IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO.

4, OCTOBER 2006 1847

Improved Phase Selector for Unbalanced

Faults During Power Swings Using
Morphological Technique
Li Zou, Qingchun Zhao, Xiangning Lin, and Pei Liu

Abstract—In order to prevent distance protection from tripping quantities, such as voltage and current, may be particularly faint.
during power swing conditions, a power swing blocking device More recently, the technique using symmetrical components
is often utilized. On the other hand, it has been an increasing and the fuzzy set approach has been employed in fault detection
requirement to achieve rapid clearance of internal faults during
power swings. Accurate phase selection is a prerequisite to selec- during power swings [6]. This method, however, needs more
tive clearance of faults. An improved unbalanced fault phase se- research and exhaustive investigation in order to yield better
lector during a power swing based on series multiresolution mor- results.
phological gradient (SMMG) transform is proposed in this paper. The faulty phase(s)-related ground distance relay and phase
As a feature extractor from raw signals, SMMG is employed to distance relay are able to operate correctly for single-phase-to-
extract superimposed (fault component) current under the power
swing condition in this paper. First, the operating characteristic ground faults and phase-to- phase (to ground) faults, respec-
of the sequence component fault phase selector during power tively, even under power-swing conditions [7]. Therefore, it is
swings is discussed. In order to overcome the disadvantage of the desirable and necessary to identify the fault phases correctly and
above selector, an improved fault phase selector for unbalanced rapidly during the power swing so that the fault can be mea-
faults during power swings is then proposed by using SMMG to
extract superimposed components of modular current. The effi-
sured accurately and cleared selectively. One of the most widely
ciency and feasibility of the proposed schemes are proven using used fault phase selection schemes in the existing digital relay
a Power Systems Computer-Aided Engineering/Electromagnetic of transmission line is the sequence component fault phase se-
Transients including DC-based simulation on a sample power lector (SCPS) [7], [8], which is based on the relationship be-
system. tween the zero-sequence and negative-sequence currents mea-
Index Terms—Distance protection, phase selector, power swing, sured by the relays. Generally, this phase selector can provide
series multiresolution morphological gradient (SMMG). satisfactory performance. According to dynamic simulation re-
sults and theoretical analysis, however, the scheme maloperates
I. INTRODUCTION in some cases during power swings [9].
An improved unbalanced fault phase selector during the
N SMALL electrical systems or over long transmission
O distances, severe disturbances can cause rapid power
swings [1]. In order to prevent the distance protection from
power swing based on the series multiresolution morphological
gradient (SMMG) transform is put forward in this paper. Other
applications of mathematical morphology (MM) in power sys-
maltripping, a power swing blocking device should be ap- tems, such as identification of transformer magnetizing inrush
plied. It has been demonstrated that conventional power swing current [10] and ultra-high-speed directional protection of
blocking schemes are inadequate for very fast power swings transmission lines [11] have also been reported. The potential
[2]. Furthermore, the blocked protective relays will fail to trip of the proposed scheme is demonstrated using a Power Sys-
any internal fault which occurs during power swing blocking tems Computer-Aided Engineering/Electromagnetic Transients
state. including DC (PSCAD/EMTDC)-based simulation of a sample
During the last decade, several schemes for the discrimina- power system. The proposed scheme can be implemented on
tion of power swings and faults have been proposed. In [3] and standard relaying hardware.
[4], the rate of resistance change is used to distinguish between
power swings and faults. A power swing blocking method with
fault detection for low- and high-impedance faults by identi- II. SINGULARITY DETECTION USING SMMG
fying the voltage phase angle is proposed in [5] and [1], re-
spectively. However, when a high-resistance fault occurs during The underlying basis of the morphological filtering technique
the swing with large amplitude, these methods may suffer in- is to process signals by a function, known generally as the struc-
sufficient sensitivity as the change of the measurable electrical turing element (SE). In MM, dilation and erosion are two basic
operations. Let and denote a one-dimensional (1-D)
Manuscript received June 20, 2005; revised November 12, 2005. This input signal and an SE (also a signal), respectively, where
work was supported by the National Natural Science Foundation of China and denote the domain set on which signals and
(50177011). Paper no. TPWRD-00357-2005. are defined. In practice, represents or in the continuous
The authors are with the College of Electric and Electronic Engineering,
Huazhong University of Science and Technology, Wuhan 430074, China. or discrete cases, respectively [11]. Discrete cases are consid-
Digital Object Identifier 10.1109/TPWRD.2006.874608 ered in this paper; in other words, the input signal and SE are
0885-8977/$20.00 © 2006 IEEE
1848 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 4, OCTOBER 2006

