Electric Power Systems Research 161 (2018) 17–25

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Zero-sequence voltage trajectory analysis for unbalanced distribution

networks on single-line-to-ground fault condition
Jun Meng a , Wen Wang b,∗ , Xin Tang b , Xianyong Xu c
a
Business School, Central South University, Changsha, China
b
School of Electrical and Information Engineering, Changsha University of Science & Technology, Changsha, China
c
Hunan Electric Power Corporation Research Institute, Changsha, China

a r t i c l e i n f o a b s t r a c t

Article history: Faulty phase recognition under single-line-to-ground (SLG) fault condition is critical for some arc suppres-
Received 14 December 2017 sion methods in medium voltage (MV) networks. Traditional methods usually assume that the distributed
Received in revised form 5 March 2018 parameters of the network are symmetrical, thus are unsuitable for the asymmetrical networks, espe-
Accepted 28 March 2018
cially when the ground-fault resistance is relatively high. In this paper, the trajectory of the zero-sequence
voltage with the change of ground-fault conductance is analyzed in detail. The magnitude and phase
Keywords:
angle variation rules of the voltage under different asymmetry ratios are presented. Then, the ranges
Distribution power system
of their unique variation rules on SLG fault for each phases are separately and strictly discussed both
Single-phase-to-ground fault
Faulty phase recognition
for the under-compensated and over-compensated grounding conditions. Furthermore, a faulty phase
High-resistance ground fault recognition method is proposed based on the variation rules of zero-sequence voltage on faulty condition.
Distribution network asymmetry Simulation results verify the concluded variation rules and validate the effectiveness of the proposed fault
Zero-sequence voltage phase recognition method, which ensures exact recognition of faulty phase in asymmetrical network and
high-resistance ground fault condition.
© 2018 The Author(s). Published by Elsevier B.V. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction Petersen coil paralleling with a reactor, which disconnects the

Power supply reliability is of great importance in distribution
networks [1]. Most of the reliability problems originate from single-
line-to-ground (SLG) faults. Arc suppression devices are designed
for distinguishing the arcs caused by SLG faults. Their performance
is critical to the safety of the power supply apparatus, and is
strongly associated with the inherent parameters of the distribu-
tion network [2–5].
Traditional arc suppression coil (ASC) compensates the capac-
itive ground current by the principle of parallel resonance,
encountering limited arc suppression performance as large har-
monic contents can also maintain arcs at the ground point [6,7].
Besides, the possible overvoltage caused by the series resonance
between ASC and inherent capacitance at the inception of ground
fault may further result in cross-country faults and insulation
failure [8–13]. Several methods have been presented to avoid over-
voltage and achieve dynamic adjustment of ASC inductance. Zeng
et al. [14] has presented a grounding method using a pre-adjusted
∗ Corresponding author.
E-mail addresses: 543977749@qq.com (J. Meng), ww csust@126.com,
wildwind6@126.com (W. Wang), 5xxy@163.com (X. Xu).
coil in normal condition and connects it to the neutral at ground
fault, to avoid overvoltage and extinguish fault arcs. This method
avoids series resonance to some extent, but lacks of flexibility to
the change of distributed parameters. Chen et al. [15] has pro-
posed a two-stage magnetically controlled reactor to enhance the
dynamic performance and mitigate the 3rd order harmonics caused
by nonlinearity of saturated iron core. Although they are more flex-
ible than the ASCs, these methods are still not able to compensate
the active and harmonic components in ground current, thus have
limited arc suppression performance.
Some active ways based on the technology of power electron-
ics are proposed to meet this requirement. Janssen et al. [16] and
Winter [17] has proposed a residual current compensation ground
fault neutralizer (RCC-GFN) which is able to compensate harmonic
ground current up to the 7th order. Tian et al. [18] has proposed
an ASC based on transformer with controlled load, which has the
ability of full ground-current compensation and rapid response. A
hybrid flexible grounding system is proposed by Peng et al. [19],
which combines the magnetic control reactor (MCR) providing
large-capacity reactive power and the active power compensator
(APC) realizing full ground current compensation. Besides, the volt-
age control method used in this device does not need capacitive
current detection, which decreases the response time and residual

https://doi.org/10.1016/j.epsr.2018.03.024
0378-7796/© 2018 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/
4.0/).
18 J. Meng et al. / Electric Power Systems Research 161 (2018) 17–25

