PAPER

Development of a DC Feeding Bus Ground Fault Detection System for Railway

Substations

Hiroaki MORIMOTO Shin-ichi HASE Toru AIHARA

Researcher, Senior Researcher, Laboratory Head, Researcher,
Power Supply Systems, Power Supply Technology Div.

High-voltage DC ground relays (64P) are commonly used at DC railway substations to

detect DC feeding bus ground faults, which may cause fire or other serious damage. There-
fore, 64P relays, which send a command to open all related circuit breakers, are very
important devices. However, 64P relays sometimes disrupt railway traffic because they
are occasionally activated by irrelevant faults, such as ground faults at external feeding
lines or other substations. We are now developing a new method of detecting DC bus
ground faults that can eliminate the influence of unrelated faults. This paper gives an
outline of the new method and the results of basic investigation.

Keywords
Keywords: DC railway substation, DC bus ground faults, high-voltage DC ground relays (64P)

1. Introduction Rectifier
Bus ground fault Ground
DC bus
DC railway substations are equipped with high-volt- mat
age DC ground relays (hereafter referred to as 64P re- Ground
lays) to detect DC feeding circuit ground faults within Return 64P resistance at
wire High-speed substation
the substation compound. If a ground fault occurs at a Ground circuit
64P (Zero
substation or on its feeding line, however, the 64P relays mat breaker
potential)
of substations near the fault point may be activated, er-
Rail equivalent
roneously indicating that the fault has occurred at a Feeding and trolley wires grounding
nearby substation. This phenomenon is referred to be- resistance
low as a 64P relay joint fault. As feeding stops completely Rail
when a 64P relay is activated, recovery is slow and train
operation may be disrupted for long periods, particularly (a) Bus ground fault
where two or more substations are involved.
We are now studying a method of detecting only bus
ground faults at the substation in question without in-
fluence from faults on the feeding line or elsewhere, and
developing a new protection system for bus ground faults. Ground
The basic study results are summarized below. 64P resistance of
Ground structure
Feeding wire mat
ground fault (Zero
potential)
2. Relationship between joint faults and rail leakage
64P Rail equivalent
resistance
grounding
Ground resistance
structure
The 64P relays currently in use monitor the voltage
between the substation ground mat and the return wire
corresponding to the rail, and judge that a ground fault (b) Feeding wire ground fault
has occurred at the substation when the voltage exceeds Fig. 1 Current composition of 64P relay and equivalent
the setting value (generally 500V in JR companies) (Fig. circuit at ground fault
1). When a ground fault occurs in the substation com-
pound, the potential of the ground mat and that of the to a ground fault on the feeding line) and a state where a
rail become positive and negative respectively, against the ground fault has occurred in the substation compound.
remote ground potential taken as the zero-potential ref- If the voltage between the ground mat and the rails ex-
erence (Fig. 1 (a)). When the voltage inbetween exceeds ceeds the setting value, the 64P relay will be activated.
the setting value, the 64P relay is activated. When viewed from a substation without incident, the
When a ground fault occurs on the feeding line, the ground fault at an adjacent substation is the same as that
ground mat potential remains at zero (as no voltage is on the feeding line. Therefore, the same phenomenon
applied), while the rail potential becomes negative. As takes place in this case. This phenomenon is known as a
the 64P relay monitors only the voltage between the 64P relay joint fault.
ground mat and the rails, it cannot distinguish between In terms of rail potential, rails are normally in in-
a state where the rail potential alone has changed (due complete contact with the ground through sleepers and

