑡 6.75 𝜇(𝑡) = 0.10
178.677.75
204 T8 2004
202 T6 2002
10.88
Isolator 𝜆(𝑡) = ( ) ∙ 𝑡 9.88 𝜇(𝑡) = 0.25
c) Zone 2 fault tree model 175.9810.88

Zone 3 failure
In the FTA simulation, an assessment of the reliability and
unavailability of the substation for 10 years was carried out. As
Upline failure Downline failure the failure rates of electrical components increase with time, the
reliability of the whole system decreases year by year. Fig. 4
Incoming lines Incoming lines
depicts the probability of the top events over time, which are
failure failure
the system unavailability and reliability over time (years). After
3002 3005
10 years of operation of the power supply system, its system
302 303 306 307 301 304 305 308 reliability is 0.962115, and its unavailability is 0.000981057.
Left part failure Right part failure Left part failure Right part failure
TPSS system reliability and unavailability over years
System reliability System unavailability

1.005 0.0012
3001 3004 3006
3003
1
0.001
301 304 0.995
302 303 306 307 305 308

System unavailability
0.99 0.0008
System reliability

d) Zone 3 fault tree model 0.985

0.0006
Fig. 3. Fault tree models of a DC traction power supply system 0.98

0.975 0.0004

0.97
1) Qualitative analysis 0.965
0.0002

Using FTA software and computational algorithms, the fault 0.96

0 1 2 3 4 5 6 7 8 9 10
0

tree model of the traction substation is qualitatively analyzed to Years

obtain the minimum cut sets. The minimal cut set number is
199, in which there are 12 second-order cut sets, 162 Fig. 4. Traction power supply system reliability and unavailability over time
fourth-order cut sets, 16 sixth-order cut sets, 8 seventh-order
cut sets, and 1 eighth-order cut set. 3) Importance factors analysis
The probability of system failure in a cut-set event is From the fault tree software and the calculation formulas
inversely proportional to its order. That is, the lower the order listed in Section II, the importance indicators of each
of the minimum cut set, the easier it is for the corresponding component to system failure are calculated, shown in Table IV.
event to fail the system, so the more important the cut set. It can
TABLE IV. IMPORTANCE FACTOR RESULTS OF RELEVANT COMPONENTS
be summarized from the minimum cut set result obtained that
the second-order minimum cut sets {X1, X29}, {X2, X29}, Bottom events Unavai Marginal Critical Risk
{X3, X30}, {X4, X30}, {X9, X29}, {X10, X29}, {X11, X30}, lability Importance Importance Reduction
Worth
{X12, X30},that is, the transformers T1, T2, T3, T4, circuit X29, X30 0.0056 0.0869 0.4959 1.9837
breakers 101, 102, 103, 104, 1005 and 1006 have the greatest X1, X2, X3, X4 0.0287 0.0534 0.1563 1.3852
influence on the reliability of the system. From a system X9, X10, X11, X12 0.0107 0.0053 0.0573 1.0025
perspective, it is necessary to avoid problems with the X25, X27, X28, X26 0.0056 0.0052 0.0298 1.0608
equipment included in the low-order minimum cut set to ensure
X13, X14, X15, X16 0.0287 8.65E-05 0.0025 1.0307
that the system can function properly. In the actual system
design, highly reliable circuit breakers should be used at 1005 X18, X19, X22, X23, 0.0287 2.28E-10 6.66E-09 1.0000
X17, X20, X21, X24
and 1006 to minimize the occurrence of minimum cut set
X5, X6, X7, X8 0.0107 8.49E-05 9.27E-04 1.0009
events. Transformers T1, T2, T3, T4 are also the weakest links X31, X32, X33, X34 0.0056 8.45E-05 4.81E-04 1.0005
of the power system, which should be enhanced.
X36, X39 0.0056 2.67E-09 1.52E-08 1.0000
X35, X37, X38, X40 0.0056 1.33E-09 7.91E-09 1.0000
2) Quantitative analysis
The devices reliability parameters were tested on a traction
substation of Beijing-Shanghai HSR in China using FTA and It can be seen from the simulation results that bottom events
FMEA method [10, 11]. This study employs the empirical X29, X30, X1, X2, X3 and X4 are of most importance to the
reliability parameters of the electrical components from the system. They have the greatest impact on the reliability of the
field tests in [10, 11] as listed in Table III. The failure rate of traction power supply system. Fig. 5 lists the Marginal and
each component is a time-related function. These data are used Critical Importance of major components, sorted by the degree
of importance. X29 and X30 have the largest Risk Reduction
IEEE Transactions on Transportation Electrification 6