both discrete sequences. The grayscale dilation and ero-

sion of by are defined by [12]
Fig. 1. Block diagram of SMMG.
(1)
(2) MMG transform cascaded in series with , the same or different
SEs are expected to magnify inconspicuous transient features
where and . It can be concluded from (1) and detect faint signal changes. The SMMG procedure is shown
and (2) that a dilation or an erosion of with a length of by in Fig. 1, where MF1, MF2, , MFn are MMG filters. Let a
a flat line SE (SE with all its values equal to zero) with a signal denote a flat line SE, is the length of , and
length of requires comparisons. arithmetic operator is defined as MMG by at level .
Apparently, the number of comparisons required is related to Correspondingly, let a set of signals ,
the length of the SE. As seen, the computational complexity is denote a set of flat-line SEs; is the length of , and
very small when using a short SE. As for a long SE, the number arithmetic operator is defined as the -order se-
of comparisons required can be remarkably reduced by some ries MMG by at level , respectively. In particular, we have
fast algorithms of grayscale morphological filters [13], by which if SEs have the same length L. As a special case,
the computational complexity is independent of the length of MMG can be regarded as first-order SMMG transform.
SE. Note that all of the operations in the proposed schemes in The filter length can be calculated from the following
this paper are based on these two operations, the computation formula:
burdens, therefore, are much smaller than conventional digital
filter algorithms. This attribute is highly desirable for real-time (8)
applications.
The basic morphological gradient (MG), also known as the
Beucher gradient, is defined as the arithmetic difference be- Based on (8), the SMMG filter length can be calculated by
tween the dilation and erosion of signal by SE
(9)
(3)

Frequently, the MG is used for edge detection in image and In signal processing of power systems, singularities and ir-
signal processing. With the purpose of extracting the ascending regular signals often carry the most important information. The
and descending edges of the transient waves, a multiresolution singularities of signals indicate disturbances or faults occurring
morphological gradient (MMG) transform has been designed in in power systems, which are particularly meaningful for protec-
[11]. The MMG is designed to have a pair of scalable flat line tion relay. By feeding a sampled signal to an SMMG filter, the
SEs, with different origins, defined as SMMG coefficients can be obtained. SMMG detects irregular
signals by means of a local modular maxima of SMMG coeffi-
cient. For an SMMG transform , if is
(4) satisfied for arbitrary when belongs to either the right or the
left neighborhood of is the point of SMMG local mod-
ular maxima.
where is the SE used for extracting the ascending edges
and is for the descending edges of the transient waves
, where indicates the level of the MMG to be processed, III. SEQUENCE COMPONENT FAULT PHASE SELECTOR
is the primary length of at level 1. The underlined samples Three-phase currents can be represented by positive-
in and show their origins. Based on the definition of sequence, negative-sequence, and zero-sequence currents on
MG in (3) and scalable flat SEs in (4), the MMG with level the basis of symmetrical component law. In SCPS, the fault
is defined as phases can be identified using the relationship between the mea-
sured zero-sequence and negative-sequence currents phasors
(5) for grounded faults and using the apparent impedances mea-
sured by the distance relay for ungrounded faults, respectively.
(6)
The relationship between the zero-sequence and negative-se-
(7) quence currents phasors measured by the relays in the event of
grounded faults can be described by Table I. In Table I, and
when is the input signal. are zero-sequence and negative-sequence current phasors at
MMG may be used to depress the steady-state components of a relay location referred to phase A, respectively. According to
a signal and extract the transient features. However, the elemen- the table, the 360 phase plane is divided into six zones. The
tary MMG filter has failed to detect faint slow signal changes table is based on the analysis of the relationship between
in some cases, as mentioned in [14]. Considering fine charac- and for various grounded faults, as given in [8]. To make
teristics of MMG of transient component extraction, an -order the understanding of Table I convenient, the characteristics of
ZOU et al.: IMPROVED PHASE SELECTOR FOR UNBALANCED FAULTS 1849

