This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

Location of Single Phase to Ground Faults in

Distribution Networks based on Synchronous Transients
Energy Analysis
Xuewen Wang, Student Member, IEEE, Hengxu Zhang, Member, IEEE, Fang Shi, Member, IEEE,
Qiuwei Wu, Senior Member, IEEE, Vladimir Terzija, Fellow, IEEE, Wei Xie and Chen Fang

 of the faulty feeders. These methods can be roughly

Abstract—As a result of small fault current, high level of noise categorized into three groups.
and a large penetration of distributed generators (DG), in the 1) Using the steady-state signal. The amplitude and phase
neutral non-effectively grounded medium-voltage (MV) difference of the fundamental frequency zero-sequence current
distribution networks, it is quite difficult to locate the faulted line are widely used to detect the faulty feeder [6]-[10]. In [11], the
section for single phase to ground faults. In this paper, using a
technique based on synchronized measurements in distribution
5th harmonic current was extracted for faulty line selection [11].
networks, a fault location method based on the analysis of the The zero-sequence admittance was calculated using zero-
energy of the transient zero-sequence current in the selected sequence voltage and current to identify the faulty line and
frequency band (SFB) is proposed. The equivalent impedance of faulty line section in [12]. These techniques only use the
the distribution network with lateral branches is studied with an steady-state signals suitable for the steady-state ground fault,
equivalent network, and the phase-frequency characteristics of the information in the very early stage of the fault is neglected.
the equivalent impedance are analyzed. The SFB, within which As a result, these methods are not reliable for the transient
the transient energy of the faulty line section is larger than that of
fault or the resonant grounded system.
the healthy line sections is determined. A combined fault-section
location criterion is proposed and the implementation scheme is 2) Using the transient signal. The transient signal is less
illustrated with the distribution level phasor measurement units. affected by the neutral grounding mode, which makes it has a
Numerical simulations of the IEEE 34 node system and the field wider potential application. It is efficient to use the zero-
experiments of a 10kV distribution network validate the sequence current travelling wave to detect the faulty feeder
feasibility and effectiveness of the proposed method. even when the fault resistance is high [13]-[14]. However, it is
difficult to detect the faulty feeder when the voltage inception
Index Terms-- Active distribution networks, fault location, angle is near zero. The correlation analysis of the transient
neutral non-effectively grounded system, single phase to ground
zero-sequence current waveform between the faulty and the
faults, transient zero-sequence current.
healthy feeder was used for faulty feeder identification in [15]-
I. INTRODUCTION [16]. The transient component of the zero-sequence current
and voltage was used for calculating the transient energy of

S INGLE phase to ground faults accounts for up to 80

percent of all the faults in distribution networks, and most
of them are transient faults [1]. Unfortunately, it is difficult to
every feeder in [17], which may be affected by the noise and
lead to mis-judgement when the fault resistance is high. In [18],
using the wavelet packet transform method, the zero-sequence
locate these faults effectively and precisely, especially for the currents of feeders were decomposed into different frequency
neutral non-effectively grounded medium-voltage (MV) bands and a selected frequency band (SFB) was utilized for
distribution networks, which exist in China, Japan and some of faulty feeder detection. Artificial intelligence (AI) methods
the European countries. The advantage is that the power were reported in [19], which largely depend on the reliability
supply can last for a certain time after the ground faults occur of the algorithms and the training database size. And it lacks of
as the fault current is limited. However, it will deteriorate the clear physical meaning and the high computational cost restrict
insulation of the equipment, cause short circuit faults [2]-[3] the practical application of the AI techniques.
and threaten the safety of people or property. Therefore, 3) Active methods [20]-[26]. By injecting various frequency
efficient and accurate fault identification and location are signals or controlling the compensation degree of the Peterson
important for electric utilities. coil, the signals monitored in the faulty feeder are changed and
Most of the current techniques for single phase to ground the faulty feeder is detected. Different from the above methods,
fault location are specified for the detection and identification supplementary devices and extra operations are needed, which
increase the potential risks and cost.
This work was supported by the National Key Research and Development When the single phase to ground fault occurs in distribution
Program of China (2017YFB0902800) and Science and Technology Project networks, the fault signals are weak and heavily contaminated
of State Grid Corporation of China (52094017003D).
by noises, which makes it difficult to identify and locate the
Xuewen Wang, Hengxu Zhang and Fang Shi are with the Key Laboratory
of Power System Intelligent Dispatch and Control of the Ministry of faults for industrial applications. The increasing penetration of
Education (Shandong University), Jinan, China (shifang@ sdu.edu.cn). the DGs and non-linear loads in the evolving active
Qiuwei Wu is with Technical University of Denmark and is visiting distribution networks further increases the difficulties. In [27],
professor at Shandong University.
a wide-area high-frequency impedance comparison method is
Vladimir Terzija is with The University of Manchester and is visiting
professor at Shandong University. proposed based on the high-frequency impedance model for
Wei Xie and Chen Fang are with State Grid Shanghai Municipal Electric DGs without knowing exact parameters of the distribution
Power Company, Shanghai, China.

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

system. In [28], the actual inverters’ control strategies during effectively grounded MV distribution system which contains
faults of the photovoltaic (PV) power plants are considered, one feeder, can be shown as in Fig. 1. R0 , L0 and C0 indicate
and the improved Bayesian compressive sensing algorithm is per-unit-length zero-sequence resistance, inductance and
used to extract the fault characteristics and then locate the capacitance, respectively; I C 0 i (i  1, 2,..., N f  1) is the
faulted node. The synchronous measurement can improve the
effectiveness and feasibility of the above two methods. equivalent zero-sequence capacitive current per-unit-length;
Besides, reliable power supply and high-speed restoration U 0 and U f are the zero-sequence voltages at the bus and the
requirements have shown the need for an on-line and accurate fault point, respectively; R f is the fault resistance; I 0 and I f
fault location scheme. Therefore, the fault characteristics
immune to noises and new centralized algorithms based on represent the zero-sequence currents to the bus and through the
multi-measurements should be carefully studied. With the fault point (for metallic ground fault, R f =0 and I f has the
accessibility of the synchronized, low-latency and high- largest amplitude); N f is the number of nodes on the feeder.
resolution measurement data from the distribution-level phasor
To properly represent the distributed capacitive current, the
measurement units (D-PMUs) [29]-[31], the synchronized
number of the PI sections in Fig. 1 should be infinite, so the
phasor and sampling data can be acquired at the distribution
number of N f is infinite.
control center for further applications.
This paper proposes a new scheme to locate the faulted line I0 If
1 2 N f 1 Nf
section in a short time after the single phase to ground fault,
which is especially designed for the neutral non-effectively R0 L0 R0 L0 R0 L0
U0 Uf
grounded MV distribution networks. By analyzing the C0 C0 I C 02
C0 Rf
I C 01 I C 0N
equivalent zero-sequence circuit for the single phase to ground f 1

fault in distribution networks, the SFB is studied within which

Fig. 1. Earth-fault zero-sequence network
the transient zero-sequence current is far greater than the
The relationship between the zero-sequence voltage and
noises. Then, the frequency range of the SFB for a general
current along the feeder can be represented as
distribution network with lateral branches is determined based
cosh  l Z c sinh  l 
-1
on the phase-frequency characteristics of the equivalent
U 0    U f 
network. The transient energy of the monitored voltage/current    sinh  l
  