Worth numerical value for RRW of 1.9837. If the probability of 𝑇𝐻𝑆 = 𝑇𝐴 + ∆𝑇𝑇𝑂 + 𝑇𝐺 (9)
failure of the bottom events X29 and X30 is reduced, the
probability of occurrence of the top event can be quickly
reduced, that is, the probability of failure of the traction power Where, 𝑇𝐴 is the ambient temperature, ∆𝑇𝑇𝑂 is the top-oil
supply system, which is more effective than reducing the temperature rise over the ambient temperature, 𝑇𝐺 is the
probability of occurrence of any other bottom events by the winding hottest-spot temperature rise over top-oil temperature.
same value. 2) Diode rectifiers (IEEE P1653.2-2009 [19])
Marginal and Critical Importance A traction rectifier is used to convert AC power to DC
0.6 power. A controlled rectifier means that the DC output voltage
Marginal and Critical Importance

0.5 can be controlled or adjusted in value based on demand,

0.4
whereas an uncontrolled rectifier will produce a fixed DC
0.3
0.2
output. Traction rectifiers are either 6, 12 or 24 pulse. The
0.1 six-phase full-wave uncontrolled diode rectifier is the basic
0 design on most 750V DC traction substations.
X1
X2
X3
X4
X9
X29
X30

X10
X11
X12
X25
X27
X28
X26
X13
X14
X15
X16
X18
X19
X22
X23
X17
X20
X21
X24
The standard rating of a rectifier unit for heavy traction
Compommets
service is as follows [19]: one hundred percent rated load
Marginal Importance Critical Importance
amperes continuously until constant temperatures have been
reached by all parts of the rectifier unit, followed by either 1)
Fig. 5. Order of Marginal and Critical Importance of major components in the 150% current for 2 h, or 2) 300% current for 1 min. Most
DC traction power supply system
modern rectifiers have been designed with proper circuits to
Through quantitative analysis of the traction power supply indicate the condition of the diodes, protect the rectifier against
system (TPSS), it can be seen that the results are consistent with over-temperature and monitor the cooling fan.
the analysis results of the minimum cut set of the qualitative The diode can withstand short periods of overload. After
analysis. To ensure a highly reliable system, the key returning to the normal load after overloading, the temperature
components should be monitored and should be repaired and and leakage current of the diode will return to the normal
replaced in time. values, so it will not have a great impact on its life. Studying the
effect of dynamic load-time characteristics on diode life does
IV. TPSS LIFE EVALUATION not make much sense, as it has little influence on diode failure.
3) Circuit breakers (IEEE Std C37.14 [21])
A. Life of components The circuit breakers form a very important part of the
There are four major components in a TPSS system, protection circuits in the substation. They will switch off the
transformers, rectifiers, circuit breakers and isolators. In this supply when a fault occurs inside the substation. The circuit
section, the failure mechanisms of these major components will breaker tripping mechanism is operated by protection relays
be described separately, and the reliability model of the traction which protect the substation from overloading and high fault
substation will be derived. currents.
1) Transformers (IEEE Std C57.91-2011 [20]) The Consumer Product Safety Commission (CPSC)
The transformer is one of the key pieces of equipment for the estimates the life expectancy of a circuit breaker to be around
normal operation of a traction power system. The load capacity 30 to 40 years. Short circuits, power surges and spikes are the
of an oil-immersed transformer is closely related to its top-oil main cause of circuit breaker failure. Circuit breakers are not
temperature and hot spot temperature. There are several factors used frequently under the normal operating conditions of a
affecting the transformer’s overall life expectancy and overload substation. Under normal operating conditions, the reliability
capability, however the hottest-spot temperature is the most and life of the circuit breaker have little to do with the load.
critical parameter for determining the transformer’s life Therefore, in this study, the impact of dynamic load on the
expectancy. A higher winding hot-spot temperature will cause circuit breaker is ignored.
degradation of the winding insulation and increase the potential 4) Isolators
for transformer failure. An inordinate temperature rise in an Isolator switches, also called disconnectors, in particular, are
oil-immersed transformer is featured by hot-spot temperature. used in power grids and substations to isolate very high voltage
It is an important parameter to determine the optimum load apparatus such as transformers and circuit breakers when they
ability of a transformer. are due for maintenance or when there is a fault. When a fault
IEEE C57.91-2011 [20] presents a commonly used thermal occurs, the isolator will cut out a portion of the substation.
model (Clause 7) for calculating the transformer hot-spot B. Loss-of-life calculation
temperature and top oil temperature. The specific formulas are
not detailed in this paper. For details, please refer to the IEEE 1) Loss of life
standard and other related papers [27-29]. The theoretical basis The Acceleration Aging Factor 𝐹𝑎𝑎 (also called the
of the model is that an increase in the loading current of the Accelerated Aging Rate or Acceleration Factor) is defined as
transformer will result in transformer losses and then cause an the ratio of the real-world lifetime to the test duration. The
overall temperature rise. The mechanism of hot-spot higher the 𝐹𝑎𝑎 , the less reliable the test is. In the IEEE
temperature change can be explained by formula (9), which has C57.91-2011 guide, the relationship between the Acceleration
been verified by IEEE C57.91-2011 standard [20]. Aging Factor 𝐹𝑎𝑎 and the hot-spot temperature 𝜃ℎ is presented
as:
IEEE Transactions on Transportation Electrification 7