TABLE I during power swings using superimposed components (fault

DESCRIPTION OF THE RELATIONSHIP BETWEEN ZERO- AND component) of modular current in conjunction with an SMMG
NEGATIVE-SEQUENCE CURRENTS AT RELAY LOCATION OF GROUNDED FAULTS
is proposed here.
Generally, a superimposed component-based phase selection
scheme can give more satisfactory performance because of its
immunity to the effects of load currents. By using an appro-
priate mathematical transform, three coupled phases can be un-
coupled, obtaining three independent quantities, known as mode
0 or ground mode, mode 1 and mode 2, or aerial modes com-
ponents, respectively. Because the modular transformation is a
linear one, the superimposed component theory is also suitable
and for single-phase-to-ground faults and two phases for the modular components. The superimposed postfault mod-
to ground faults are presented briefly in Appendix A. ular voltage sources , and , for various fault types are
It can be seen that the fault type and phases can be imme- calculated in [15] and [16]. For a single-phase- (e.g., phase A)
diately determined for cases ii), iv), and vi), whereas for case to-ground fault, we have
i), iii), and v), the fault type cannot be directly identified since
there are two possible fault types–a single-phase-to-ground fault (10)
and a two-phases-to-ground fault, in each case. Since there are
two potential fault types that are for each, cases i), iii), and v)
do not possess common phases, a widely adopted method for For two-phases (e.g., phase B and phase C)-to-ground fault
discriminating between these two fault types is to examine the
apparent impedance between the two faulty phases related to the
possible two-phases-to-ground fault measured by the phase dis- (11)
tance relay. If the measured impedance is smaller than the set-
ting impedance, the two-phases-to-ground fault is selected; oth-
erwise, the single-phase-to-ground fault is selected. Field oper- For a phase-to-phase (e.g., phase B to phase C) fault
ation experience shows that this SCPS can provide satisfactory
performance in most scenarios. (12)
According to dynamic simulation results and theoretical anal-
ysis, however, the scheme maloperates in some cases during
power swings. For instance, consider a dual-generator power where and are the mode 0 and mode 1 impedances seen
transmission system with a generator at each end of the trans- from the fault point, and is the prefault voltage of phase
mission line, respectively, and suppose a phase-A-to-ground A at the fault point. The derivation of (10)–(12) is given in
fault occurs during a power swing. The distance relay installed Appendix B. The superimposed modular currents in the branch
at a busbar using SCPS is possible to wrongly identify the fault to which the relay is connected are stimulated completely by the
as a phase B and C to ground one. If there is a long electrical respective superimposed modular voltage sources; hence, the
distance between the swing center and the relay location, when characteristics of the former are similar to those of the latter.
a phase A to ground fault occurs and when the power angle It can be seen from (10)–(12) that superimposed mode 2 mod-
between the two generators reaches approximately 180 , the ular current referred to as the faulty phase will be zero for a
apparent impedance between phase B and C calculated by single-phase-to-ground fault, while for the phase-to-phase fault,
the phase distance relay may be lower than the setting the superimposed mode 1 modular current referred to as the
impedance, which results in an invalid judgment. In order to healthy phase will be zero. These characteristics are useful for
solve this problem, a time delay is proposed in [7]. However, the fault phase identification. One of the most widely used algo-
more response time is required by this measure. And the time rithms in digital protection for the extraction of superimposed
delay could be a considerable one since may be still well current can be described using the following equation:
lower than the setting impedance even when the power angle
between the two generators deviates alot from 180 if the swing (13)
center is close to the relay location.
In regards to ungrounded faults, the impedance measurement where is the latest sampling number, and is the samples
method also faces a similar problem. per cycle. However, this algorithm will no longer be applicable
under power swing conditions due to the continuous change of
the measured currents’ amplitudes and the frequency deviation
IV. ALGORITHM caused by the power swing. SMMG is used to extract super-
imposed current in this paper. Simulation results show that the
A. Principle of the Improved Phase Selector During
variation of a signal can be reflected by the local modulus max-
Power Swings
imum of its SMMG coefficients. Thus, the superimposed current
In order to overcome the disadvantage mentioned in components can be effectively estimated by SMMG even under
Section III, an improved unbalanced fault phase selector the power swing condition.
1850 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 4, OCTOBER 2006