current level. Wang et al. [20] has proposed a single-phase inverter

based full ground current compensator with time-domain dual-
loop zero-sequence voltage control method, which is able to realize
fast and reliable ground-fault arc suppression.
The performance of the active methods depends on the pre-
cise detection of the faulty phase on SLG fault. A misjudgment
may result in cross-country fault as the arc suppression device
may short-circuit the sound line to the ground, which will cause
the immediate operation of current protection to cut off the faulty
feeder [21].
The faulty phase recognition is straightforward if the distri-
bution network is symmetrical and the ground-fault resistance
is low enough. That is, in under-compensated grounding system,
e.g., ungrounded or resistance grounded system, the faulty phase
is the phase with the system voltage lagging its counterpart whose
line-to-ground voltage is most significant among the three line-
Fig. 1. Typical medium distribution network on SLG fault condition.
to-ground voltages on SLG fault condition. In over-compensated
grounding system, on the other hand, the faulty phase has the sys-
tem voltage heading its counterpart whose line-to-ground voltage are assumed to be equal, as R, for simplicity. RE and GE are Ground-
is most significant among three phases [22]. fault resistance and conductance, respectively. C is the sum of the
Nevertheless, the distribution network has the inherent feature
of asymmetry and various ground-fault resistance. It means in some line-to-ground capacitances. LN and RN are the neutral grounding
high ground-fault resistance conditions the criterion above goes inductance and resistance, respectively. The expression is unified
invalid. As the zero-sequence voltage on SLG fault follows the par- as it applies for all the grounding methods discussed previously. It
ticular rules that might indicate the faulty phase [23], it is useful can be simplified by using the rotating coefficient a = ej2␲/3 .
to analyze the voltage trajectories with the change of ground-fault CA +a2 CB +aCC GE
resistances. Xu et al. [22] has proposed a novel criterion by estab- j C + 
ωC
lishing and comparing three-phase voltage phasors, however, the U 0A = −E A   (2)
three-phase asymmetry ratio is considered to be scalar which is 3 
impractical enough as the ratio is a vector in most cases. Liu et al. j 1− 1
ω2 LN C  + ωC
1
 R
+ 1
RN
+GE
[21] has practically improved the zero-sequence voltage trajectory
by introducing three-phase asymmetry ratio vector and proposed a
The three-phase asymmetry ratio ra is used to describe the
faulty phase recognition method based on estimating and plotting
unbalance ratio of the distribution network, which is defined as
the zero-sequence voltage under SLG fault in each phase. However,
the boundaries of the sectors are ambiguous and complicated to CA + a2 CB + aCC
ra = = ra ∠˛
C
calculate, which makes the method difficult to apply in practice. (3)
In this paper, the magnitude and phase angle variation rules
of the zero-sequence voltage are analyzed in detail. The change
of the ground-fault resistance together with the distribution net- Theoretically, its magnitude ra has the maximum values of 5% in
work parameters are taken into consideration. The invalid region of distribution network with pure overhead line, 0.5% with pure cable
the traditional criterion above is discussed. A practical faulty phase line and 1.5% with mixed line. It should be noticed that the phase
recognition method based on the variation rules of zero-sequence angle of ra (defined as ˛) changes with the distribution line, which
voltage is proposed, which can be easily implemented in digital makes great difference when considering the zero-sequence volt-
processors. Simulation results are presented for verification of the age trajectory. Meanwhile, if its magnitude is specified, the trace of
proposed method. ra as ˛ changes from 0◦ to 360◦ is a circle with a circle center of the
origin point.
The detuning ratio  and damping ratio d mainly depend on the
2. Zero-sequence voltage trajectory analysis
distributed capacitive current and system grounding methods.

A typical medium voltage distribution network is shown in 1

 =1− (4)
Fig. 1. Either a neutral grounding inductor (Peterson Coil), a resis- ω2 LN C
tor or their combination can be chosen to ground the neutral point
of the distribution network. Some small-scale networks use non-
1
3 1

d= +
ωC
grounded method as the distributed capacitive current is relatively (5)
R RN
low. Therefore, the detuning ratio differs with the grounding meth-
ods. Assume that the SLG fault occurs in phase A, the neutral voltage
or the zero-sequence voltage can be described by the following To avoid series resonance between the Peterson Coil and dis-
expression. tributed capacitance, the detuning ratio should not be too small in
resonant grounded systems. Typically, a negative detuning ratio
jω(CA E A + CB E B + CC E C ) + GE E A of 10%–20% is set to over-compensate the capacitive current in
U 0A = − (1)
jωC + 3
R
+ GE + 1
jωLN
+ 1
RN case of overvoltage caused by series resonance when power system
operation mode changes from the maximum one to the minimum
U0 is the neutral voltage or zero-sequence voltage. U0X is one.
the Zero-sequence voltage when single-line-to-ground (SLG) fault The damping ratio has the range of 1.5%–5% in ungrounded sys-
occurs on phase X (X = A, B or C). EX , CX and RX are the line-to- tems. A damping resistor is necessary to connect in parallel with the
neutral voltage, line-to-ground capacitance and resistance of phase Peterson coil to damp the aforementioned series resonance, which
X, respectively. The line-to-ground resistances on the three phases consequently increases the system damping ratio. The damping
 J. Meng et al. / Electric Power Systems Research 161 (2018) 17–25 19

considering the specified ground resistance (RE (O1 )) which allo-

cates the zero-sequence voltage −U0A on the point of O1 , we can see
that the line-to-neutral voltage of phase C is no longer the largest
one. Indeed, as UBG is the largest one, phase C is inferred to be the
faulty phase by the traditional criteria, which is contradictory to
the assumed faulty phase. That is to say, a new criteria should be
established when the asymmetry of the network is not negligible
and the ground-fault resistance is relatively high.

3. Rules of the zero-sequence voltage variation

It is useful to analyze the variation rules of the magnitude and

phase angle of the zero-sequence voltage with the change of the
ground-fault resistance on SLG fault. As the faulty phase voltage
Fig. 2. Zero-sequence voltage trajectory for a specified distribution network when does not change with the resistance, the variation rules of ground-
SLG fault happens on phase A ( = +10%, d = 4%, ra = 3.5%, ˛ = 60◦ , RE (O1 ) = 1.5 k). fault ratio K is mainly discussed.

ratio also increases as the grounding method changes to resistive
grounding. Therefore, the zero-sequence voltage can be written to
jra cos ˛ − ra sin ˛ + GEs
U 0A = −E A = −E A K.
j + d + GEs
GEs is the standardized ground-fault conductance by the admit-
tance of the overall distributed capacitance, i.e., GE /(ωC ). The
3.1. Magnitude variation rules
Expression (11) shows the magnitude of ground-fault ratio. It
indicates that the sign of detuning ratio has no impact on M(K),
(6)
which means the variation rules of M(K) apply for both under-
compensated and over-compensated grounding methods.