15000 15000
𝐹𝑎𝑎 = 𝑒 110+273
−
𝜃ℎ +273 (10) segments with equal length to calculate the dynamic parameters
This equation is an adaption of the Arrhenius reaction rate at different locations. In this study, the purpose is to explore the
theory. It indicates that the aging of the transformer is a impact of traction load on the power system reliability, and
function of hot-spot temperature because of the load current. whether there is a way to maximize system reliability without
𝐹𝑎𝑎 is a per-unit quantity for a reference temperature of 110°C. affecting the normal operation of the power supply system. The
The equivalent aging factor (𝐹𝑒𝑞𝑎 ) of the transformer with proposed simulation framework is shown in Fig.6. The
respect to the reference temperature 110°C over a given period dynamic inputs are train driving strategies and timetables, the
of time (T) is calculated as: outputs are reliability-related system reliability and remaining
𝑡=𝑇 useful life.
∫𝑡=0 𝐹𝑎𝑎 𝑑𝑡 (11) Fixed inputs:
𝐹𝑒𝑞𝑎 = - Train traction parameter
𝑇 (Mass, tractive effort, resistance parameter)
The loss-of-life percentage is the ratio of the loss-of-life over - Route data
(Gradient, route speed limit, stations, dwell
a period of time T and its normal insulation life, as shown in time)

(12). Normal insulation life is generally an empirical fixed

value, which could be found in the device user manual. Outputs:
- Train speed profile
𝐹𝑒𝑞𝑎 × 𝑇 Dynamic inputs: Single-train movement - Traction energy consumption
𝐿𝑜𝑠𝑠 𝑜𝑓 𝐿𝑖𝑓𝑒 % = × 100 (12) - Driving strategies simulation - Journey time
𝑁𝑜𝑟𝑚𝑎𝑙 𝐼𝑛𝑠𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝐿𝑖𝑓𝑒 - Train power requirement

The remaining useful life (RUL) is an estimate of the number Train power
of years remaining for a component or system to operate
without failure before replacement. In this study, the system’s Outputs:
remaining useful life can be calculated as: Dynamic inputs: Power network - System power flow
- Timetables simulation - Energy consumption
𝑅𝑈𝐿 = 𝑁𝑜𝑟𝑚𝑎𝑙 𝐼𝑛𝑠𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝐿𝑖𝑓𝑒 − (𝐹𝑒𝑞𝑎 × 𝑇) (13) - Power loss