cosh  l   I f 
(1)
component within the SFB is calculated for each feeder section,  I0   Z 
 c
which shows that the faulty feeder section has the largest
transient energy. It is proven that the polarities of the zero- where Z c  ( R0  j L0 ) jC0 is the wave impedance;
sequence currents are opposite between the faulty feeder
 jC0 R0   L0C0 is the transmission coefficient; and l
2
section and the healthy ones. A new criterion combining the
above two characteristics is proposed and the settings are is the length of the feeder. As there is only one feeder in the
established to determine the faulty line section. The non-effectively grounded system, no other electrical
procedures for practical application are illustrated with the components supply zero-sequence current to the bus bar. As a
increasing emergence of D-PMUs. result, I 0 =0 . For the fundamental frequency impedance of the
The rest of the paper is organized as follows. Section II distribution feeder, the resistance is comparable to the
describes the determination of the frequency range of the SFB reactance. However, the frequency range of the dominated
by analyzing the phase-frequency characteristics of the transient components during the fault is typically more than
equivalent network. In section III, a new criterion is proposed 200Hz [33, 34], in which the zero-sequence resistance is small
and the implementation of the new fault location scheme is with respect to the reactance. Thus, when considering the
illustrated. Section IV presents the testing results of the phase-frequency characteristics, the input zero-sequence
proposed scheme both with simulation data and the field data impedance seen at the fault point can be represented as
from fault experiments. Section V demonstrate the comparison
 
Uf L
results with three other methods. Conclusions are drawn in Z oc   Z c coth(l )   j 0 cot l L0 C0 (2)
Section V. If C0
where  is the angular frequency.
II. PROBLEM FORMULATION
The simplification is reasonable when the phase-frequency
characteristics is investigated, as the resistance in the RLC
A. Phase-frequency characteristics of the distribution feeders
circuit is related to the amplitude damping. It can be seen that
On the study of zero-sequence network of the neutral non- the phase-frequency characteristics of Z oc when R0 is
effectively grounded systems, there is no necessary to consider
considered and ignored are only slightly different when series
the effect of loads [25]. Furthermore, the primary sides of DG
transformers are ungrounded [4], hence the zero-sequence resonance and parallel resonance occur as shown in Fig. 2,
circuit of the distribution networks with DGs are the same as where
that without DGs. Consequently, when a single phase to k   2l L0 C0 (3)
ground fault occurs at the end of the feeder, the zero-sequence Therefore, the influence of R0 can be ignored when
distributed parameter equivalent circuit of the neutral non- considering the phase-frequency characteristics of Z oc .

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

90 inductance of the Peterson coil. Since the degree of

R0 is considered
compensation in the actual system is limited, parallel
Angle (°)
-70 R0 is ignored
0
-80 resonance occurs at about 1.8 times 0 , as shown in Fig. 4①.
-90
ωk The frequency at which the series resonance occurs for the first
-90
time is about k , as shown in Fig.4②. As in the frequency
ωk 2ωk 3ωk that the equivalent circuit behaves inductive, the transient
frequency (Hz) components will vanish quickly due to attenuation. So the
Fig. 2. Effect of R0 on phase-frequency characteristics of Zoc transient signals in the frequency above k is trivial but the
Consequently, Z oc is capacitive if
noise could be severe. Therefore, the frequency range of the
2nk     2n  1 k (4) SFB in neutral ineffectively grounded system is chosen as
and is inductive if 2 f B    k , where f B is the boundary frequency and
 2n  1 k    2nk (5) chosen as 200 Hz [33, 34].
90
where n = 0, 1, 2, ... Uncompensated
Cp = -10%
The impedance of the equivalent circuit shows capacitive

Angle (°)
Cp = +10%
characteristic in the frequency range of 0    k , and then 0

exhibits inductive and capacitive characteristics alternately

with the increase of frequency in the range of   k . -90
0 ω0 ωk Frequency (Hz)
Based on the zero-sequence network of the neutral isolated 90 90

Angle (°)
system in Fig. 1, the zero-sequence network of the resonant 0 0
grounded system is shown in Fig. 3. y p is the zero-sequence
-90 -90
admittance of the Peterson coil, and U p is the zero-sequence ω0 1.4ω0 1.8ω0 2.2ω0 2.4ω0 ωk Frequency (Hz)

voltage on the Peterson coil. The relationship between the

Fig. 4. Phase-frequency characteristics Considering the influence of Peterson
voltage and current on the feeder can be represented as
coil
U 0   1 0  U p  In the frequency range of the SFB, the zero-sequence
    y 0   (6)
 I0   p  0  voltage and zero-sequence fault current satisfy
I0 If  U f ( )  
N f 1 ang  
 I ( ) 
1 2 Nf (8)
R0 L0 R0 L0 R0 L0  f  2
Up
U0 Uf
yp C0 C0 C0 Rf where ang is the function that calculates the phase angle of
the phasor. And in the frequency range of   0 , the voltages
Fig. 3. Zero-sequence network of resonant grounded system and the currents in the network except the fault point satisfy
Combine (1) and (6), the input zero-sequence impedance  U C ( )  
ang  0 i    , i  1, 2,...N f  1 (9)
Zoc in the resonant grounded system is expressed as  I C i ( )  2
 0 
cosh  l   y p Z c sinh  l 
Z oc  Z c (7) where U C0 i and I C0i represent the zero-sequence voltage and
sinh  l   y p Z c cosh  l  the zero-sequence capacitive current at node i .
Based on (2) and (7), the phase-frequency characteristics of The node with the maximum phase shift of the zero-
the zero-sequence impedance of the neutral isolated system sequence voltage is the one farthest from the fault point as
and the resonant grounded system are obtained, as shown in shown in Fig. 1, hence the following equations hold:
Fig. 3. For resonant grounded systems, the compensation range    U ( )    U f ( ) 
of the Peterson coil is [-10%, +10%], where "-" stands for  max  ang  f    ang  
 
under-compensation and "+" stands for over-compensation.    U C0 i ( )    U 0 ( ) 
Fig. 4 shows the phase-frequency characteristics of the zero-  (10)
 Uf
sequence impedance in the neutral isolated system, resonant  U0 
grounded systems with 10% over-compensation and 10%  cosh  0l
under-compensation, respectively. The phase shift of the U 0 ( ) with the increase of frequency
In Fig. 4, 0 is the fundamental frequency of the system, is shown in Fig.5(a). In the frequency range of 0    s , the
and C p is the compensation degree of the Peterson coil. Due phase shift is 0. The boundary frequencies s and k are
to the influence of the Peterson coil, the phase-frequency equal, and the variation trend with the increase of l is shown
characteristics of the Zoc and Z oc are slightly different in Fig.5(b).
around 0 and k . Zoc is inductive in the full frequency
band, and the bandwidth of this band is related to the

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

Phase shift of U 0 ( )
iL2 u L2 u L3
π ωs L2
2

Frequency
u L2
ωk iL1 L1 L3
iL3 l1 + l2
u L1 uL1 l1 + l3
0 L4 P
uL3
L5 l1
π L6
2
Uf l1 uL2
ωs 2ωs 3ωs 4ωs L7 uL3
Frequency Line Length
(a) Phase shift with frequency changes (b) s with different line length
(a) The zero-sequence network (b) The equivalent circuit unit
Fig. 5. Phase shift of zero-sequence voltage
Therefore, in the frequency range of the SFB, there is no Fig. 7. Zero-sequence network and the transformation
phase shift between the zero-sequence voltages in the zero- Set
sequence network. Furthermore, the phase of I C0i   and 1 (n  1) 0  1 0  n (n  1) 0
  
1  0 1   0
(14)
I f   are equal within the SFB. Therefore, the amplitudes of  0 0
(13) can be converted into
I C0i   and I f   satisfy:
cosh  (l1  l2 ) Z c sinh  (l1  l2 ) 
U L1    U L2 
N f 1
    sinh  (l1  l2 )   0 
I f     I  
i 1
C0 i (11)  I L1  
 Z c
cosh  (l1  l 2 )
  