(GEs − ra sin ˛)2 + ra2 cos2 ˛

zero-sequence voltage becomes the asymmetry voltage U00 by set- M(K) = 2
(11)
ting GEs to zero, which corresponds to the non-fault condition, as (GEs + d) + 2
follows:
If we fix the network coefficients, i.e., ra , , d and ˛, then M(K) is
jra cos ˛ − ra sin ˛ a function of the standardized ground-fault conductance GEs . Thus,
U 00 = −E A . (7)
j + d the inflection points of M(K) is determined by its derivative
It is notable that the trajectory of U00 as ˛ changes from 0 to dM(K)
= 0. (12)
360◦ is a circle as shown in Fig. 2, where all the vectors are nor- dGEs
malized by EA . K denotes the ground-fault ratio as shown in (6),
The roots of Eq. (12) can be expressed as
whose trajectory indicates the change of U0A with the ground-fault  2
resistance ranging from 0 to ∞, as EA is fixed. The trajectory of the ra2 − 2 − d2 ra2 − 2 − d2 ra (sin ˛(2 + d2 ) + dra )
GEs = ± + (13)
zero-sequence voltage i.e., −U0A , for a typical under-compensated 2(ra sin ˛ + d) 2ra sin ˛ + 2d ra sin ˛ + d
distribution network with SLG fault in Phase A is shown in Fig. 2.
Some features of the trajectory should be addressed. It starts from From the aforementioned discussion, the detuning ratio  is
the circle of the asymmetry voltage U00 when GEs is set to zero, as always larger than the magnitude of asymmetry ratio ra . Take notice
that GEs is always positive, therefore if the following inequality
jra cos ˛ − ra sin ˛
KS = . (8) sin ˛(2 + d2 ) + dra
j + d <0 (14)
ra sin ˛ + d
Its terminus KT is (1,0) as GEs approaches infinite, which corre-
sponds to the bolted SLG fault condition. The trajectory of K is part is met, there will be no inflection points in the trace of M(K). As
of a circle with the center of this condition makes the derivative in (12) positive, the magnitude
 r cos ˛ +  r sin ˛ + d  of ground-fault ratio increases as the ground-fault conductance
a a becomes larger. Inequality (14) is equal to the following expres-
KO = , , (9)
2 2 sions.
and the radius of ⎧ d dra
 d
⎪
⎨ − ra , − 2 + d2 , ra < 1
2
(ra cos ˛ − )2 + (ra sin ˛ + d) sin ˛ ∈ (15)
r= . (10)
2 ⎪
⎩ dra d
[−1, − ), ≥1
It can be seen that if the network is symmetrical, KS will be the 2 + d 2 ra
origin and the entire −U0A trajectory will be lagging EA . This is the Consider the condition that damping ratio is larger than asym-
base of the traditional criteria for faulty phase recognition. metry ratio, which is practical as the damping resistance RN exists,
From the traditional criteria, the line-to-neutral voltage of the then the range of ˛ to make M(K) monotonically increasing is
faulty phase is 120◦ leading the line-to-neutral voltage of the phase  
dra dra
with a maximum line-to-ground voltage in an over-compensated ˛ ∈  + arcsin , 2 − arcsin 2 . (16)
network; whereas, in an under-compensated network, the line-to-  2 + d2  + d2
neutral voltage of the faulty phase is 120◦ lagging that of the phase The zero-sequence voltage traces with different ˛ are shown in
with a maximum line-to-ground voltage. Fig. 3. From the specified network parameters, the monotonically
Fig. 2 shows the area of −U0A which validates the traditional cri- increasing range of M(K) is (186.93◦ , 353.07◦ ). The angle corre-
teria. In this area, the line-to-neutral voltage of phase C is the most sponding to curve I is within the range and it is verified from
significant one among three phases, which leads to a correct infer- the curve that the magnitude of zero-sequence voltage increases
ence that phase A is the faulty phase. However, if the ground-fault straightly from the start to the terminus. Nevertheless, curve II
resistance is high enough, the criteria may be invalid. For instance, shows a different way in magnitude variation. Its corresponding
20 J. Meng et al. / Electric Power Systems Research 161 (2018) 17–25

Fig. 5. Integrated zero-sequence voltage magnitude and phase angle variation rules
Fig. 3. Zero-sequence voltage magnitude variation region with different asymmetry when SLG fault happens on phase A ( = +10%, d = 4%, ra = 3.5%).
ratio phases when SLG fault happens on phase A ( = +10%, d = 4%, ra = 3.5%; Curve I:
˛ = 270◦ ; Curve II: ˛ = 45◦ ).