TABLE II one point in a cycle. The modulus maximum of the SMMG

CHARACTERISTICS OF THE OUTPUT PATTERNS OF THE SMMG coefficients during the cycle is denoted by ;
FILTER FOR GROUNDED FAULTS
2) multiplied by a reliability coefficient are
taken as a criterion, denoted by ;
3) each SMMG coefficient calculated afterwards is compared
with and . By updating the data window by
one sample, a new SMMG coefficient can be cal-
culated. If , return to the beginning of
step 3). If , assign the value of
to , and return to step 2). If
TABLE III
CHARACTERISTICS OF THE OUTPUT PATTERNS OF THE SMMG , record the latest sampling number by and go to
FILTER FOR UNGROUNDED FAULTS step 4);
4) the SMMG-based improved phase selector for either
grounded faults or ungrounded faults will be launched
depending on whether the zero-sequence current exceeds
a setting threshold.
In the event of grounded faults, SCPS is applied first. If a
grounded fault is identified as either a single-phase-to-ground
one or a two-phases-to-ground one according to the existing
SCPS (i.e., case i), iii), or v) in Table I), s1–s3 are then em-
Using a matrix transform, six modular current signals
ployed for further phase identification. The SMMG transforms
can be derived from three-phase currents,
are carried out one point by one point from the sampling point
described as follows:
to . Thus, for each of the three modular currents,
ten SMMG coefficients are generated. The modulus maximum
(14)
of the ten SMMG coefficients, or three for s1, s2,
where or s3, respectively, is used to form the phase identification crite-
rion as follows. If the fault is a single-phase-to-ground one, we
have

(15)