QR of RTRI, Vol. 46, No. 2, June. 2005 103

 ballast. The contact resistance in this state, expressed
Rail Sleeper by the conductance between the ground and rails, is called
the rail leakage conductance. It should be noted that the
rail leakage conductance is treated as the conductance
between the ground and the left and right rails as a pair.
Ballast Since the rail leakage conductance exists along the en-
tire track length, the rails are considered to be grounded
as a distributed parameter system (Fig. 2). Therefore the
r (ǡ/km), l (mH/km) rail potential that has emerged for any reason at a par-
ticular point propagates to remote places while decaying.
When the rail resistance is constant in the same type
rail, the rail potential decays quickly when rail leakage
conductance value G is large, and propagates over a long
distance when G is small.
Discussed below is a case where five substations A to
G (S/km) E exist at intervals of 10km along a single track of infi-
Fig. 2 Equivalent electric circuit of track nite length with rectifiers in operation without feeding,
and a ground fault occurs at substation C. The distribu-
tion of potential can be obtained by solving a model that
approximates each 1km-long rail segment with an equiva-
Approximate equivalent circuit Rectifier lent π circuit in Fig. 3 (a), in which the rail inductance is
per 1km of single-track 1500V4000kW
omitted as it is irrelevant to the calculation of the volt-
Resistance of 50N rail Voltage regulation 8%
age distribution in a steady state.
r = 0.017Ω/km Figure 3 (b) shows the simulation results when the
Rail leakage conductance Bus ground fault
64P rail leakage conductance of a ballast section is set at
G = 0.05S/km
0.05S/km, for example. As the rail potential exceeds
0.017Ω Grounding resistance
-500V along the entire track length in this case, the 64P
0.2ǡ
relay will be activated at all substations A to E where the
0.025S 0.025S
setting value is 500V.
In recent years, vibration-proof materials have been
10 pieces Track of infinite length pasted on the bottom or sides of sleepers or used to float
the track as a whole. Track structures of this type often
claim high insulation against the ground and a low rail
leakage conductance of 0.001S/km or less. Although such
64P 64P 64P 64P
low rail leakage conductance is desirable in preventing
electrolytic corrosion due to leakage current under nor-
mal conditions, it has unexpectedly caused concern due
10km 10km 10km 10km to increases in 64P relay joint faults.
Substation A Substation B Substation C Substation D Substation E

(a) Simulation model 3. Use and limitation of discharging gaps as a means to

suppress rail potential

750 A 64P relay joint fault is caused when a ground fault

Rail potential, Ground mat potential [V]

Rail potential occurs and makes the rail potential negative, which then
500
Ground mat propagates to remote places. It is thought therefore that
250 potential the suppression of rail potential at ground faults is a way
to prevent 64P relay joint faults.
0 For this purpose, a method of introducing resistance
-250 and discharging gaps (connected in series between the
ground mat and rails) was examined (Fig. 4). The gap
-500 discharges at a voltage equivalent to the conventional
setting value for the 64P relay. It was thought that with
-750
this arrangement the gap would discharge to suppress
-1000 the voltage between the ground mat and rails when a bus
-20 -10 0 10 20 ground fault occurs.
A simulation was performed using a model installed
Distance from the ground fault suffered substation [km]
with a discharging gap (at a discharge starting voltage of
(b) Simulation results 500V) and a current-limiting resistance of 0.05Ω at sub-
Fig. 3 Potential distribution at substation where bus ground station C in Fig. 3 (a). Figure 5 shows the results of the
faults occurred simulation. The rail potential was substantially lower
than the value without the discharging gap, and was kept

104 QR of RTRI, Vol. 46, No. 2, June. 2005

 propriate under the conditions set in this study.1)
To rectifier However, it is difficult to keep the rail potential at
negative pole
adjacent substations higher than -500V using discharg-
ing gaps when: a) the rail leakage conductance is less than
in the present case, b) the grounding resistance is lower
at the substation where the ground fault occurs, or c)
Discharging gap
substations have larger capacities and are located at short
64P intervals in urban areas, for example. In such cases, the
Current-limiting
64P relays at adjacent substations will be activated, or
resistance 0.05ǡ gaps will discharge where they exist. Discharge gaps will
Return wire also be activated when the rail potential exceeds the dis-
Ground mat charge starting voltage in the case of feeding line ground
faults. Therefore the installation of discharging gaps alone
Grounding resistance
cannot completely eliminate the joint fault of 64P relays.
0.2ǡ It is possible to judge only from the activation or non-
activation of discharging gaps whether a fault has oc-
Rail curred at the substation in question or elsewhere.