Fig. 4. Zero-sequence voltage phase variation region with different asymmetry ratio
phases when SLG fault happens on phase A ( = +10%, d = 4%, ra = 3.5%; Curve I:
˛ = 287.56◦ , Curve II: ˛ = 90◦ , Curve III: ˛ = 72.44◦ ).
zero-sequence voltage tends to decrease firstly, followed by an
increase. This again verifies the aforementioned rules as ˛ is out
of the monotonically increasing range.
rises after GEs reaches the root specified by (19) as GEs increases
from 0 to infinite. The corresponding range is shown by the blue
arc in Fig. 4, which is ˛ ∈ (−72.4◦ , 72.4◦ ) for the specified net-
work parameters. Zero-sequence voltage curves as ˛ is at the
two boundaries (I and III) are particularly displayed, from which
it can be seen that the intersection points of the curves and U00
circle are just the tangent points from the origin to the curves.
Thus,the boundariescorrespond to the phase shift points.
2) ˛ ∈ arccos 2ra 2 , 2 . This condition ensures that the roots of
 +d
(18) are two positive reals. Therefore, the curve of P(K) has firstly
a rise, then a fall and finally a rise as GEs increases. Particularly,
the roots of (18) are equal when ˛ is /2, thus the fall and rise
 simultaneously
happen  origin point (see curve II in Fig. 4).
at the
3) ˛ ∈ ␲ 2
, 2 − arccos 2ra 2 . On this condition the roots of (18)
 +d
are either negative or imaginary, which means P(K) monotoni-
cally increases as GEs increases.
3.3. Integrated variation rules

3.2. Phase angle variation rules For practical network parameters, the following inequalities
stand.
From Eq. (6), the phase angle of ground-fault ratio can be
 dra ra
expressed as (17). <  + arcsin 2 < 2 − arccos 2 (20)
⎧ 
2  + d2  + d2
⎨ arctan ra cos ˛
− arctan , as GEs > ra sin ␣
GEs − ra sin ˛ GEs + d
P(K) = (17) ra dra ra
⎩  + arctan ra cos ˛
− arctan

, as GEs < ra sin ˛ −arccos < −arcsin 2 < arccos 2 (21)
GEs − ra sin ˛ GEs + d  2 + d2  + d2  + d2
Similarly, we can fix the network coefficients and obtain the
derivative of P(K) over ground-fault conductance GEs . The inflection Therefore, the zero-sequence voltage magnitude and phase
point of P(K) is determined by angle variation rules can be integrated in one plane. The ranges
indicating the unique magnitude and phase variation rules of zero-
dP(K) sequence voltage can then be concluded as shown in Table 1.
=0 (18)
dGEs Fig. 5 aims to illustrate the ranges in Table 1. As each range cor-
Assume that the phase angle of asymmetry ratio ˛ is /2 or 3/2, responds to unique zero-sequence voltage magnitude and phase
from (9) and (10) it can be seen that the origin point is on the trace of angle variation rules, the voltage trajectory is predictable if the start
zero-sequence voltage, which means an 180◦ phase change occurs point is specified. Actually, the point is determined by the phase
around the point. This phase shift can also be observed from the angle and magnitude of the asymmetry ratio ra . The rules shown in
roots of (18), which are two equal ones with positive real value. Table 1 specifies the zero-sequence voltage feature of SLG fault on
If ˛ is neither /2 nor 3/2, the roots of Eq. (18) can be expressed phase A. The faulty feature on phase B and C can also be concluded,
as which will be discussed in the next section.
 2
 tan ˛ + d  tan ˛ + d 2 + d2 − ra / cos ˛
GEs = ± + (19)
/(ra cos ˛) − 1 /(ra cos ˛) − 1 /(ra cos ˛) − 1
4. Faulty phase recognition method
The variation rules of P(K) with ˛ can be concluded as follows.
Typical zero-sequence voltage traces are shown in Fig. 4 for explicit As the faulty phase corresponds to the specific zero-sequence
explanation. voltage trace and the traces for the three phases never cross each
  other, an effective way to recognize the faulty phase is to com-
1) ˛ ∈ −arccos 2ra 2 , arccos 2ra 2 . On this condition, Eq. (18) pare the zero-sequence voltage on SLG fault with that on normal
 +d  +d
has just one positive real root, which means P(K) falls first, then condition.
 J. Meng et al. / Electric Power Systems Research 161 (2018) 17–25 21

Table 1
Zero-sequence voltage magnitude and phase angle variation rules on under-compensated grounding condition.

Range number Asymmetry ratio angle ˛ (˛ + 2/3 Zero-sequence voltage Zero-sequence voltage phase
for phase B; ˛ − 2/3 for phase C) magnitude variation rules angle variation rules
 
dra ra
I −arcsin , arccos Fall → rise Fall → rise
  2 + d2   2 + d2
ra 
II arccos 2 , Fall → rise Rise → fall → rise
  + d 2 2

 dra
III ,  + arcsin 2 Fall → rise Rise
2  + d2 
dra ra
IV  + arcsin 2 , 2␲ − arccos 2 Rise Rise
  +d 2  + d 2

ra dra
V −arccos 2 , −arcsin 2 Rise Fall → rise
 + d2  + d2

Fig. 6. Zero-sequence voltage curves with SLG fault on different phases. ( = +10%,
d = 4%, ra = 3.5%).