As a result, in the frequency range of the SFB, the cosh  (l1  l3 ) Z c sinh  (l1  l3 )  (15)
+  sinh  (l1  l3 )  U L3 
amplitudes of I C0i   and I f   satisfy:  
cosh  (l1  l3 )   0 

 Zc 
N f 1
cosh  l1 Z c sinh  l1   cosh  l1 Z c sinh  l1 
 I f     
i 1
I C0 i   (12)
+  sinh  l1  U L2  +  sinh  l
 
 U L3 
 
SFB  SFB  cosh  l1   0   1
cosh  l1   0 
which means that the amplitudes of the fault current equals to  Z c   Z c 
the summation of the capacitive current within the SFB. where
 cosh  0 l2
B. The SFB for a general MV network U L2 
U L2 (16)
The topology of the zero-sequence network of a general 2
 cosh  0 l3
MV distribution network with lateral branches and DGs is U L3  U L3 (17)
shown in Fig. 6. In order to clarify the simplification process 2
of the zero-sequence network, the equivalent circuit is After four successive transformations as above, the topology
presented as in Fig. 7(a). The circuit comprises a series of the of the zero-sequence network can be transformed to the
basic unit as in the dotted box, which can be equivalent to the equivalent network, as shown in Fig. 8. uP indicates the
network shown in Fig. 7(b). transient zero-sequence voltage at the Peterson coil connection
L2 point; u L6 ( u L7 ) and iL6 ( iL7 ) are the transient zero-sequence
L1 L3
DG voltage and current at the end of the line L6 ( L7 ), respectively;
and li is the length of the ith line section.
L4 L5 L6
P DG l4 + l1 + l2 u L2
L7
uf l4 + l1 + l3
uL3
l4 uP
Fig. 6. The topology of the zero-sequence network
l5 + l6 yp
The equation of the equivalent circuit unit is l5 + l7 u L6
cosh  l1 Z c sinh  l1  uL7
U L1    1 2 0 
    sinh  l1
 I L1   Z cosh  l1   0 1  Fig. 8. The equivalent topology of the zero-sequence network
  (13)
c
Therefore, considering the phase-frequency characteristics
 cosh  l2 Z c sinh  l2  cosh  l3 Z c sinh  l3   of the network, a general distribution network can be regarded
  U L2  +  sinh  l  U L3  
  sinh  l2      as 2n  1 lines connected in parallel, n is the number of the T
 Z cosh  l2   0   3
cosh  l3   0  
 c   Z c   node with branches. As a result, the phase-frequency
characteristics of the distribution network is the same as the
where li is the length of Li , u L1 and iL1 indicate the transient characteristics of the equivalent network as shown in Fig. 8.
The frequency range of the SFB for neutral non-effectively
zero-sequence voltage and current at the beginning of the line
grounded system is 2 f B    k . k is determined by the
L1 ; u L2 ( u L3 ) and iL2 ( iL3 ) represent the transient zero-
longest line in the equivalent network, which can be calculated
sequence voltage and current at the end of the line L2 ( L3 ). via (3).

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

III. THE SCHEME FOR LOCATING THE FAULTY LINE SECTION conditions that indicate the fault occurs in the line section F
BASED ON D-PMUS can be expressed as follows
 FLSI  1
A. Fault section location criterion based on transient energy  (25)
 SGN  1
of the zero-sequence current within the SFB
Otherwise, the fault occurs at the substation bus if SGN  1 .
Assume that there are N nodes in a distribution network and
the D-PMUs are installed in the downstream of several nodes, B. The implementation procedures of the proposed scheme
the network is divided into M sections. When a single phase to The flow chart of the faulty line section location scheme is
ground fault occurs in the line section F, the relation of the shown in Fig. 9, which includes the following steps:
zero-sequence current between faulty line section and healthy 1) Determination of the SFB. Based on the topology of the
line sections within the frequency range of the SFB can be zero-sequence network, the maximum length of the equivalent
represented as line LLONG is calculated, then k is estimated by
 iS SFB 
  iS , j 
SFB

 , F k   2 LLONG L0 C0 (26)
jM , j  F
  SFB  , j  1, 2,...M (18) where L0 and C0 denote per-unit-length zero-sequence
iS , F    iS , j
 SFB 
 jM , j  F inductance and capacitance of the network. And the frequency
where iS , j  iupstream, j   idownstream
 SFB   SFB   SFB  range of the SFB can be determined as in the range of 200 Hz
, j is the SFB component of
to k /2π.
the transient zero-sequence current difference between that
 SFB   SFB  Single phase to ground fault Design the digital filter
flowing in and out the line section j; iupstream , j and idownstream , j identification
Search the longest path in the topology of
the network and calculate l LONG
are the SFB components of transient zero-sequence current, Real-time acquisition of the zero-
which can be monitored with the D-PMUs deployed at the sequence voltage in the D-PMUs
Caiculate k by (26)
upstream node and downstream nodes of the line section j.
N
Denote the transient energy in line section j as: U0 kU N
Combined with the grounding mode
of the network, the frequency range
TF of the SFB is defined by (16)
 SFB  2
ESFB , j   i
0
S, j (t )  dt (19) Y

Single phase-earth fault occurs Design the digital filter

where TF is the duration of the fault transient process. Then,

an index for faulty line section identification is defined as
Calculate the transient energy
 
FLSI  (max ESFB , j )   ESFB , j  max ESFB , j  (20) Get waveforms of the transient zero-sequence currents at all nodes from D-PMUs
jM jM
 jM 
According to (18), when the fault occurs in the line section F, Data preprocessing

the relation of the SFB components between healthy line

sections and faulty line section can be expressed as Calculate is , j (k ) in line-section j by (27)

2
 SFB 
   iS , j  
iS , F
SFB 
2
(21) ( SFB )
Extract iS , j (k ) in line-section j by the digital filter

 jM , j  F 
Therefore, the following in-equation holds for the Calculate ESFB , j in line-section j by (28)

instantaneous data obtained with the D-PMUs

 i    i  
2 2
SFB
S ,F  SFB
S, j (22) Faulty Line-section identification
jM , j  F
ESFB ,m  max ESFB , j
jM
And in the transient process after the fault, it holds that
TF TF
 SFB 
  i SFB 
2 2

 i
0
S ,F (t )  dt 
jM , j  F 0
S, j (t )  dt (23) FLSI  1?
N

Y
As a result, when a single phase to ground fault occurs in line
Y N
section F , the correspondent FLSI is greater than 1. SGN  1?