Fig. 4 Discharging gap and current-limiting resistance

4. Protection system against bus ground faults using
discharging gaps
750
Rail potential, Ground mat potential [V]

Rail potential
500 The discharging gap installed between the ground mat
Ground mat and rails discharges at bus ground faults to constitute a
250 potential short circuit that quickly increases the rectifier output
0 current. It is thought therefore that bus ground faults
can be detected by checking the increased rectifier cur-
-250 rent rather than the voltage between the ground mat and
-500 rails. Consequently, a protection system against bus
ground faults using the rectifier current was devised.
-750 Figure 6 shows the composition of the system 2) 3), which
-1000 has the following characteristics.
-20 -10 0 10 20 (1) A discharging gap and a current-limiting resistance
are connected in series between the ground mat and
Distance from the ground fault suffered substation [km]
rails in place of the conventional 64P relay, constitut-
Fig. 5 Simulation results of the potential distribution with ing a short circuit when a voltage equivalent to the
discharging gaps setting value of the 64P relay is applied (Fig. 6 (a)).
(2) A fault selectve relay (50N) is installed on the return
higher than -500V at adjacent substations B and D. Thus, wire of the substation to detect the increased rectifier
the 64P relays at these substations were not activated, current at bus ground faults.
and joint faults were prevented at the 64P relay with set- (3) Use of the logic circuit in Fig. 6 (b) enables us to judge
ting value of 500V. At substation C, the potentials of both that a fault has occurred on the feeding line of the
rail and ground mat were suppressed when the gap was feeding circuit when its fault selectve relay (50F) is
discharged. This was caused by a large current flowing activated, and that a bus ground fault has occurred
due to the current-limiting resistance connected in par- when the 50N relay is activated and other 50F relays
allel with the route of the ground fault current, reducing are not.
the rectifier output voltage in accordance with its per- If a feeding line ground fault or a bus ground fault at
cent voltage regulation. another substation occurs, the discharging gap discharges
If the current-limiting resistance is omitted, the posi- when the potential of the substation return wire exceeds
tive and negative poles of the rectifier will be connected the discharge starting voltage. This is equivalent to the
directly, inducing an excessive current when a bus ground rails of the substation being temporarily grounded. From
fault occurs. To prevent this phenomenon and protect a macroscopic viewpoint, however, it is simply equivalent
the components, a higher resistance is therefore desir- to the rail leakage conductance being increased in the
able. On the other hand, as high resistance causes a large conventional system. This means that there are no
voltage drop for the rectifier due to the fault current, a changes except for a slight decrease in the rail potential
lower resistance is required to suppress the rail poten- from its value prior to the introduction of the discharg-
tial at bus ground faults. This means that the optimal ing gap. If the 50F relay can detect feeding line faults
value of the current-limiting resistance depends on track correctly, therefore, the logic circuit in Fig. 6 (b) will judge
conditions and the levels of protection coordination. If it that a feeding line fault has occurred regardless of
is a prerequisite to keep the rail potential higher than - whether the 50N relay is activated or not.
500V at adjacent substations when a bus ground fault
occurs, a current-limiting resistance of 0.05Ω will be ap-

QR of RTRI, Vol. 46, No. 2, June. 2005 105

 5. A field ground fault test with discharging gaps

Rectifier To confirm that installation of a discharging gap be-

DC bus tween the ground mat and the return wire at substa-
Return Bus tions reduces the rail potential at bus or feeding line
50N 64P
wire ground ground faults and prevents the 64P relay joint fault, a
fault 54F 54F low-voltage ground fault test was performed on a 12km-
GP R
long double-track section between Kyobashi and
50F 50F Amagasaki on the JR Tozai line 2) 4).
Ground mat