where is a setting threshold, which will be discussed in

Factually, s1, s2, and s3 are mode 2 modular currents of the the next section; subscript are 1, 2, 3, respectively,
Clark transform referred to as phase A, B, and C respectively; for phase-A-to-ground fault; 2, 1, 3 for phase-B-to-ground
s4, s5, and s6 are mode 1 modular currents of the Clark trans- fault; and 3, 1, 2 for phase-C-to-ground fault. If the above
form referred to phase A, B, and C, respectively. A criterion is not satisfied, the fault is simply discriminated as a
filter is designed for the modular currents in this paper. two-phases-to-ground one. Obviously, criterion (15) is equiva-
The filter length is ten samples according to (9). By feeding lent to the following one:
these modular currents to the SMMG filter, the SMMG coef-
ficients of the six modular currents, m1–m6, can be obtained. (16)
The characteristics of the output patterns of the SMMG filter for
Hence, , denoted by , can be regarded as
grounded faults and ungrounded faults are presented in Tables II
the redundancy of the criterion.
and III, respectively. “1” in the tables denotes a major modulus
With regard to ungrounded faults, the other three modular
maximum occurring in the SMMG coefficient after fault, while
currents, s4–s6, are used to perform phase discrimination. Anal-
“0” means no obvious major modulus maximum.
ogous to that for grounded faults, phase identification criterion
B. Description of the Phase Selection Procedure for ungrounded faults is constituted using the modulus maxima
of the SMMG coefficients of the modular currents
The synchronous frequency is 50 Hz and the sampling rate described as follows:
used in this paper is 1.8 kHz (i.e., 36 samples per cycle).
(17)
Once the power swing blocking has been initiated, a self-
adaptive fault detection module is activated, described as fol- or
lows. For any one of the six modular currents: (18)
1) the signal processing based on SMMG starts from half a
cycle after the beginning of the power swing blocking state. where is a setting threshold; subscript are 4, 5, 6, re-
Using a data window of ten samples, an SMMG coeffi- spectively, for the phase B to C fault, 5, 4, 6 for the phase C to
cient can be obtained. By updating the data window by one A fault, and 6, 4, 5 for the phase A to B fault. The redundancy
sample, a new SMMG coefficient can be calculated, and of this criterion is , denoted by .
so on. The SMMG coefficients are calculated one point by Fig. 2 shows the flowchart of the improved phase selector.
ZOU et al.: IMPROVED PHASE SELECTOR FOR UNBALANCED FAULTS 1851

extensive series of fault scenarios has been carried out on the

model system. Nine types of fault (single phase to ground, two
phases to ground, phase to phase fault) are simulated. There are
ten fault locations along circuit 2 line with steps of 34 km. The
fault inception angles of 0 , 90 , 180 , and 270 (these angle
values are estimated with reference to a synchronous frequency
of 50 Hz) from phase-A voltage zero crossing are taken into ac-
count. The ground resistances of single-phase-to-ground faults
considered are , 50, and 200 .

B. Analysis of Simulation Results

It is assumed that the first zones of the ground distance relay
and the phase distance relay at R3 are both set to protect the
circuit 2 line of 80% of the line length. The setting impedance
accordingly.
Suppose a solid phase-A-to-ground fault occurs on circuit 2
at a distance of 102 km from R3. The fault occurs at 350 ms. The
power swing frequency is approximately 5 Hz. The normalized
zero-sequence and negative-sequence currents and are
plotted in Fig. 4(a), where the solid line represents and the
Fig. 2. Flowchart of the improved phase selector. dashed line represents . The negative-sequence component
is obtained by the following algorithm:

(19)

where is the latest sampling number, and is the sam-

ples per cycle. As seen, a data window of samples is
required to obtain negative-sequence current series. The locus
of calculated by the full-cycle Fourier approach
Fig. 3. Model power system. (FCFA) is plotted in Fig. 4(b). According to the simulation
results, the error induced by the frequency deviation caused
by the power swing is generally too small to upset the phasors
V. SIMULATION AND ANALYSIS comparison result. Obviously, this fault is of case i) in Table I.
A. Description of the Simulation Model Let us examine the apparent impedance between phase B and
C then. Because measured by the distance relay at R3 is
Using PSCAD/EMTDC, the simulation model of a typical after fault inception, the fault is wrongly identi-
500-kV transmission system is set up. The model system con- fied as a phase-B- and C-to-ground fault by the existing SCPS.
sists of an 800-MVA generator supplying power to an infinite Fig. 4(c)–(e) shows the normalized three modular currents, s1,
bus. The generator is connected to the infinite bus through a s2, s3, (dotted lines) measured by relay R3 and their SMMG
transformer and a 500-kV, 340-km double-circuit trans- coefficients (solid lines). The horizontal axis represents the
mission line as illustrated in Fig. 3. The distance relay consid- time (in milliseconds), and the vertical one is the normalized
ered is installed at point R3. A disturbance created by a three- magnitude. It is apparent that the magnitude of is much
phase fault occurring at F1 on circuit 1, which is cleared after smaller than that of and after the fault inception. The
a time delay of 300 ms by disconnecting the faulty line is by fault type and phase are clearly identified by the proposed
opening the breaker CB1 and CB2. This disturbance will cause approach.
the generator to accelerate or decelerate resulting in a power Fig. 5 presents the case of a phase-B-to-ground fault with a
swing. The parameters of each circuit of the transmission line ground fault resistance at a distance of 306 km
are as follows: from point R3. The fault occurs at 360 ms. The locus of
is shown in Fig. 5(a). The normalized three
modular currents s1, s2, s3, (dotted lines) and their SMMG
coefficients (solid lines) are shown in Fig. 5(b)–(d). It is ob-
vious that the fault phase is correctly identified by SMMG with
In order to verify the proposed phase selector, a variety of sufficient sensitivity. It should be indicated that the redundancy
faults is applied on circuit 2 during the following power swing. of the criterion (15) and (16), namely , is the smallest
In the simulation, the phase voltages and currents at any node are for this scenario among all of the single-phase-to-ground
provided by EMTDC. All faults on circuit 2 are applied when faults we have simulated. In this case, is equal to 29.99,
the power angle between the two generators is close to 180 . An denoted by . Fig. 6 gives the case of a phase-C- and
1852 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 4, OCTOBER 2006