But if the length of one line section is almost equal to the Fault occurred in the
Fault occurred in
sum length of the rest line sections, FLSI may still be greater line-section m substation bus

than 1 when the fault occurs at the substation bus bar. In order
Fig. 9. Flow chart of the proposed fault location scheme
to avoid the misjudgment, an auxiliary criterion is added: 2) Single phase to ground fault detection. The instantaneous
 TF   

value of zero-sequence voltage acquired by the D-PMUs is
SGN  sgn   iS SFB 
, F (t )  iS SFB 
, j (t ) dt  (24)
  monitored, the fault is considered detected if it is greater than
0 jM , j  F  
the threshold kU N , where U N is the rated voltage and k is in
where sgn is a function for calculating numeric symbols.
the range of 0.05 ~ 0.10.
According to (18), when the fault occurs in the line section 3) Transient energy calculation. The waveforms of the zero-
F, SGN is -1. Therefore, the necessary and sufficient

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

sequence currents during the fault are obtained by D-PMUs between healthy line sections and faulty line section.
and uploaded to the distribution control center. Before being Table II shows the faulty line section detection results for
used for fault location, the data is preprocessed to eliminate different faulty line sections and faulty distance from the bus.
the impact of outliers in the data. For each sampling point k , The initial phase angle, fault resistance, fault phase and SNR is
the instantaneous current difference iS , j  k  can be obtained as 0°, 0Ω, phase A and 5dB, respectively. The bold line of Table
II, named Case B that the fault occurs at the bus, is taken as an
iS , j  k   iupstream , j  k    idownstream , j  k  (27) example to illustrate the results. The transient energy is shown
iS , j  k  can also be seen as the zero-sequence current flowing in Fig. 12, and there is no obvious difference between the
to the earth from line section j ( j  M ). And the concerned transient energy of different line sections. The fault at the bus
bar is also located correctly. It can be seen from Table II that
( SFB )
component iS,j  k  of line section j is extracted by a digital the method can achieve faulty line section identification in
filter according to the value of the SFB. Then the transient various faulty distances, even for those at the bus bar, at the
energy in line section j is calculated by end of the line as well as the laterals connected to T nodes.
N
ESFB , j   iS , j  (k ) 
2
SFB DG1
(28) k1 Fault Location
8
822 DG4 848

k 1 846
D-PMU 820 15
4) Faulty line section identification. The maximum of ESFB , j is 1 Line-section 7 844
DG2
864
calculated by k2 k3 6
818
k4 13
842
14 k5
ESFB , m  max ESFB , j
802 806 2 808 4 812 814 850 824 834 860 836 840
(29) k1
1 5 816 826
858
jM 832
888 890 16 862
3
800 9
And the fault occurs in the line section m if it satisfies 810 852 DG5 838

 FLSI  ESFB , m  ESFB , j  1 12

 jM , j  m DG3

 828 10 830 854 11 856

   N 
(30)
 SGN  sgn    iS( SFB  S, j
Fig. 10. Single-line diagram of IEEE-34 system
)
( k )  i ( SFB )
( k )    1

,m
 k 1  jM , j  m   TABLE I
LOCATION RESULTS FOR DIFFERENT F AND FR
For the distribution networks including several feeders, the
Faulty Location
proposed method can be used to determine the faulty feeder F (°) FR (  ) SNR (dB) FLSI SGN
Line Section  Results
and the faulty line section simultaneously.
0 0 5 9.26 -1 5
0 0 -1 9.15 -1 5
IV. CASE STUDY
0 50 5 8.92 -1 5
A. Simulation Test 5 0 500 5 9.05 -1 5
0 500 -1 9.31 -1 5
The IEEE 34 node distribution benchmark network is used
60 0 5 8.73 -1 5
to verify the proposed scheme, which is modeled and modified
90 0 5 9.17 -1 5
with PSCAD/EMTDC software. All loads are three-phase
10 -3 10 -3
balanced constant power loads. There are five DGs in the Line-section 4
6
Line-section 4
2 4
network. Three photovoltaic power stations are connected to
Current (kA)

Line-section 5 Line-section 5
Current (kA)

Line-section 6 2 Line-section 6
0
node 822, node 826 and node 848, of which the power 0
-2
supplies are 0.15MW, 0.3MW and 0.2MW. Two wind -2
-4 -4
generators are added in node 838 and node 856, of which the -6
1.995 2 2.005 2.01 2.015 2.02 1.995 2 2.005 2.01 2.015 2.02
power supplies are 0.3MW and 0.2MW. 16 D-PMUs are Time (s) Time (s)
supposed randomly installed along the primary feeder buses as (a) Original signals (b) Polluted signals
shown in Fig. 10. The sampling frequency is set as 10 kHz. 1.5
10 -3
120
A single phase to ground fault is set to occur at k3, which is 1 Line-section 4
100
Current (kA)

Line-section 5
ESFB, j ( A2 s )

0.5 Line-section 6
22.57 km away from the bus bar. Table I shows the results of 0
80
60
-0.5
different situations with different initial angles ( F ), -1 40
-1.5 20
grounding resistance ( FR ), and SNR. The proposed scheme is -2 0
1.995 2 2.005 2.01 2.015 2.02 1 2 3 4 5 6 7 8 9 10 11121314 15 16
tested using the fault scenarios listed in Table I. All the faulty Time (s) Line section
line section location results are correct, which show the
proposed method suitable for the faults in different initial (c) SFB components (d) Transient energy distribution
Fig. 11. Signals and transient energy distribution of Case A
angles and grounding resistances. Fault Case A, as in the
The influence of DG is also taken into consideration. When
bolded line in Table I, is taken as an example to illustrate the
a single phase to ground fault occurs at k3, the initial phase
results in detail. Fig. 11(a) and Fig. 11(b) are the original and
angle, fault resistance, fault phase and SNR is 0°, 0Ω, phase A
polluted transient zero-sequence currents, respectively. The
and 5dB. Table III shows the faulty line detection results for
SFB components are extracted, which is shown in Fig. 11(c).
different allocation of DGs. Table IV shows the faulty line
And the transient energy distribution of all line sections is
detection results for different power injected by the DGs. In
shown in Fig. 11(d), which give a good discrimination degree

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

Table IV, there is no change in Test Number 1. The power 2 0.1 MW (DG3) 8.72 -1 5
supply of DG3 is changed to 0.1MW and 0.5MW in Test 3 0.5 MW (DG3) 9.03 -1 5
Number 2 and Test Number 3. In the Test Number 4, the 4 0.4 MW (DG4) 8.79 -1 5
power supply of DG4 is changed to 0.4MW. In the Test 0.1 MW (DG3) and 0.4 MW
5 9.08 -1 5
Number 5, the power supply of DG3 and DG4 is changed to (DG4)
0.1MW and 0.4MW. The proposed method can determine the
TABLE V
fault line sections for different allocation of DGs and different
LOCATION RESULTS FOR RESONANT GROUNDED SYSTEM WITH
power injected by the DGs. DIFFERENT COMPENSATION RATE
Table V shows the faulty line detection results for different Faulty Location
CP (%) FLSI SGN
compensation degrees ( CP ) of the Peterson coil. It shows that Line section Results
fault line sections can be correctly identified in a wide 5 -8 9.88 -1 5
compensation range. -5 9.84 -1 5
The effect of different load type is evaluated, as shown in 5 9.72 -1 5
Table VI. When a single phase to ground fault occurs at k5, 8 9.81 -1 5
the initial phase angle, fault resistance, fault phase and SNR is TABLE VI
0°, 0Ω, phase A and 5dB. The load conditions considered in LOCATION RESULTS FOR DIFFERENT LOADS
Test Number 1 is that all loads are three-phase balanced Faulty
FLSI SGN
Location
Number Load Condition
constant power loads. In Test Number 2, the loads connected Line Section Results
to node 836, node 838, node 840, and node 860 are removed. 1 No change 9.08 -1 16
In Test Number 3, the load connected to the node 836 is 2 Loads removed 8.83 -1 16
16
changed to a two-phase constant power load, and the loads 3 Unbalanced Loads 8.93 -1 16
connected to the node 838, node 840 and node 860 are 4 Nonlinear Loads 9.11 -1 16
changed to one-phase constant power loads. In Test Number 4, When analyzing the robustness to measurement, the
the loads connected to the node 836 and node 840 are changed influence of noise and outliers have been taken into
to three-phase balanced nonlinear power loads. According to consideration. When SNR is -1dB, the fault location can still
the fault location results in Table VI, it can be seen that there be accurately located as shown in Table I, which proves that
is no influence on the FLSI and the fault location results for the method has good noise immunity. The impact of the
different load types. outliers also has been analyzed. When a single phase to ground
8 fault occurs at k3, the initial phase angle, fault resistance, fault
7 phase and SNR is 0°, 0Ω, phase A and 5dB. The results are
ESFB, j ( A2 s )