Feeding and trolley wires

5.1 Test method
Test

Rail Figure 7 outlines the test circuit. A discharging gap

was installed at the Mitejima substation and another at
50F : Fault selective relay for feeding circuit GP : Discharging gap the Shin-Fukushima substation, and feeding line/bus
50N : Fault selective relay for return wire R : Current-limiting resistance ground faults were simulated at the two substations us-
54F : DC high-speed circuit breaker
ing the former as a power source and the latter as an
(a) Composition of main circuit and relays adjacent substation.
In the test, 20 or 24 x 12V automobile lead batteries
were used, arranged in series and connected with a resis-
50N tance of about 1Ω to limit the short circuit current to ap-
proximately 200A. The discharge starting voltage of the
Detection of
discharging gap was as low as about 230V, and the cur-
DC bus rent-limiting resistance was omitted in the test. At both
50F11 ground fault ends of the test section, the impedance bonds were dis-
connected in order to separate the rails of the test section
from other rails and set a state of low rail leakage con-
50Fnn ductance.
To supply a current to the test system, we connected
(b) Logic circuit to detect bus ground faults the DC bus line or the feeding line with a grounding pole
Fig. 6 Composition of a new DC bus ground fault detection (substation ground mat, station distribution board
system grounding wire or water-resistant iron door structure) at
one of Kashiima station, Mitejima substation, Ebie sta-

Mitejima
substation
Batteries

Amagasaki Closing switch Shin-Fukushima Kyobashi

1ǡ
substation substation substation

A A

V Discharging gap V Discharging gap

Osakajo-
Kashima Ground Ebie Ground Kitazume
station mat station mat station
Down track feeding line

Ground Ground Ground

structure structure structure
V V V
Down track rail

Up track rail
Connection/disconnection Connection/disconnection
of impedance bond of impedance bond
Fig. 7 Ground fault simulation test on JR Tozai line
Tozai