Fig. 5. b-G fault with R = 200 at 306 km and 360-ms fault inception.
(a) The angle by which the negative-sequence current phasor leads the zero-
sequence current phasor. (b) Modular current s1 and its SMMG coefficients.
(c) Modular current s2 and its SMMG coefficients. (d) Modular current s3 and
its SMMG coefficients.

the proposed criterion both highly sensitive and reliable. For

instance, let a reliability coefficient , then
Fig. 4. Solid a-G fault at 102-km and 350-ms fault inception. (a) Zero-sequence , the sensitivity coefficient
and negative-sequence currents. (b) The angle by which the negative-sequence .
current phasor leads the zero-sequence current phasor. (c) Modular current s1
and its SMMG coefficients. (d) Modular current s2 and its SMMG coefficients. In Fig. 7, the normalized modular currents, s4, s5, and s6
(e) Modular current s3 and its SMMG coefficients. (dotted lines), measured at R3 and their SMMG coefficients
(solid lines) are presented for a phase-C-to-A fault occurring
at a distance of 306 km from point R3. The fault inception
A-to-ground fault at a distance of 102 km from point R3. time is 360 ms. Because the magnitude of is much smaller
The fault occurs at 360 ms. It can be seen from the figure than that of and after the fault inception, the fault is
that and are all of high values. Hence, the fault correctly identified as a phase-C-to-A fault. Among all of the
is correctly discriminated as a two-phases-to-ground one by phase-to-phase faults simulated, for this scenario is the
the proposed algorithm. for this case has the largest value smallest, which is equal to 36.65, denoted by . Thus,
among all of the two-phases-to-ground faults we have simu- in (17) and (18) can be set as follows. Let a sensitivity
lated. In this case, is equal to 2.68, denoted by . coefficient , then .
According to the simulation results, the proposed scheme is The performance of the proposed approach with a different
able to make correct judgments even for the most unfavorable power swing center location and power swing frequency has
scenarios. for a single-phase-to-ground fault is much larger also been verified. By modifying the parameters of the generator
than that for a two-phases-to-ground fault. By this virtue, it is and the transformer in Fig. 3, the power swing center is moved
easy to set an appropriate coefficient in (15) and (16) to make to point R3. Fig. 8 gives the case of a solid phase-A-to-ground
ZOU et al.: IMPROVED PHASE SELECTOR FOR UNBALANCED FAULTS 1853

Fig. 7. c–a fault at 306 km and 360-ms fault inception. (a) Modular current s4
and its SMMG coefficients. (b) Modular current s5 and its SMMG coefficients.
(c) Modular current s6 and its SMMG coefficients.