2) Failure rate Multi-train simulator

According to the definition of conditional probability, the
probability of failure during time interval Δt after T equivalent Reliability-related
simulation
operation time can be calculated:
𝜆(𝑡) = 𝑃(𝑇 ≤ 𝑡 ≤ 𝑇 + ∆𝑡 | 𝑡 > 𝑇, 𝜃0 )
(14) System evaluation & optimal operation
𝐹𝑒𝑞𝑎 (𝑇 + ∆𝑡|𝜃0 )−𝐹𝑒𝑞𝑎 (𝑇|𝜃0)
= decision-making
1−𝐹𝑒𝑞𝑎(𝑇|𝜃0 )
Fig. 6. Diagram of the overall simulation structure
Where 𝜃0 is the hot-spot temperature at time t. Here, the failure
rate is a function related to time and temperature, rather than an B. Load parameters and timetable
empirical fixed value. In this paper, the Singapore East West line is used as a
simulation example. It is a high-capacity metro line operated by
V. TPSS RELIABILITY AND LIFE EVALUATION CONSIDERING SMRT, which is 49.46km in length with 27 substations, 8 tie
LOAD CHARACTERISTICS stations and 2 stations without a DC-link connection. Fig. 7
A. Simulation structure shows the Singapore line’s altitude and TPSS locations.
Different to the substation in a general power system, the
TPSS has a single load, and the impact on the traction
transformer is very large, while the load of a general power
system is mostly industrial or household load, and the
transformer is relatively more stable. After understanding the
operating modes and failure modes of the main components,
the load-time characteristics of a DC traction substation can be
determined by Matlab simulation. According to the fault
mechanism of the rectifier and isolator, only transient high
currents can damage the life of the rectifier and isolator. In
general, during normal operation, the current does not exceed
the maximum allowable threshold of the rectifier and isolator. Fig. 7. Line altitude and station locations of the Singapore line
The load current-time characteristics have a greater impact on
the life of the transformer than the other components. The C. Results and analysis
influence of instant load on the rectifier and isolators can be The normal lifespan of a distribution transformer is 20.55
ignored. Therefore, this research focuses on the impact of load years, which is 180,000 h. Suppose the power system is at its
characteristics on the traction transformer. first day of operation. The rated power of the TPSS is 4MW.
A dynamic model of single train simulation and power The ambient temperature is held constant at 25°C. A full day’s
supply system was established in Matlab to study the influence timetable is adopted in this simulation. For more accurate
of traction load. The simulator was developed for the use of results, the simulation runs every 120s and repeats to reach a
railway energy evaluation [30-33]. It is based on a discrete time full day’s timetable.
operation at which the power and energy consumption are
calculated over 0.1s time period. The train route is divided into
IEEE Transactions on Transportation Electrification 8

1）Scenario 1 (No. 1 TPSS)

Fig. 11. No. 1 TPSS loss-of-life vs. time

Fig. 8. No. 1 TPSS power vs. time
A distribution of power usage for traction substation Tuas 2）Scenario 2 (No. 7 TPSS)
Link (No. 1 TPSS) is shown in Fig. 8. These data are for a Under the premise that the entire metro line runs on the same
single train journey of 120s and show the amount of power that schedule, the reliability and life loss of different substations are
is drawn from the substation. There is a peak between 20 s and different due to the different output power and load cycle of
40 s where the load is higher. The power system experiences an each substation. Scenario 2 gives the simulation results from
overload for a short period of time. The hot-spot temperature of traction substation Lakeside (No. 7 TPSS), in which the system
the traction transformer changes with time, as shown in Fig. 9. is under normal operation and experiences no overload.

Fig. 9. No. 1 TPSS hot-spot temperature vs. time Fig. 12. No.7 TPSS power vs. time
At the maximum load of the substation, between 30 s and 40 The Aging Acceleration Factor, as shown in Fig. 13, is less
s, the Aging Acceleration Factor of the transformer is highest, than 1 when the hot-spot temperature is lower than 110 °C,
meaning that the system is most likely to fail at this point of meaning there is little loss of life.
time.

Fig. 13. No. 7 TPSS Faa vs. time

Fig. 10. No.1 TPSS Faa vs. time In this case, the total loss of life over a single run of 120 s (2
Fig.11 shows the loss-of-life percentage over given periods min) is 3.5× 10−6 years, which is equivalent to 0.03066 h (1.839
of time compared to the system’s normal life. From the min). Compared to scenario 1, in which the loads are slightly
simulation results, it can be seen that the total loss of life over a over the rated power, the total loss of life of the system in this
single run of 120 s (2 min) is 5.2156× 10 −6 years, which is simulation is very small. There will be a slight extension of
equivalent to 0.0457 h (2.74 min). lifespan compared to its normal life under rated power
operation. It can be seen that operating below the rated load of
the transformer has a negligible impact on its life and
reliability; overloading has a great influence on the system,
which will reduce the life of the transformer.
IEEE Transactions on Transportation Electrification 9

optimizing the timetable and train running diagram, the impact

of the traction load on the system can be reduced without
compromising energy consumption and travel time. It can also
provide guidance for future power supply system design and
optimal operation of the system once it is built.