106 QR of RTRI, Vol. 46, No. 2, June. 2005

tion, Shin-Fukushima substation and Osakajo-Kitazume tween the two poles of the discharging gap) at the Shin-
station in advance to simulate a ground fault, and then Fukushima and Mitejima substations when a bus ground
connected the batteries with a closing switch as a power fault was simulated at the Mitejima substation in the
source. above tests. It took about 15ms for the gaps to discharge
after the closing switch was turned on. Meanwhile, the
5.2 Test results (steady state)
Test relationship between the discharging gap voltages at the
two substations became more complex in that the voltage
Table 1 summarizes the results of measurement. was higher at the Mitejima substation for some periods
No gaps discharged when 20 x 12V lead batteries were and lower for others. It is therefore unknown which gap
connected in series to simulate a bus ground fault at the discharges first, as the situation depends on unknown
Mitejima substation. The voltage between the ground- conditions. After the discharge gap has discharged, vi-
ing pole and rails was about 185V at all measuring points. bration waveforms are observed. The reason is that it
If a bus ground fault occurs under the same conditions as takes some time to charge the power filter capacitor (LC
those of this test, therefore, a 64P relay joint fault will resonance circuit) installed at the substations through
also occur. As a current of 46A was found to flow in this the series resistance of the power source, and resonance
case, the rail leakage conductance was about 0.01S/km takes place between the power filter and the feeding cir-
when grounding resistance at the Mitejima substation cuit inductance.
was omitted.
When 24 x 12V lead batteries were connected in se- 5.4 Simulation to reproduce the field test
ries, one of the two gaps discharged, increasing the ground
fault current to about 200A and significantly reducing To discuss the system of protection against ground
the voltage between the grounding pole and rails at all faults, it is necessary to assess the phenomenon when a
measuring points. This was due to the grounding mat fault occurs at an actual power source. Since the object
and rails being short-circuited by the discharging gap, in this case is a ground fault rather than a metallic short-
substantially reducing the resistance of the grounding circuit, however, it is difficult to use an actual power
current route and increasing the voltage drop due to the source for tests. A model was therefore made to simulate
resistance (1Ω) installed at the power source. As an the transient phenomenon and reproduce the test circuit
equivalent internal resistance amounting to 20-90 mΩ in Fig. 6 as closely as possible, and the power source was
exists in actual rectifiers, therefore, it is thought that the then replaced with a rectifier to simulate and estimate
rail potential can be reduced to some extent through the the actual conditions of a bus ground fault.
use of discharging gaps at bus ground faults. Figure 9 shows the results of the test to simulate a
The test was repeated with the impedance bonds con- bus ground fault at the Mitejima substation (correspond-
nected at both ends of the section, but the potential was ing to the case of Fig. 8). The waveforms between the
not emerged at a ground fault current of about 200A due turning of the closing switch and the gap discharge cor-
to the high rail leakage conductance beyond each imped- respond closely to the measurement results shown in Fig.
ance bond. As a result of this test, the discharging gaps 8. Conversely, the waveforms after the gap discharge dif-
were not activated. fer from the measurement results in that their amplitude
is larger and their duration longer. Nevertheless, the two
Table 1 Result of ground fault simulation (voltage and current cases are close to each other as a whole, which means
steady) that the simulation model sufficiently reproduces the con-
Number of 　
Voltage between grounding pole and rails ditions observed in the actual circuit. It is thought that
storage Ground
Ground fault
point
batteries fault Osakajo- the minor differences between the simulation and mea-
in series current Kashima Mitejima Ebie Shin-Fukushima
Kitazume
connection station
　
substation station substation
station surement results are caused by the negligence of numeri-
Kashima station 24 199A 21.0V 8.7V Discharge 13.2V 13.1V cal errors in circuit elements, the discharging gap volt-
　
20 46A 185V 186V 186V 184V
age at discharge, the skin effect of conductors and the
Mitejima
substation
resistance component of the ground return circuit imped-
24 205A 10.4V 8.6V Discharge 10.4V 10.4V
　 ance.
Ebie station 24 190A 22.6V 25.4V 21.1V 14.1V Discharge 15.5V Based on the above, the power source was replaced
Shin-Fukushima
substation
24 190A 21.0V 23.8V 14.0V Discharge 14.2V with an actual 1,500V rectifier, and the discharge gap was
Osakajo-
24 180A 21.8V 24.4V 13.9V Discharge 22.9V
set at a starting voltage of 550V and a current-limiting
Kitazume station
resistance of 0.05Ω. Feeding through the feeding line was
stopped, and the AC circuit breaker of the rectifier trans-
5.3 Test results (transient state)
Test former was turned on in a state of bus ground fault. Fig-
ure 10 shows the results of simulation under these condi-
Since discharging gaps activate when the voltage be- tions. As the power source impedance was small in this
tween the two poles reaches the discharge starting volt- case, the discharging gap of the Mitejima substation
age, it is not possible to explain the discharging phe- quickly discharged with little influence from the power-
nomenon by the voltage in the steady state alone (as in separating filter, inducing a large fault current. At the
the conventional method) - the transient phenomenon Shin-Fukushima substation, the voltage between the
must also be taken into consideration. ground mat and rails exceeded 550V a little later, when
Figure 8 shows the waveforms of the voltage between the discharging gap discharged to induce a current flow.
the ground mat and rails (equivalent to the voltage be- When these results are cross-checked with Fig. 6, the

QR of RTRI, Vol. 46, No. 2, June. 2005 107

 300
Voltage between ground mat and rails [V]

1000
Mitejima substation

Current at the discharging gap[A]

800
Mitejima substation
200
600
400
100
200

0 0
Shin-Fukushima -200
substation Shin-Fukushima substation
-100 -400
-0.01 0 0.01 0.02 0.03 0.04 -0.01 0 0.01 0.02 0.03 0.04
Time [s] Time [s]
(a) Voltage waveform between ground mat and rails
Voltage (b) Current waveform at the discharging gap
Fig. 8 Transient phenomenon when a bus ground fault is simulated at the Mitejima substation
Transient