4.1. Zero-sequence voltages when SLG fault happens on phase B

or C

Similarly to the previous discussion, the zero-sequence voltages

when SLG fault happens on phase B or C can be expressed as follows.
To simplify the discussion, they are transferred to the same plane
with that when SLG fault happens on phase A.

jar a + GEs jr a + a2 GEs

U 0B = −E B = −E A (22)
j + d + GEs j + d + GEs

ja2 r a + GEs jr a + aGEs

U 0C = −E C = −E A (23)
j + d + GEs j + d + GEs
It can be seen that the start point for the three voltage curves
are unique if the network parameters are specified. Nevertheless,
the terminal points are specified by the negative vector of line-to-
neutral voltages as shown in Fig. 6.
Comparing (22) and (23) with (6), it can be concluded that the
magnitude and phase angle of U0B and U0C follow the same varia-
tion rules with U0A . The only difference is the relationship between
the asymmetry ratio angle and the variation range. Unlike the rela-
tionship listed in Table 1, we have to substitute ˛ for ˛ + 2/3 and
˛ − 2/3 for SLG fault on phase B and C, respectively. For instance, if Fig. 7. Flowchart of the faulty phase recognition method on under-compensated
we consider the zero-sequence voltage on phase-B SLG fault when condition.
the asymmetry ratio angle is zero, a simple way to judge the varia-
tion rules is to use Table 1 with an asymmetry ratio angle of 2/3,
which results in the condition in range III. The flowchart of the faulty phase recognition method based on
the analysis above is shown in Fig. 7. Firstly, the network parame-
4.2. Faulty phase recognition method ters should be measured, i.e., the asymmetry ratio, damping ratio
and detuning ratio. We can obtain the three zero-sequence volt-
Comparing the zero-sequence voltages in (6), (22) and (23), it is age variation rules R(A), R(B) and R(C) from Table 1 by assuming
obvious that the three voltage traces can only cross each other when SLG fault occurs on phase A, B and C, respectively. Then, we mea-
GEs is zero, i.e., no ground fault happens in the network. Therefore, sure the zero-sequence voltage twice with an interval of around
the zero-sequence voltage variation rules can be used for faulty 1 s. If the deviation ratio of zero-sequence voltage magnitude 20%,
phase recognition. an SLG fault is claimed to happen. The deviation ratio limit is set
22 J. Meng et al. / Electric Power Systems Research 161 (2018) 17–25
Fig. 8. Zero-sequence voltage phase variation region with different asymmetry ratio
phases when SLG fault happens on phase A ( = −10%, d = 4%, ra = 3.5%; Curve I:
˛ = 107.56◦ , Curve II: ˛ = 90◦ , Curve III: ˛ = 252.44◦ ).
to deal with the SLG fault with high ground-fault resistance which
is the situation this paper mainly focuses on. As for the low resis-
tance situation, the aforementioned traditional method is precise
enough to find the faulty phase. Then, we calculate the change of the
magnitude and phase angle of the zero-sequence voltage. Finally,
the rule that matches the real variation of the voltage indicates the
corresponding phase to be on fault.
4.3. Discussion on over-compensated grounding method
We can do the same analysis for the over-compensated ground-
ing method, i.e., the detuning ratio is smaller than zero, and then
some similar conclusion can be obtained. As for the zero-sequence
voltage magnitude variation rules, the expressions of (6), (22) and
(23) indicate that the magnitude is not affected by the sign of detun-
ing ratio, i.e., the range specified by (16) also corresponds to the
monotonically increasing magnitude and the rest of the plane cor-
responds to the fall-and-rise rules.
The phase angle variation rules of the zero-sequence voltage
under over-compensated grounding condition are a little compli-
cated as shown in Table 2. Fig. 8 aims to illustrate the variation
ranges listed in Table 2. It can be seen that there are also three
boundaries of the ranges, which are also either the tangent points or
the cross-origin point. However, the variation rules are just inverse
to those of the under-compensated condition.
5. Simulation verification
To verify the theoretical analysis and faulty phase recognition
method, a 10.5 kV distribution network with specified parameters
is simulated in MATLAB. Its topology is the same as Fig. 1. The
capacitive current of the network is fixed to 100 A.
The simulation results are classified in four groups as shown in
Tables 3–6. All the phase angles are relative values to the line-to-
neutral voltage of phase A, i.e., EA . The first table aims to discuss the
five ranges corresponding to different zero-sequence voltage vari-
ation rules when phase-A SLG fault occurs on under-compensated
grounding condition. The second is to extend the conclusion to
Phase-B SLG fault condition. The third group addresses the over-
compensated condition. Finally, the conditions when traditional
faulty phase recognition criterion is invalid are illustrated and com-
pared with the proposed faulty phase recognition method.
5.1. Zero-sequence voltage variation rules verification
The magnitude and phase angle variation of zero-sequence volt-
age on phase-A SLG fault and under-compensated condition is
shown in Table 1. Five values of asymmetry voltage angle are cho-
sen to illustrate the five variation trends analyzed above, each
Fig. 9. Typical asymmetry ratio angles and the allocation of their corresponding U00 .
Fig. 10. Detailed zero-sequence voltage curves on phase-B SLG fault condition
( = 10%, ra = d = 1.5%, ˛ = 240◦ ).
of which is within one of the five ranges, as shown in Fig. 9.
Take notice that the values of grounding conductance is limited
to [10−5−1 ][10−5 ,10−1 ], this on the one hand is for the practical
reason that the ground-fault resistance is greater than 0.1 M, on
the other is because the zero-sequence voltage is very close to the
terminal point as the ground-fault conductance reaches its upper
limit. It should also be noticed that the magnitude and phase angle
are both increasing if the ground-fault conductance is large enough.
This not only verifies the theoretical analysis, but also indicates
that high ground-fault resistances should be taken special con-
siderations as the variation rules of corresponding zero-sequence
voltages are totally different to those of their lower counterparts.
The second group shows the zero-sequence voltage on phase-B
SLG fault and the results are listed in Table 4. A smaller asymme-
try ratio is used to simulate mixed distribution lines. Three typical
asymmetry ratio angle are chosen to verify the analysis above. The
three angles from the top to the bottom correspond to range no.
III, IV and I in Table 1, respectively. The results match the analyzed
variation rules well except for the condition when ˛ equals 240◦ .
This is because the range of ˛ for M(U0B ) to be falling is too small
and the magnitude difference is so tiny that it is hardly possible
for the simulation software to demonstrate due to the calculation
precision (as shown in Fig. 10). Therefore, it is practical to neglect
this range and consider the magnitude to be rising at this particular
asymmetry ratio angle.
As the inference for zero-sequence voltage variation rules on
phase-B SLG fault condition is verified, it is reasonable to say that
the concluded rules on phase-C SLG fault condition is also correct
due to the interchangeability between three phases. Thus, the SLG
fault simulation on phase-C is unnecessary to be conducted.
The rules under over-compensated grounding condition are also
evaluated. In this group, three typical angles are also chosen to rep-
resent the three ranges from the top to the bottom in Table 2, i.e.,
˛1 , ˛2 and ˛3 correspond to range I, II and III, respectively. The
phase angle variation rules in Table 2 are well verified from the data
 J. Meng et al. / Electric Power Systems Research 161 (2018) 17–25 23