TABLE V. SUBSTATION DATA FOR EAST WEST LINE

Maximum Failure rate Average Remaining
Platform Faa reliability useful life
(years)
Tuas Link 65.91 4.15 × 10−3 0.995 14.599
Fig. 14. No. 7 TPSS Loss of Life vs. time
Tuas West Road 3.75 5.39 × 10−4 0.999 19.312
3）Scenario 3 – Full year operation Tuas Crescent 67.84 4.32 × 10−3 0.995 14.215
The daily average failure rate and reliability of each TPSS Gul Circle 55.58 9.89 × 10−4 0.997 16.943
are predicted using the reliability evaluation indexes listed in Joo Koon station 25.36 7.34 × 10−4 0.997 17.874
Section III. The remaining useful life of the substations is also Pioneer Station 24.54 7.29 × 10−4 0.998 17.925
calculated, as illustrated in Table V. According to the Lakeside 0.14 8.36 × 10−5 0.999 21.754
Chinese Garden 1.59 5.30 × 10−4 0.999 19.578
simulation data listed, under the condition that the same
Jurong East 0.87 7.48 × 10−5 0.999 20.146
timetable is operated on the whole line for 1 year, the third Sungei Uiu 20.42 4.46 × 10−4 0.998 19.024
substation Tuas Crescent will bear the largest load, and the Pandan Online
estimated remaining service life is 14.215 years, which is much Commonwealth 34.87 6.65 × 10−4 0.998 18.344
shorter than for other substations. Based on the FTA results in Buona Vista 55.36 8.87 × 10−4 0.997 17.036
Section III, the simulation results can more accurately reflect Queenstown 49.87 8.25 × 10−4 0.997 17.874
Delta Online 44.69 8.17 × 10−4 0.997 17.364
the real-time operation conditions of TPSS. When developing a Outram Park 50.74 8.32 × 10−4 0.997 17.567
maintenance and replacement plan, emphasis should be placed Raffles Place 20.15 4.67 × 10−4 0.998 19.176
on this substation. It can be concluded that due to the influence City Hall 10.81 6.47 × 10−4 0.998 18.674
of dynamic traction load, each substation on the same line has Lavender 25.87 1.25 × 10−3 0.997 16.368
different life effects. Predictive maintenance could be carried Aljunied 14.35 7.02 × 10−4 0.998 18.341
out when necessary. Paya Lebar 10.69 5.27 × 10−4 0.999 19.394
In this way, running a 24 h timetable and analyzing the Eunos 29.36 7.75 × 10−4 0.998 17.964
simulation results, we can find out which time slot for each Kembangan 28.65 6.83 × 10−4 0.998 18.156
Bedok 44.58 8.06 × 10−4 0.998 17.440
TPSS is the most risky and which substation is most likely to
Sungei Bedok 30.58 6.19 × 10−4 0.998 18.343
fail. During this time period, it is important to monitor Online
substations that have the highest risk of failure. If the utilization Simei 1.87 3.65 × 10−4 0.999 19.890
rate and reliability of the current power supply system can be Tampines 0.56 8.04 × 10−5 0.999 20.694
improved, it can not only save the operation cost for railway Pasir Ris 0.19 8.96 × 10−5 0.999 21.866
utilities, but also optimize the allocation of power resources and
make contributions to energy conservation. Theoretically, by
This experimental method can be used for simulation
verification when designing a new power supply system. The
VI. CONCLUSION developed simulator can be used to verify whether the location
This paper proposes a method to investigate the influence of and sizing of the traction substations are reasonable. Similarly,
traction load on TPSS reliability and life. A comprehensive this method can be used to upgrade the existing power supply
evaluation and simulation model of a DC traction power system systems and the capacity of substations to cope with overload.
is proposed. In this paper, the reliability of a typical 750 V DC If the utilization rate and reliability of the current power supply
traction power supply system is evaluated by using fault tree system can be improved, it can not only save the operation cost
model and a series of reliability indices. According to the fault for railway utilities, but also optimize the allocation of power
mechanism of the electrical components, the thermal model of resources. The study of TPSS reliability considering load
transformer and simulation model of traction load are characteristics provides a way to pre-diagnose a system and
established. The model is used for evaluating the railway power make predictions. It can help determine when to maintain or
system and for optimal operational plan decision-making. The replace certain components based on present load, future
impact of traction load on system reliability is clearly shown. growth and ambient operation conditions.
The weakest links and time slot of the system is found. Through
the actual case analysis of Singapore metro line, it is found that REFERENCES
although the same timetable is operated on the same line, the [1] J. H. Saleh and K. Marais, "Highlights from the early (and pre-)
history of reliability engineering," Reliability Engineering & System
impact of dynamic load on substation life is different. Overload
Safety, vol. 91, no. 2, pp. 249-256, 2006.
will reduce the life span of devices of traction substation. Due [2] W. Denson, "The history of reliability prediction," IEEE
to the large random fluctuation and non-linear characteristics of Transactions on Reliability, vol. 47, no. 3, pp. 321-328, 1998.
traction load, the change of load will have a certain impact on
the reliability of the system and equipment.
IEEE Transactions on Transportation Electrification 10