6
5 presented in Table VII. There is no bad data for Test Number
4
3 1. In Test Number 2, part of the data measured at node 806 is
2 lost during transmission, and part of the data measured at node
1
0 828 has impulse noise in Test Number 3. In Test Number 4,
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Line-section part of the data measured at node 814 during transmission is
Fig. 12. Transient energy distribution of Case B lost, and part of the data measured at node 860 has impulse
TABLE II noise. Taking Test Number 4 in Table VII as Case C, the
LOCATION RESULTS FOR DIFFERENT LINE SECTIONS original waveform, the polluted waveform and the corrected
Faulty Distance Location waveform of the signals measured at node 814 and node 860
FLSI SGN
Line Section (km) Results are as shown in Fig. 13(a) and Fig. 13(b). The transient energy
bus bar 0 0.424 1 bus bar distribution of the system is as shown in Fig. 13(c). Comparing
1 0.79 9.73 -1 1 the corrected waveform with the original waveform, it is seen
5 22.57 9.13 -1 5 that the outliers can be effectively corrected by using the
9 35.76 8.91 -1 9 robust local regression smoothing filtering method [35]. Using
16 57.68 8.77 -1 16 the corrected signals, the transient energy of each line section
TABLE III are calculated and the faulty line section can be accurately
LOCATION RESULTS FOR DIFFERENT ALLOCATION OF DGS located. After the data preprocessing, the outliers will not
Faulty Allocation of Location affect the location results.
FLSI SGN
Line Section DGs Results TABLE VII
5 DG1~DG5 9.37 -1 5 LOCATION RESULTS WITH OUTLIERS
DG2~DG5 9.52 -1 5 Faulty Line Location
Number Outliers FLSI SGN
9 DG1~DG5 9.08 -1 9 Section Results
DG2~DG5 9.44 -1 9 1 No Outliers 9.15 -1 5
2 Impulse Noise 8.67 -1 5
TABLE IV
3 5 Lost Data 8.88 -1 5
LOCATION RESULTS FOR DIFFERENT POWER INJECTED BY THE DGS
Faulty Line Location Impulse Noise
Number DGs’ Power Adjustment FLSI SGN 4 9.10 -1 5
Section Results and Lost Data
1 5 No change 9.11 -1 5

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

20 20
Lost Polluted waveform
network in China, the structure of the network is shown in Fig.
Original waveform
10 data Corrected waveform 14. The transformer is 0.4/10kV with Y-delta connection, and
10 Corrected waveform
Current (A)

Current (A)
0 0
the neutral point of the grounding transformer has two working
modes: ungrounded or via the Peterson coil. There are five
-10 -10
feeders, including cable, overhead line and hybrid feeder. Arc
-20
0 2 4 6 8 10
-20
0 2 4 6 8 10
suppression coil is WZXC-1000/10.5, the parameters are as
Time (ms) Time (ms)
follows: rated capacity 1000kvar, rated current 165A, rated
(a) Original waveform, polluted waveform and corrected waveform (node 814) frequency 50Hz. The parameters of the grounding transformer
10 1
Polluted waveform Original waveform are as follows: rated capacity 1200/200kVA, rated voltage
8 0.5
Corrected waveform Corrected waveform
10.5±2×2.5%/0.4kV, rated current 65.98/288.68A, neutral
Current (A)

Current (A)
6
0
4 Impulsive current 165 A, rated frequency 50Hz. The D-PMUs are
noise -0.5
2 deployed at the beginning of all the feeders and a D-PMU is
-1
0 installed at the junction of the overhead line and the cable in
-2 -1.5
0 2 4 6 8 10 0 2 4 6 8 10 feeder 2. The sampling frequency of the D-PMUs is 200 points
Time (ms) Time (ms)
per cycle. In the test, the fault location procedure is triggered
(b) Original waveform, polluted waveform and corrected waveform (node 860) according to the zero-sequence voltage at the bus. Whether it

180 is a bus bar fault or a line fault is firstly identified and then the
ESFB, j ( A s )

160
140
faulty line section is located via (30).
2

120 0.5 km overhead line

100 k1 Fault Location
80 0.5 km overhead line 3 km cable
60 D-PM U
k1 1 k2 2
40 1 Line-section
20 15 km cable
0 Cable
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Overhead Line 3 km cable
Line section
1.5 km overhead line
(c) Transient energy distribution
Fig. 13. Signals and transient energy distribution of Case C 1.5 km overhead line

The robustness to parameter uncertainty is also discussed. Fig. 14. The structure of the 10kV network for fault experiment
According to the parameters given by the system, the A various of single phase to ground fault experiments are
frequency range of the SFB is calculated to be 200-770Hz, and implemented on feeder 2, including different neutral grounded
the parameters of the system are changed but the SFB is not
mode, different grounding resistance and different
recalculated. When a single phase to ground fault occurs at k3,
compensation degree. Fault occurs at k1 on the overhead line
the initial phase angle, fault resistance, fault phase and SNR is
of feeder 2 and k2 on the cable of feeder 2. k1 is 150 m away
0°, 0Ω, phase A and 5dB. There is no change in the parameters
from the bus, and k2 is 600 m away from the bus. Some photos
of the line in Test Number 1. In Test Number 2 and Test
of the on-field fault experiments are shown in Fig. 15. The
Number 3, the line resistance of the entire system was set to
faulty feeder selection and the faulty line section location are
80% and 120% of the original resistance. In Test Number 4
implemented separately. The results are presented in Table IX.
and Test Number 5, the line inductance of the entire system
was set to 80% and 120% of the original inductance. In Test Taking Case D, as in the bolded line in Table IX, as an
Number 6 and Test Number 7, the shunt capacitance of the example to illustrate the effectiveness of the proposed method
entire system was set to 80% and 120% of the original in real system, Fig. 16 shows the related waveform for Case D.
capacitance. The results are presented in Table VIII. Fig. 16(a) and Fig. 16(b) are the original transient zero-
According to the location results in the Table VIII, it can be sequence currents and the SFB components, respectively.
seen that there is no influence on the FLSI and the results when Transient energy of every line and every line section is shown
any parameter of the entire network line changes by 20%. in Fig. 16(c). The transient energy of faulty line is larger than
the transient energy of other lines, and the transient energy of
TABLE VIII
LOCATION RESULTS WITH DIFFERENT PARAMETERS
faulty line section is larger than the transient energy of other
Faulty Location line sections, so the faulty feeder and faulty line section can be
Number Parameters FLSI SGN identified based on the transient energy. The fault location
Line Section Results
1 No change 9.17 -1 5 results in Table IX are correct, which indicates that the
2 80% resistance 8.95 -1 5 proposed scheme can detect faulty line section effectively in
3 120% resistance 8.98 -1 5 radiation network including several feeders.
4 5 80% inductance 7.21 -1 5
5 120% inductance 9.18 -1 5
6 80% capacitance 9.57 -1 5
7 120% capacitance 9.01 -1 5