Fig. 6. c-a-G fault at 102 km and 360-ms fault inception. (a) Angle by which
the negative-sequence current phasor leads the zero-sequence current phasor.
(b) Modular current s1 and its SMMG coefficients. (c) Modular current s2 and
its SMMG coefficients. (d) Modular current s3 and its SMMG coefficients.

fault occurring at R3 at a lower power swing frequency of ap-

proximately 1 Hz. It is obvious that the fault type and phase are
clearly identified by the proposed approach.
The fault phase can be identified reliably by the proposed
scheme within one-and-a-half cycles after the fault inception
during a power swing if the current phasors are calculated using
FCFA. In this sense, this improved phase selector during power
swings can be regarded as a fast operation element.

VI. CONCLUSION
An improved phase selector for an unbalanced fault during a
power swing is proposed in this paper by utilizing the SMMG
transform to extract the superimposed components of modular
currents. The theoretic analysis and the simulation results enable
the following conclusions.
Fig. 8. A–G fault at 0 km and 800-ms fault inception. (a) Modular current s1
1) The superimposed currents can be effectively and rapidly and its SMMG coefficients. (b) Modular current s2 and its SMMG coefficients.
extracted during the power swing condition by SMMG; (c) Modular current s3 and its SMMG coefficients.
hence, the proposed phase selection scheme has satisfac-
tory responding speed, sensitivity, and reliability.
2) The SMMG-based phase selection schemes are able to 3) The fault location and fault inception angle have little in-
cope with large ground fault resistances. fluence on the phase selection approach.
1854 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 21, NO. 4, OCTOBER 2006

4) The proposed algorithm is simple and has no special hard- where

ware requirement; hence, the schemes can be implemented
on standard relaying hardware.

APPENDIX A
Therefore, the boundary condition can be converted into
For a single-phase (e.g., phase A)-to-ground fault, we have
(23)
(20)

where denotes the superimposed postfault

where and are zero-sequence and negative-sequence mode modular voltage source referred to as phase A;
current in the fault branch; and are the distribution fac- is the mode modular impedance seen from the fault point;
tors of the zero-sequence and negative-sequence current, respec- and is the superimposed mode modular current in the
tively. Because the phase angle of is equal to that of , the fault branch.
difference between the phase angles of and is deter- In addition, we have
mined by the relationship of . In extremely high-voltage
(EHV) or ultra-high-voltage (UHV) power systems, the phase
angle of is generally very close to that of ; the differ- (24)
ence between the phase angles of and , therefore, is very
small. Then, (10) can be obtained by solving (23) and (24).
For a two-phases (e.g., B and C)-to-ground fault, we have For two-phases (e.g., phase B and phase C)-to-ground fault,
we have boundary condition