[3] A. Coppola, "Reliability engineering of electronic equipment a [25] Z. Li, J. Gu, T. Xu, L. Fu, J. An, and Q. Dong, "Reliability analysis
historical perspective," IEEE Transactions on Reliability, vol. R-33, of complex system based on dynamic fault tree and dynamic
no. 1, pp. 29-35, 1984. Bayesian network," in 2017 Second International Conference on
[4] E. Zio, "Reliability engineering: Old problems and new challenges," Reliability Systems Engineering (ICRSE), 2017, pp. 1-6.
Reliability Engineering & System Safety, vol. 94, no. 2, pp. 125-141, [26] S. Kabir, "An overview of fault tree analysis and its application in
2009. model based dependability analysis," Expert Systems with
[5] P. P. D. Meyer, The Reliability of the Electric Transmission Applications, vol. 77, pp. 114-135, 2017.
Infrastructure in the 21st Century. IEEE, 2006, p. 1. [27] O. E. Gouda, G. M. Amer, and W. A. A. Salem, "Predicting
[6] K. Hou et al., "A Reliability Assessment Approach for Integrated transformer temperature rise and loss of life in the presence of
Transportation and Electrical Power Systems Incorporating Electric harmonic load currents," Ain Shams Engineering Journal, vol. 3,
Vehicles," IEEE Transactions on Smart Grid, vol. 9, no. 1, pp. no. 2, pp. 113-121, 2012.
88-100, 2018. [28] M. Hakirin Roslan, N. Azis, Z. Kadir, J. Jasni, Z. Ibrahim, and A.
[7] F. Zheng, H. J. v. Zuylen, X. Liu, and S. L. Vine, "Reliability-Based Ahmad, "A Simplified Top-Oil Temperature Model for
Traffic Signal Control for Urban Arterial Roads," IEEE Transformers Based on the Pathway of Energy Transfer Concept
Transactions on Intelligent Transportation Systems, vol. 18, no. 3, and the Thermal-Electrical Analogy," Energies, vol. 10, p. 1843,
pp. 643-655, 2017. 2017.
[8] S. Sagareli, "Traction power systems reliability concepts," in [29] D. Susa, M. Lehtonen, and H. Nordman, "Dynamic thermal
Proceedings of the 2004 ASME/IEEE Joint Rail Conference, 2004, modelling of power transformers," IEEE Transactions on Power
pp. 35-39. Delivery, vol. 20, no. 1, pp. 197-204, 2005.
[9] Z. Wang, D. Feng, S. Lin, and Z. He, "Research on reliability [30] Z. Tian et al., "Energy evaluation of the power network of a DC
evaluation method of catenary of high speed railway considering railway system with regenerating trains," IET Electrical Systems in
weather condition," in 2016 International Conference on Transportation, vol. 6, no. 2, pp. 41-49, 2016.
Probabilistic Methods Applied to Power Systems (PMAPS), 2016, [31] G. Zhang, Z. Tian, P. Tricoli, S. Hillmansen, Y. Wang, and Z. Liu,
pp. 1-6. "Inverter Operating Characteristics Optimization for DC Traction
[10] D. Feng, S. Lin, Q. Yang, X. Lin, Z. He, and W. Li, "Reliability Power Supply Systems," IEEE Transactions on Vehicular
Evaluation for Traction Power Supply System of High-Speed Technology, vol. 68, no. 4, pp. 3400-3410, 2019.
Railway Considering Relay Protection," IEEE Transactions on [32] Z. Tian, N. Zhao, S. Hillmansen, C. Roberts, T. Dowens, and C.
Transportation Electrification, vol. 5, no. 1, pp. 285-298, 2019. Kerr, "SmartDrive: Traction Energy Optimization and Applications