B. On field fault experiments test

We carried out on field test at a real 10kV distribution

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

Method A: It uses the amplitude and phase of the zero-

sequence admittance of the line section to identify the faulty
line section, and the zero-sequence admittance is obtained by
dividing zero-sequence current and zero-sequence voltage.
The comparison of identification accuracy for Method A
Fig. 15. The on-field single phase to ground fault experiments and the proposed method in different conditions is shown in
TABLE IX Table X. In Table X (a), the fault conditions of SNR and initial
LOCATION RESULTS OF FIELD TEST phase angle are 20 dB and 0°, respectively. In Table X (b), the
Faulty
Results
fault scenarios of SNR, initial phase angle and grounding
Fault
Neutral Faulty Feeder Results Line
(Faulty resistance are 20 dB, 0° and 5  , respectively. In order to
Grounding Rf Selection (Faulty Section
Location Line verify the noise immunity of the method, each group of the
Mode Feeder) Detection
FLSI SGN FLSI
Section) simulations was repeated 1000 times with random noises.
0 2.01 -1 2 48.75 1 TABLE X
125 1.72 -1 2 59.65 1 (a) Accuracy Comparison for Different Location and FR
Ungrounded 500 1.68 -1 2 4.33 1
Faulty Accuracy of Accuracy of
1000 1.59 -1 2 4.31 1 FR (  )
2000 1.54 -1 2 3.57 1 Line section Method A the proposed Method
k1
0 1.81 -1 2 11815.85 1 0 100% 100%
Resonant 5
125 1.72 -1 2 42978.17 1 500 100% 100%
grounded
500 1.37 -1 2 92.87 1 0 100% 100%
(Over- 16
1000 1.95 -1 2 1.21 1 100% 100%
compensation) 500
2000 1.46 -1 2 1.44 1
0 3.46 -1 2 1087.65 2 (b) Accuracy Comparison for Different Location and CP
167 2.48 -1 2 897.33 2 Faulty Accuracy of Accuracy of
CP (%)
Ungrounded 500 2.56 -1 2 778.50 2 Line section Method A the proposed Method
1000 2.88 -1 2 411.22 2 97.1% 100%
5
2000 1.99 -1 2 312.45 2 5
8 97.2% 100%
0 1.81 -1 2 635.89 2
Resonant 66.1% 100%
167 1.79 -1 2 418.95 2 5
grounded 16
k2 500 1.52 -1 2 107.18 2 8 64.9% 100%
(Over-
1000 1.88 -1 2 23.06 2
compensation)
2000 1.54 -1 2 2.59 2 The results show that Method A and the proposed method
0 1.83 -1 2 570.49 2 both correctly detect the single phase-earth fault line section in
Resonant
grounded
167 1.95 -1 2 1543.84 2 the neutral isolated system. But Method A produces some
500 2.03 -1 2 899.74 2 misjudgments in resonant grounded system, especially when
(Under-
1000 2.25 -1 2 11.89 2
compensation) the fault is at the end of the line. The reason is that the steady
2000 2.15 -1 2 2.51 2
4 1
state zero-sequence current is weak in resonant grounded
Line section 1 Line section 1 system, the weak steady-state zero-sequence current is
Line section 2 Line section 2
Current (A)

Current (A)

2 0.5 disturbed and the phase of the zero-sequence admittance has

0
large errors. The bold part of Table X(b) is taken as an
0
example, and 1 group phasors of zero-sequence admittance
-2 -0.5 based on Method A are shown in Fig.17, which indicate that
0 20 40 60 0 20 40 60 the noises lead to large errors in the phase of zero-sequence
Time (ms) Time (ms)
admittance. The proposed method is immune to strong noises.
(a) Original signals (b) SFB components
90
50 2.5e-05
120 60
Transient Energy

2e-05
40 1
2
（A2·s）

1.5e-05
30 150 4e-06 150 30 3
1e-05 4
3e-06 5
20 2e-06 5e-06 6
7
10 180 180 8
9
10
0 11
210 12
13
210 330 14
15
(c) Transient energy of Case D 16

240 300
Fig. 16. Signals and transient energy of Case D 270

Fig. 17. Phasors of Zero-sequence Admittance

V. COMPARISON WITH OTHER METHODS Method B: It uses the corrected correlation coefficient ( Ccor )
In order to present the reliability advantage of faulty line of the transient zero-sequence currents to identify the faulty
section identification method with strong noise background, line section, of which the threshold is 0.5. However, Method B
we compared this method with three methods introduced in is not suitable for some scenarios. Using the benchmark
[12], [36] and [16], which have been verified in distribution network shown in Fig. 10, more D-PMUs are deployed in the
network and are denoted as Method A, B and C, respectively. node 810, node 822, node 856, node 848 and node 840. The