Table 2
Zero-sequence phase angle variation rules on over-compensated grounding condition.

Range number Asymmetry ratio angle ␣ Zero-sequence voltage phase angle variation rules
 
 ra
I , arccos 2 Fall → rise → fall
2  + d2 
ra ra
II arccos 2 , 2 − arccos 2 Rise → fall
  +d 2
  +d 2

ra 
III 2 − arccos 2 , Fall
 + d2 2

Table 3
Zero-sequence voltage on phase-A SLG fault ( = +10%, d = 4%, ra = 3.5%).

˛ Corresponding range number GE /S M(U0A )/kV Variation of M(U0A ) P(U0A )/◦ Variation of P(U0A )

˛1 = 45◦ I 10−5 1.94 – −115.9 –

10−4 1.72 ↓ −120. 0 ↓
10−3 1.92 ↑ −189.9 ↓
10−2 5.41 ↑ −186.3 ↑
10−1 6.0 ↑ −180. 7 ↑

˛2 = 85◦ II 10−5 1.94 – −75.2 –

10−4 1.60 ↓ −74.0 ↑
10−3 1.12 ↓ −215.1 ↓
10−2 5.30 ↑ −188.3 ↑
10−1 5.99 ↑ −180.9 ↑

˛3 = 105◦ III 10−5 1.94 – −54.9 –

10−4 1.61 ↓ −49.9 ↑
10−3 1.18 ↓ 119.0 ↑
10−2 5.30 ↑ 170.5 ↑
10−1 5.99 ↑ 178.9 ↑

˛4 = 225◦ IV 10−5 1.99 – 65.52 –

10−4 2.17 ↑ 73.98 ↑
10−3 3.77 ↑ 118.3 ↑
10−2 5.85 ↑ 168.8 ↑
10−1 6.05 ↑ 178.7 ↑

˛5 = 300◦ V 10−5 2.00 – 139.3 –

10−4 2.23 ↑ 138.4 ↓
10−3 3.98 ↑ 145.3 ↑
10−2 5.91 ↑ 172.62 ↑
10−1 6.05 ↑ 179.3 ↑

Table 4
Zero-sequence voltage on phase-B SLG fault ( = +10%, d = 1.5%, ra = 1.5%).

˛ GE /s M(U0B )/kV Variation of M(U0B ) P(U0B )/◦ Variation of P(U0B )

0◦ 10−5 0.87 – 189.2 –

10−4 0.60 ↓ 207.9 ↑
10−3 2.32 ↑ 358.4 ↑
10−2 5.71 ↑ 50.2 ↑
10−1 6.03 ↑ 59.0 ↑

120◦ 10−5 0.93 – −50.8 –

10−4 1.2 ↑ −40.4 ↑
10−3 3.57 ↑ 1.1 ↑
10−2 5.97 ↑ 50.2 ↑
10−1 6.06 ↑ 59.0 ↑

240◦ 10−5 0.90 – 65.7 –

10−4 0.97 ↑ 49.1 ↓
10−3 3.03 ↑ 20.9 ↓
10−2 5.84 ↑ 52.2 ↑
10−1 6.05 ↑ 59.2 ↑

in Table 5. As to the magnitude variation rules, the monotonically
increasing range as analyzed above is (186.9, 353.1), in which the
second and third asymmetry ratio angles are included. The results
perfectly match the presented rules.
5.2. Discussion on faulty phase recognition method
Four conditions indicating that the traditional faulty phase
recognition criterion is invalid are shown in Table 6. Two of them
are on under-compensated grounding condition and the others are
on over-compensated one. The ground-fault resistances are rela-
tively high but practical enough. Take notice that phase A is set
to be in SLG fault on all the conditions. UAX is the line-to-ground
voltages of phase X.
The zero-sequence variation rules can be used to recognize the
faulty phase. For condition I, the asymmetry ratio angle corre-
sponds to range I in Table 1 while assuming phase-A SLG fault. The
range changes to III and V while considering SLG faults on phase
B and C, respectively. As the zero-sequence voltage magnitude on
faulty condition decreases from the normal, i.e., M(U0A ) < M(U00 ),
24 J. Meng et al. / Electric Power Systems Research 161 (2018) 17–25

Table 5
Zero-sequence voltage on phase-A SLG fault under over-compensated grounding condition ( = −10%, d = 4%, ra = 3.5%).