300 1000
Voltage between ground mat and rails [V]

Current at the discharging gap [A]

Mitejima substation 800
Mitejima substation
200 600
400
100
200
0
0
Shin-Fukushima -200
substation Shin-Fukushima substation
-100 -400
-0.01 0 0.01 0.02 0.03 0.04 -0.01 0 0.01 0.02 0.03 0.04
Time [s] Time [s]
(a) Voltage waveform between ground mat and rails
Voltage (b) Current waveform at the discharging gap
Fig. 9 Results of the simulation to reproduce the ground fault test
Voltage between ground mat and rails [V]

1000 20000
Return wire current at Mitejima substation
800 Mitejima substation 15000
Current [A]

600 10000 Discharging gap current

at Mitejima substation
400 5000

200 Shin-Fukushima substation 0 Discharging gap current at

Shin-Fukushima substation
0 -5000
-0.01 0 0.01 0.02 0.03 0.04 -0.01 0 0.01 0.02 0.03 0.04
Time [s] Time [s]
(a) Voltage waveform between ground mat and rails
Voltage (b) Current waveform at the discharging gap and return
wire
Fig. 10 Results of the simulation of a bus ground fault at the Mitejima substation with actual power source

50F relay was not activated since a current did not flow 6. Summary and future subjects
in the feeding circuit at the Mitejima substation, but the
50N relay was activated due to the fault current. At the To detect bus ground faults at a substation itself with-
Shin-Fukushima substation, a current flowed only out being affected by ground faults on the feeding line
through the discharging gap. As a result, neither the or elsewhere, a system involving discharging gaps and
50F relay nor the 50N relay was activated. This method return wire fault selectve relays was proposed, and its
can detect bus ground faults only at the substation it- basic function was confirmed through simulation and
self, thus avoiding influence from faults elsewhere. field tests.

108 QR of RTRI, Vol. 46, No. 2, June. 2005

 However, the system within the range discussed in 2) Kawahara, K., Hase, S., Morimoto, H., Yazawa, H.,
this report cannot be applied to feeder sectioning sta- Itou, T., Takahashi, N., Study of Countermeasure for
tions without rectifiers. In addition, a number of mat- Unnecessary Operation of Ground-fault Overvoltage
ters remain unresolved in ensuring satisfactory perfor- Relay on DC Substation, 2002 National Convention,
mance under different system operating conditions. Such Institute of Electrical Engineers of Japan, 47, 2002.
issues include methods to determine the discharge start- 8.
ing voltage of discharging gaps and the value of current- 3) Kawahara, K., Morimoto, H., Hase, S., Takigawa, T.,
limiting resistance, the performance required from fault Itou, T., Takahashi, N., Study on a Protection System
selectve relays for feeding and return wires, and the lev- for Ground Fault using ∆I Return Current in DC Trac-
els of coordination required for protection. These mat- tion Substation, 2003 National Convention, Institute
ters will be addressed further with a view to the com- of Electrical Engineers of Japan, 5-235, 2003. 3.
mercial development of the system. 4) Suzuki, M., Nobuhara, T., Takigawa, T., Kawahara,
K., Hase, S., Morimoto, H., Development Protection
System for Ground Fault in DC Traction Substation,
References Workshop on Transport and Electric Railways, Insti-
tute of Electrical Engineers of Japan, TER-03-33,
1) Kawahara, K., Hase, S., Morimoto, H., Yoshihara, S., 2003. 7.
Itou, T., Takahashi, N., Study on a Protection Sys- 5) Kawahara, K., A Protection System for Ground Fault
tem for Ground Fault in DC Traction Substation, in DC Traction Substation, Railways and Electrical
2002 National Convention, Institute of Electrical En- Engineering, Vol. 14, No. 11, pp. 18-23, 2003. 11.
gineers of Japan, 5-200, 2002. 3.

QR of RTRI, Vol. 46, No. 2, June. 2005 109