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

faulty conditions of SNR, initial phase angle and grounding doesn’t work properly sometime, especially when the
resistance are 20 dB, 0°and 5  . grounding resistance is bigger than 200 Ω. The noise in the
As shown in Table XI, method B is completely ineffective actual distribution grid is very strong, resulting in incorrect
because the zero-sequence current near the substation bus bar location results if the grounding resistance is large. The
and the end of the line. Meanwhile, the proposed method can proposed method is more reliable in detecting the faulty feeder
determine the faulty line section correctly. Compared with and identifying the faulty line section in the actual distribution
method B, the proposed method is suitable for more fault network.
scenarios, and needs less measuring equipment.
TABLE XI VI. CONCLUSIONS
LOCATION RESULTS OF METHOD B
Fault True or This paper proposes an integrated technique based on the D-
Ccor
Location False PMUs to identify the faulty line section for neutral non-
[0.0223, 0.8492, -0.0297, 0.8353, 0.8597, effectively grounded MV networks. The paper investigates the
(0.7534, 0.6224), 0.9448, 0.0169, 0.9533, determination of the SFB within which the transient energy of
k2 (0.8336, 0.8872), 0.0088, 0.9440, 0.9935, False
the monitored signal is far greater than the energy of the noise,
(0.9912, 0.9920), 0.0192, -0.0035]
the transient energy of faulty line section is proven to be
[0.0248, 0.9466, -0.0222, 0.9028, -0.2986,
(0.5926, 0.3955), 0.9107, 0.0090, 0.9151, greater than the energy of the healthy line sections. A method
k3 (0.6823, 0.7832), -0.0240, 0.8839, 0.9863, False based on a combined criterion is proposed, and the
(0.9812, 0.9829), 0.0126, -0.0201] implementation of the scheme is illustrated. Both the
[0.0572, 0.9177, 0.0368, 0.8322, 0.8291, simulation and on-field fault experiment results show that the
(0.5457, -0.5983), 0.8288, -0.0252, -0.4980,
k4 False proposed method is not affected by strong background noise,
(0.6027, 0.7557), 0.0311, 0.8457, 0.9802,
(0.9716, 0.9742), 0.0240, -0.0332] fault locations, load types, grounding resistances, inception
[0.0133, 0.9691, 0.0407, 0.9316, 0.9239, angles, locations and power supply of DGs. And the method is
(0.7803, -0.8273), 0.9371, -0.0029, 0.9396, also robust to line parameters and outliers. Three methods are
k5 (-0.5335, 0.8007), 0.0052, 0.8152, 0.9594, False
compared with the proposed method to prove that the
(-0.2998, 0.9245), 0.0137, -0.0022] proposed method is more accurate and applicable.
Method C: The protection criterion established by using the
slope relation matrix can be applied to detect the faulty feeder VII. EFERENCES
of single phase to ground faults. The threshold is set equal to [1] S. Jamali and A. Bahmanyar, "A new fault location method for
−0.5. The location results for Method C in different conditions distribution networks using sparse measurements," International
are shown in Table XII using the field data in Table IX. Journal of Electrical Power & Energy Systems, vol. 81, pp. 459-468,
Oct. 2016.
TABLE XII
[2] H. Chen, P. D. S. Assala, Y. Cai, and P. Yang, "Intelligent transient
LOCATION RESULTS OF METHOD C overvoltages location in distribution systems using wavelet packet
Neutral decomposition and general regression neural networks," IEEE Trans.
Fault
Grounding Rf i [ 1 2 3 4 5 6 ] Result Industrial Informatics, vol. 12, no. 5, pp. 1726-1735, Oct. 2016.
Location
Mode [3] D. Topolanek, M. Lehtonen, M. R. Adzman and P. Toman, "Earth fault
0 [0.16, -1.82, 0.55, 0.50, 0.56, 0.05] Right location based on evaluation of voltage sag at secondary side of medium
125 [-0.01, -0.35, 0.05, 0.08, 0.14, 0.09] False voltage/low voltage transformers," IET Gener. Transmission &
Ungrounded 500 [0.19, -0.54, 0.19, 0.32, -0.20, 0.03] Right Distribution, vol. 9, no. 14, pp. 2069-2077, Nov. 2015.
1000 [-0.07, -0.05, -0.07, 0.05, -0.24, 0.37] False [4] Y. Xue, L. Guo, L. Zhang, B. Xu and T. Li, "Selection Problems of
2000 [0.02, -0.08, 0.27, -0.28, -0.05, 0.13] False Neutral Grounding Mode in Active Distribution Networks. Automation
k1
0 [0.59, -1.65, 0.37, 0.34, 0.05, 0.29] Right of Electric Power Systems," vol. 39, no. 13, pp. 129-126, May, 2015. (in
125 [0.25, -1.12, 0.05, 0.30, 0.34, 0.19] Right Chinese)
Over-
compensation
500 [-0.01, -0.14, -0.04, -0.01, 0.05, 0.15] False [5] P. Liu and C. Huang, “Detecting single-phase-to-ground fault event and
1000 [0.08, 0.00, -0.09, 0.17, 0.05, -0.22] False identifying faulty feeder in neutral ineffectively grounded distribution
2000 [0.16, -0.30, -0.14, 0.21, 0.04, 0.03] False System,” IEEE Trans. Power Delivery, vol. 33, no. 5, 2265-2273, Oct.
0 [0.65, -1.23, 0.10, 0.16, 0.53, -0.21] Right 2018.
167 [0.22, -0.26, 0.09, 0.17, 0.07, -0.29] False [6] D. Griffel, V. Leitloff, Y. Harmand and J. Bergeal, "A new deal for
Ungrounded 500 [0.29, -0.80, 0.08, 0.21, -0.06, 0.30] Right safety and quality on MV networks," IEEE Trans. Power Delivery, vol.
1000 [0.17, -0.26, 0.25, 0.13, -0.11, -0.18] False 12, no. 4, pp. 1428-1433, Oct. 1997.
2000 [0.10, -0.33, 0.06, 0.07, -0.01, 0.12] False [7] Z. Zhang, Z. Wang, and K. Li, "The summarization of fault line
0 [0.64, -1.77, 0.47, 0.41, 0.43, -0.18] Right selection of Small current grounding system," International Symposium
167 [0.30, -0.57, 0.17, 0.18, -0.31, 0.23] Right on Vehicle, Mechanical, and Electrical Engineering (ISVMEE), 21-22,
Over-
k2 500 [0.11, 0.06, -0.18, -0.03, 0.04, 0.00] False Dec, 2013, pp. 1775-1778, Taiwan, China.
compensation
1000 [-0.04, -0.11, 0.13, -0.03, -0.00, 0.07] False [8] F. Han, Y. Zhang, S. Hao, and X. Zhao, "Research on single-phase-to-
2000 [0.35, -0.13, -0.59, 0.23, 0.10, 0.04] False earth fault location based on Hilbert-huang transform," 2016 IEEE
0 [0.58, -1.76, 0.62, 0.64, -0.26, 0.18] Right Advanced Information Management, Communicates, Electronic and
167 [0.26, -0.32, 0.34, 0.34, 0.21, -0.84] False Automation Control Conference (IMCEC), 3-5 Oct. 2016, pp. 1338-
Under-
500 [0.08, -0.64, 0.26, 0.37, 0.08, -0.15] Right 1441, Xi’an, China.
compensation
1000 [0.13, -0.26, 0.01, 0.13, -0.03, 0.02] False [9] X. Zeng, K. Li, W. Chan, S. Su and Y. Wang, "Ground-Fault Feeder
2000 [0.21, -0.33, 0.32, -0.24, -0.06, 0.10] False Detection With Fault-Current and Fault-Resistance Measurement in
Because the data processing method of Method A is not Mine Power Systems," IEEE Trans. Industry Applications, vol. 44, no.
2, pp. 424-429, Mar. 2008.
mentioned in [16], the results show that Method C can detect
the faulty feeder if the grounding resistance is 0 Ω but it