˛ GE /s M(U0A )/kV Variation of M(U0A ) P(U0A )/◦ Variation of P(U0A )

◦ −5
˛1 = 100 10 1.93 – 75.78 –
10−4 1.60 ↓ 74.7 ↓
10−3 1.12 ↓ 214.6 ↑
10−2 5.30 ↑ 188.3 ↓
10−1 5.99 ↑ 180.9 ↓

˛2 = 220◦ 10−5 1.99 – −163.6 –

10−4 2.15 ↑ −160.0 ↑
10−3 3.70 ↑ −153.7 ↑
10−2 5.82 ↑ −173.7 ↓
10−1 6.04 ↑ −179.3 ↓

˛3 = 340◦ 10−5 1.98 – −45.4 –

10−4 2.08 ↑ −55.6 ↓
10−3 3.45 ↑ −111.4 ↓
10−2 5.75 ↑ −168.1 ↓
10−1 6.04 ↑ −178.7 ↓

Table 6
Conditions when traditional criterion does not stand (d = 4%, ra = 3.5%, Phase A is set to be the faulty phase).

Cond. I Cond. II Cond. III Cond. IV

◦ ◦ ◦
˛ 60 105 100 150◦
◦ ◦
 +10% +10% −10% −10%
GE /S 7 × 10−4 5 × 10−4 5 × 10−4 6 × 10−4
RE /k 1.4 2 2 1.7
M(UAG )/kV 4.58 6.49 6.01 4.41
M(UBG )/kV 6.99 5.81 5.83 6.89
M(UCG )/kV 6.86 5.92 6.37 7.18
M(U0A )/kV 1.49 0.43 0.32 1.66
P(U0A )/◦ −176.4 8.5 101.5 173.5
M(U00 )/kV 1.97 1.94 1.97 1.97
P(U00 )/◦ −115.6 −55.4 76.0 126.0
Faulty phase by traditional criterion Phase C Phase B Phase B Phase B

SLG fault on phase-C must not happen. In the same way, as the
zero-sequence voltage angle falls from the normal condition to the
faulty one, SLG fault on phase-B must not happen.
For condition II, the specified asymmetry ratio angle corre-
sponds to range III, IV and V while assuming SLG faults on phase
A, B and C, respectively. It is easy to exclude the possibility of SLG
faults on phase B and C as the magnitude of zero-sequence voltage
falls from the normal condition to the faulty condition.
The same analysis can be conducted for the over-compensated
grounding conditions. As to the specified network parameters,
the ranges for different zero-sequence voltage angle variation in
Table 2 are (90◦ , 107.6◦ ], (107.6◦ , 252.4◦ ] and (252.4◦ , 450◦ ] from
the top to the bottom. For condition III, as the zero-sequence
voltage angle increases from the normal to faulty condition,
the SLG fault on phase-C must not happen. As the asymme-
try ratio angle forces the zero-sequence voltage magnitude in
the monotonically increasing region for SLG fault on phase-B,
i.e., (186.9◦ < ˛ + 120◦ = 196.0◦ < 353.1◦ ), the falling magnitude from
normal to faulty condition makes this type of fault impossible to
happen.
For condition IV, the asymmetry ratio angle P(U00 ) forces both
the zero-sequence voltage angle on phase B and phase C into the
third range, i.e., the angle has a monotonically falling nature. How-
ever, as the data of P(U0A ) shows, the zero-sequence voltage angle
just increases from the normal condition. Therefore, the SLG faults
on phase-B and phase-C must not happen.
and phase angle variation rules of zero-sequence voltage with the
change of ground-fault conductance on SLG fault condition for both
under-compensated and over-compensated grounding systems are
presented in detail. The ranges for unique magnitude and phase
angle variation rules are strictly addressed. A faulty phase recog-
nition method based on the rules is proposed and the flowchart
of the method is provided and described in detail. Comparisons
are made between the proposed method with the traditional one,
which indicates the proposed one is immune to the asymmetry
and high ground-fault resistance. As the unique ranges of zero-
sequence voltage under SLG fault are clearly presented, the method
is easily to be implemented, especially for digital processors. How-
ever, the SLG fault with low ground-fault resistance should be dealt
with carefully, as a result, the traditional method is recommended
if the resistance is relatively low.
Acknowledgments
This research was sponsored in part by the National Natu-
ral Science Foundation of China (Grant Nos. 51407014, 51577014
and 51707013), in part by the Hunan Provincial Natural Science
Foundation of China (Grant No. 2018JJ2427), and in part by the
Key Laboratory of Renewable Energy Electric-Technology of Hunan
Province (Grant No. 2017ZNDL004). The authors would like to
acknowledge Ziqiang Sheng, Mingwen Ma and Lingjie Yan for their
help on finding the variation rules of zero-sequence voltage.