[11] S. K. Chen, T. K. Ho, and B. H. Mao, "Reliability evaluations of in Rail Systems," IEEE Transactions on Intelligent Transportation
railway power supplies by fault-tree analysis," IET Electric Power Systems, vol. 20, no. 7, pp. 2764-2773, 2019.
Applications, vol. 1, no. 2, pp. 161-172, 2007. [33] Z. Tian, N. Zhao, S. Hillmansen, S. Su, and C. Wen, "Traction
[12] B.-H. Ku and J.-M. Cha, "Reliability Assessment of Electric Power Substation Load Analysis with Various Train Operating
Railway Substation by using Minimal Cut Sets Algorithm," Journal Styles and Substation Fault Modes," Energies, vol. 13, no. 11, p.
of International Council on Electrical Engineering, vol. 1, no. 2, pp. 2788, 2020.
135-139, 2011.
[13] H. Hayashiya, M. Masuda, Y. Noda, K. Suzuki, and T. Suzuki,
"Reliability analysis of DC traction power supply system for electric
railway," in 2017 19th European Conference on Power Electronics
and Applications (EPE'17 ECCE Europe), 2017, pp. 1-6.
[14] K. Li, Q. Yang, Z. Cui, Y. Zhao, and S. Lin, "Reliability Evaluation
of a Metro Traction Substation Based on the Monte Carlo Method,"
IEEE Access, vol. 7, pp. 172974-172980, 2019.
[15] H. Xiao-qing, W. Gai-ping, W. Zhong, and W. Jie-xin, "Influence of
large-scale impulsive load on adjacent power plant generating
units," in 2009 International Conference on Sustainable Power
Generation and Supply, 2009, pp. 1-5.
[16] R. D. White, "DC electrification supply system design," in 6th IET
Professional Development Course on Railway Electrification
Infrastructure and Systems (REIS 2013), 2013, pp. 57-85.
[17] M. Rios and G. Ramos, "Power System Modelling for Urban
Massive Transportation Systems," in Infrastructure Design,
Signalling and Security in Railway, D. X. Perpinya, Ed.: InTech,
2012, pp. 179-202.
[18] R. D. White, "DC electrification supply system design," in 7th IET
Professional Development Course on Railway Electrification
Infrastructure and Systems (REIS 2015), 2015, pp. 1-29.
[19] "IEEE Standard for Uncontrolled Traction Power Rectifiers for
Substation Applications Up to 1500 V DC Nominal Output," IEEE
Std 1653.2-2009, pp. 1-48, 2009.
[20] "IEEE Guide for Loading Mineral-Oil-Immersed Transformers and
Step-Voltage Regulators," IEEE Std C57.91-2011 (Revision of
IEEE Std C57.91-1995), pp. 1-123, 2012.
[21] "IEEE Standard for DC (3200 V and below) Power Circuit Breakers
Used in Enclosures," IEEE Std C37.14-2015 (Revision of IEEE Std
C37.14-2002), pp. 1-80, 2015.
[22] W. S. Lee, D. L. Grosh, F. A. Tillman, and C. H. Lie, "Fault Tree
Analysis, Methods, and Applications & A Review," IEEE
Transactions on Reliability, vol. R-34, no. 3, pp. 194-203, 1985.
[23] E. Ruijters and M. Stoelinga, "Fault tree analysis: A survey of the
state-of-the-art in modeling, analysis and tools," Computer Science
Review, vol. 15-16, pp. 29-62, 2015.
[24] F. Abdul Rahman, A. Varuttamaseni, M. Kintner-Meyer, and J. C.
Lee, "Application of fault tree analysis for customer reliability
assessment of a distribution power system," Reliability Engineering
& System Safety, vol. 111, pp. 76-85, 2013.