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

[10] X. Lin, J. Huang and S. Ke, "Faulty Feeder Detection and Fault Self- [29] H. Zhang, Z. Jin and Y. Liu, "Wide-area Measurement System Light
Extinguishing by Adaptive Petersen Coil Control," IEEE Trans. Power and Its Application in China," Automation of Electric Power Systems,
Delivery, vol. 26, no. 2, pp. 1290-1291, Apr. 2011. vol. 38, no. 22, pp. 85-90, Nov. 2014. (in Chinese)
[11] Z. Zhang, X. Liu and Z. Piao, "Fault line detection in neutral point [30] Z. Jin, H. Zhang and C. Li, "WAMS light and its deployment in China,"
ineffectively grounding power system based on phase-locked loop," IET 2015 5th International Conference on Electric Utility Deregulation and
Gener. Transmission & Distribution, vol. 8, no. 2, pp. 273-280, Feb. Restructuring and Power Technologies (DRPT), Changsha, 2015, pp.
2014. 1373-1376.
[12] L. Zhang, H. Gao, B. Xu and Y. Xue, "Fault location method based on [31] A. v. Meier, E. Stewart, A. McEachern, M. Andersen, and L.
zero sequence admittance measurement in non-effectively earthed Mehrmanesh, "Precision Micro-Synchrophasors for Distribution
system," IEEE PES Innovative Smart Grid Technologies, Tianjin, 21-24 Systems: A Summary of Applications," IEEE Transactions on Smart
May 2012, pp. 1-4. Grid, vol. 8, no. 6, pp. 2926-2936, Nov. 2017.
[13] X. Dong and S. Shi, "Identifying Single-Phase-to-Ground Fault Feeder [32] A. von Meier, D. Culler, A. McEachern, R. Arghandeh. "Micro-
in Neutral Noneffectively Grounded Distribution System Using Wavelet Synchrophasors for Distribution Systems," 2014 IEEE Pes Innovative
Transform," IEEE Trans. Power Delivery, vol. 23, no. 4, pp. 1829-1837, Smart Grid Technologies Conference (ISGT), 19-22 Feb. 2014,
Oct. 2008. Washington, DC, United states.
[14] C. Zhou, Q. Shu and X. Han, "A single-phase earth fault location [33] Y. Xue, B. Xu, Y. Chen, Z. Feng, and P. Gale, "Earth fault protection
scheme for distribution feeder on the basis of the difference of zero using transient signals in non-solid earthed network," in Proceedings.
mode traveling waves," International Trans. on Electrical Energy International Conference on Power System Technology, 14-22 Feb.
Systems, vol. 27, no. 5, pp. e2298. 2002, pp. 1763-1767, Kunming, China.
[15] Y. Tao, and Y. Juan, "A Survey on Single-Phase Faulty Line Selection [34] M. F. Abdel-Fattah, M. J. J. o. E. Lehtonen, and P. Engineering, "A new
Techniques for Small Current Grounding System in Medium-voltage transient impedance-based algorithm for earth fault detection in medium
Grid," Applied Mechanics and Materials, vol. 492, pp. 407-411, Jan. voltage networks," Journal of Energy and Power Engineering, vol. 6, no.
2014. 2, pp. 240-249, Jun. 2012.
[16] Y. Wang, Y. Huang, X. Zeng, G. Wei, J. Zhou, T. Fang, and H. Chen, [35] H. Zhang, Z. Jin and V. Terzija, "An Adaptive Decomposition Scheme
"Faulty feeder detection of single phase-earth fault using grey relation for Wideband Signals of Power Systems Based on the Modified Robust
degree in resonant grounding system," IEEE Trans. Power Delivery, vol. Regression Smoothing and Chebyshev-II IIR Filter Bank," IEEE Trans.
32, no. 1, pp. 55-61, Feb. 2017. Power Delivery, vol. 34, no. 1, pp. 220-230, Feb. 2019.
[17] Y. Fang, Y. Xue, H. Song, T. Guan, F. Yang and B. Xu, "Transient [36] T. Li and C. Huang, "New transient fault location method in non-solidly
Energy Analysis and Faulty Feeder Identification Method of High earthed system for distribution network," CIRED - Open Access
Impedance Fault in the Resonant Grounding System". Proceedings of Proceedings Journal, vol. 2017, no. 1, pp. 1330-1333, Oct. 2017.
CSEE, vol. 38, no. 19, pp. 5636-5645, Oct. 2018. (in Chinese)
[18] X. Wang, J. Gao, M. Chen, X. Wei, Y. Wei and Z. Zeng, "Faulty Line Xuewen Wang received B.E degree in electrical
Detection Method Based on Optimized Bistable System for Distribution engineering from Shandong University. He is now a post-
Network," IEEE Trans. on Industrial Informatics, vol. 14, no. 4, pp. graduate student majoring in electrical engineering in
1370-1381, Apr. 2018. Shandong University. His research interest includes fault
[19] M. Pourahmadi-Nakhli and A. A. Safavi, "Path Characteristic location for distribution networks, power system
Frequency-Based Fault Locating in Radial Distribution Systems Using simulation.
Wavelets and Neural Networks," IEEE Trans. Power Delivery, vol. 26,
no. 2, pp. 772-781, Apr. 2011.
[20] X. Lin, J. Huang and S. Ke, "Faulty feeder detection and fault self-
extinguishing by adaptive petersen coil control," IEEE Trans. Power Fang Shi (M’ 10) received the Ph.D. degree from
Delivery, vol. 26, no. 2, pp. 1290-1291, Apr. 2011. Shanghai Jiao Tong University, Shanghai, China, 2014.
[21] W. Wang, K. Zhu, P. Zhang and W. Xu, "Identification of the Faulted He is currently a lecturer in the Key Laboratory of Power
Distribution Line Using Thyristor-Controlled Grounding," IEEE Trans. System Intelligent Dispatch and Control of the Ministry of
Power Delivery, vol. 24, no. 1, pp. 52-60, Jan. 2009. Education (Shandong University), P. R. China. His
[22] P. Wang, B. Chen, H. Zhou, T. Cuihua and B. Sun, "Fault Location in research interests include theory and applications of D-
Resonant Grounded Network by Adaptive Control of Neutral-to-Earth PMU, and power system stability and control.
Complex Impedance," IEEE Trans. on Power Delivery, vol. 33, no. 2,
pp. 689-698, Apr. 2018.
[23] C. Huang, T. Tang, Y. Jiang, L. Hua and C. Hong, "Faulty feeder Hengxu Zhang (M’06) received his B.E. degree in
detection by adjusting the compensation degree of arc-suppression coil electrical engineering from Shandong University of
for distribution network," IET Gener. Transmission & Distribution, vol. Technology, in 1998, and his M.S. and Ph.D. in electrical
12, no. 4, pp. 807-814, Feb. 2018. engineering from Shandong University, in 2000 and 2003,
[24] X. Zeng, Y. Xu,, and Y. Wang, "Some novel techniques for insulation respectively. He is now a professor with the Key
parameters measurement and Petersen-Coil control in distribution Laboratory of Power System Intelligent Dispatch and
systems," IEEE Trans. Industrial Electronics, vol. 57, no. 4, pp. 1445- Control of the Ministry of Education (Shandong
1451, Apr. 2010. University), P. R. China. His main research interests are
[25] P. Wang, B. Chen, H. Zhou, T. Cuihua and B. Sun, "Fault Location in power system security and stability assessment, power system monitoring and
Resonant Grounded Network by Adaptive Control of Neutral-to-Earth numerical simulation.
Complex Impedance," IEEE Trans. Power Delivery, vol. 33, no. 2, pp.
689-698, Apr. 2018. Qiuwei Wu (M’08–SM’15) received the Ph.D. degree in
[26] J. Zhou, and K. Cao, "Locating method of single-phase grounding fault power system engineering from Nanyang Technological
in small current grounding system based on distribution automation," University, Singapore, in 2009. He is an Associate
International Conference on Mechatronics and Industrial Informatics Professor with the Department of Electrical Engineering,
(ICMII), Applied Mechanics and Materials, vol, 321, pp. 1423-1428, Technical University of Denmark. He is also a Visiting
Jun. 2013. Professor named by Y. Xue, with Shandong University,
[27] K. Jia, Z. Ren, L. Li, Z. Xuan and D. Thomas, "High-frequency China. His research interests are integration of wind
transient comparison based fault location in distribution systems with power and electric vehicle into power systems, active
DGs, " IET Gener. Transmission & Distribution, vol. 11, no. 16, pp. distribution networks, and real time power system
4068-4077, Nov. 2017. simulation and analysis.
[28] K. Jia, C. Gu, T. Bi and D. Thomas, "Sparse Voltage Amplitude
Measurement Based Fault Location in Large-Scale Photovoltaic Power
Plants, " Applied Energy, vol. 211. pp. 568-581, Feb. 2018.

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSG.2019.2938667, IEEE
Transactions on Smart Grid

Vladimir Terzija (M’95, SM’2000, F’16) was born in

Donji Baraci (former Yugoslavia). He received the Dipl-
Ing., M.Sc., and Ph.D. degrees in electrical engineering
from the University of Belgrade, Belgrade, Serbia, in 1988,
1993, and 1997, respectively. He is the Engineering and
Physical Science Research Council Chair Professor in
Power System Engineering with the School of Electrical
and Electronic Engineering, The University of Manchester,
Manchester, U.K., where he has been since 2006. From 1997 to 1999, he was
an Assistant Professor at the University of Belgrade, Belgrade, Serbia. From
2000 to 2006, he was a Senior Specialist for switchgear and distribution
automation with ABB AG, Inc., Ratingen, Germany. His current research
interests include smart grid application of intelligent methods to power
system monitoring, control, and protection; wide-area monitoring, protection,
and control; switchgear and fast transient processes; and digital signal
processing applications in power systems. Prof. Terzija is the Editor-in-Chief
of the International Journal of Electrical Power and Energy Systems, an
Alexander von Humboldt Fellow, as well as a DAAD and Taishan Scholar.

Wei Xie received the B.Sc. degree in Shanghai Jiaotong

University in 1990. He is currently the chief engineer of
the State Grid Shanghai Municipal Electric Power
Company, his current research interests include renewable
energy integration of smart grid and power storage
technology, management of the power supply company.

Chen Fang received the B.Sc. and Ph.D. degrees from the
Department of Electrical Engineering, Tsinghua University,
Beijing, China, in 2006 and 2011, respectively.
He is currently an Electrical Engineer with the Electric
Power Research Institute, State Grid Shanghai Municipal
Electric Power Company, Shanghai, China. His current
research interests include renewable energy integration of
smart grid and power storage technology.

1949-3053 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.