6. Conclusion References

[1] W. Wang, L. Yan, X. Zeng, B. Fan, J.M. Guerrero, Principle and design of a
Traditional criterion for faulty phase recognition on SLG fault is single-phase inverter based grounding system for neutral-to-ground voltage
invalid as the asymmetry of the distribution network exists, espe- compensation in distribution networks, IEEE Trans. Ind. Electron. 64 (2)
cially when the ground-fault resistance is high. The magnitude (2017) 1204–1213.
 J. Meng et al. / Electric Power Systems Research 161 (2018) 17–25
[2] Y. Li, X. Meng, X. Song, Application of signal processing and analysis in
detecting single line-to-ground (SLG) fault location in high-impedance
grounded distribution network, IET Gener. Transm. Distrib. 10 (2) (2016)
382–389.
[3] M. Brenna, E. De Berardinis, L. Delli Carpini, P. Paulon, P. Petroni, G. Sapienza,
G. Scrosati, D. Zaninelli, Petersen coil regulators analysis using a real-time
digital simulator, IEEE Trans. Power Deliv. 26 (July (3)) (2011) 1479–1488.
[4] Q. Wang, M. Cheng, Y. Jiang, W. Zuo, G. Buja, A simple active and reactive
power control for applications of single-phase electric springs, IEEE Trans.
Ind. Electron. 65 (August (8)) (2018) 6291–6300.
[5] X. Xia, Y. Zhou, C. Fu, Z. Zhou, Y. He, Research on high voltage DC transmission
system optimal control based on MMC, Int. J. Electr Power Energy Syst. 82
(November (11)) (2016) 207–212.
[6] A.S. Al-Abdulwahab, K.M. Winter, N. Winter, Decreasing the harmonic
content of the fault current during single-phase to grounding faults in
compensated network, in: Proc. IEEE Buch. Power Tech. Conf., June 28–July 2,
2009, pp. 2497–2501.
[7] Xiangning Lin, Jingguang Huang, Shuohao Ke, Faulty feeder detection and
fault self-extinguishing by adaptive Petersen coil control, IEEE Trans. Power
Deliv. 26 (April (2)) (2011) 1290–1291.
[8] A. Kalyuzhny, Analysis of temporary overvoltages during open-phase faults in
distribution networks with resonant grounding, IEEE Trans. Power Deliv. 30
(February (1)) (2015) 420–427.
[9] Q. Wang, M. Cheng, Z. Chen, Steady-state analysis of electric springs with a
novel ␦ control, IEEE Trans. Power Electron. 30 (December (12)) (2015)
7159–7169.
[10] B. Gustavsen, J.A. Walseth, A case of abnormal overvoltages in a Petersen
grounded 132-kV system caused by broken conductor, IEEE Trans. Power
Deliv. 18 (January (1)) (2003) 195–200.
[11] F.M. Gatta, A. Geri, S. Lauria, M. Maccioni, Analytical prediction of abnormal
temporary overvoltages due to grounding faults in MV networks, Elect. Power
Syst. Res. 77 (August (10)) (2007) 1305–1313.
[12] L. Kai, L. Mengshu, L. Zhengfeng, D. Zhenghua, T. Shaoju, Analytical
closed-form expressions of DC current ripple for three-level neutral point
clamped inverters with space-vector pulse-width modulation, IET Power
Electron. 9 (April (5)) (2016) 930–937.
25
[13] Alberto Cerretti, F.M. Gatta, Alberto Geri, Stefano Lauria, M. Maccioni,
Giovanni Valtorta, Grounding fault temporary overvoltages in MV networks:
evaluation and experimental tests, IEEE Trans. Power Deliv. 27 (July (3))
(2012) 1592–1600.
[14] X. Zeng, Y. Xu, Y. Wang, Some novel techniques for insulation parameters
measurement and Petersen-coil control in distribution systems, IEEE Trans.
Ind. Electron. 57 (April (4)) (2010) 1445–1451.
[15] Xuxuan Chen, Baichao Chen, Cuihua Tian, Jiaxin Yuan, Yaozhong Liu,
Modeling and harmonic optimization of a two-stage saturable magnetically
controlled reactor for an Arc suppression coil, IEEE Trans. Ind. Electron. 59
(July (7)) (2012) 2824–2831.
[16] M. Janssen, S. Kraemer, R. Schmidt, K. Winter, Residual current compensation
(RCC) for resonant grounded transmission systems using high performance
voltage source inverter, IEEE PES Transm. Distrib. Conf. Expos. 2 (2003)
574–578.
[17] K.M. Winter, The RCC ground fault neutralizer — a novel scheme for fast
earth-fault protection, in: International Conference and Exhibition on
Electricity Distribution, Turin, Italy, 2005, pp. 1–4.
[18] J. Tian, Q. Chen, L. Cheng, Y. Zhang, Arc-suppression coil based on transformer
with controlled load, IET Elect Power Appl. 5 (September (8)) (2011) 644–653.
[19] W. Peng, C. Baichao, T. Cuihua, B. Sun, M. Zhou, J. Yuan, A novel neutral
electromagnetic hybrid flexible grounding method in distribution networks,
IEEE Trans. Power Deliv. 32 (June (3)) (2017) 1350–1358.
[20] W. Wang, X. Zeng, L. Yan, X. Xu, J.M. Guerrero, Principle and control design of
active ground-fault arc suppression device for full compensation of ground
current, IEEE Trans. Ind. Electron. 64 (6) (2017) 4561–4570.
[21] R. Burgess, A. Ahfock, Minimising the risk of cross-country faults in systems
using arc suppression coils, IET Gener. Transm. Distrib. 5 (July (7)) (2011)
703–711.
[22] B. Xu, J. Zhang, X. Cai, Grounding phase determination in non-effective
grounding systems in case of unsymmetrical voltage, Trans. China
Electrotechn. Soc. 26 (December (12)) (2011) 175–182.
[23] B. Liu, C. Wang, X. Li, Analysis and application of zero-sequence voltage of
single-phase ground fault asymmetrical system, Proc. CSEE 34 (October (28))
(2014) 4959–4967.