(21)
Using (22), this condition can be converted into

where and are the system impedances seen from the (25)
fault port in the positive- and zero-sequence equivalent network,
respectively. Thus, the difference between the phase angles of Then, (11) can be obtained by solving (24) and (25).
and are mainly determined by the relationship of For a phase-to-phase (e.g., phase-B-to-phase-C) fault, we
and . The angle by which leads , increases have boundary condition
with the increase of the ground resistance. The angle will ap-
proach 90 when the ground resistance approaches infinity.
Using (22), this condition can be converted into
APPENDIX B
The inception of a fault causes the prefault steady-state volt- (26)
ages and currents to deviate to their postfault values. A postfault
power system can be expressed as the superposition of a super-
Then, (12) can be obtained by solving (24) and (26).
imposed fault power system on the prefault power system. The
superimposed fault power system can be considered as that with
REFERENCES
a fictitious superimposed voltage source equal in magnitude and
[1] A. Mechraoui and D. W. P. Thomas, “A new principle for high resis-
opposite in sign to the prefault voltage at the fault point. tance earth fault detection during fast power swings for distance pro-
For a single-phase (e.g., phase A)-to-ground fault, we have tection,” IEEE Trans. Power Del., vol. 12, no. 4, pp. 1452–1457, Oct.
the boundary condition 1997.
[2] A. Mechraoui and D. W. P. Thomas, “The influence of power swing
on transmission line distance protection, measurement of power and
voltage phasor oscillation at the relaying point,” in Proc. UPEC, Sep.
1993, vol. 2, pp. 585–588.
[3] Z. D. Gao and G. B. Wang, “A new power swing block in distance pro-
where and denote the phase superimposed tection based on a microcomputer principle and performance analysis,”
postfault voltage source; is the phase superimposed current in Proc. Inst. Elect. Eng., Int. Conf. Advances in Power System Control,
in the fault branch; is the prefault voltage of phase at the Operation and Management, Hong Kong, China, 1991, pp. 843–847.
[4] J. Machowski and D. Nelles, “New power swing blocking method,”
fault point. in Proc. Inst. Elect. Eng., Int. Conf. Developments in Power System
Three-phase quantities can be expressed by three modular Protection, Mar. 1997, pp. 218–221.
quantities [5] A. Mechraoui and D. W. P. Thomas, “A new blocking principle with
phase and earth fault detection during fast power swings for distance
protection,” IEEE Trans. Power Del., vol. 10, no. 3, pp. 1242–1248,
(22) Jul. 1995.
ZOU et al.: IMPROVED PHASE SELECTOR FOR UNBALANCED FAULTS 1855

[6] Q. X. Yang, Z. Y. Xu, L. L. Lai, Z. H. Zhang, and N. Rejkumar, “Fault Li Zou was born in Wuhan, China, on August 10, 1978. He received the M.E.
identification during power swings with symmetrical component,” degree from the Huazhong University of Science and Technology (HUST),
in Proc. EMPD Int. Conf. Energy Management and Power Delivery, Wuhan, China, where he is currently pursuing the Ph.D. degree.
1998, pp. 108–111. His research interests include digital protection relaying, power quality anal-
[7] S. Zhu, Principles and Applications of Protective Relaying of High- ysis, and advanced digital signal processing applications in power systems.
Voltage Grid (in Chinese). Beijing, China: China Power Press, 1995.
[8] Z. Y. Xu, Q. X. Yang, W. S. Liu, and Z. H. Zhang, “A sequence fault
phase selector for transmission line protective relay,” in Proc. CSEE (in
Chinese), May 1997, vol. 17, no. 3, pp. 214–216. Qingchun Zhao received the M.E. degree from the Huazhong University of
[9] J. L. Suonan, Q. Q. Xu, G. B. Song, R. S. Li, and Y. Z. Ge, “Analysis on Science and Technology (HUST), Wuhan, China.
the operating characteristic of the sequence-component fault phase se- His research interests include digital protection relaying and advanced digital-
lector during power swings,” (in Chinese) Automat. Elect. Power Syst., signal-processing applications in power systems.
vol. 27, no. 2, pp. 52–55, Jan. 2003.
[10] P. Sun, J. F. Zhang, D. J. Zhang, and Q. H. Wu, “Morphological iden-
tification of transformer magnetizing inrush current,” Inst. Elect. Eng.
Electron. Lett., vol. 38, no. 9, pp. 437–438, Apr. 2002.
[11] Q. H. Wu, J. F. Zhang, and D. J. Zhang, “Ultra-high-speed directional Xiangning Lin received the M.Sc. and Ph.D. degrees in electrical engineering
protection of transmission lines using mathematical morphology,” from the Huazhong University of Science and Technology (HUST), Wuhan,
IEEE Trans. Power Del., vol. 18, no. 4, pp. 1127–1133, Oct. 2003. China, in 1996 and 1999, respectively.
[12] J. Serra, Image Analysis and Mathematical Morphology. London, Currently, he is an Associate Professor with HUST. His research interests
U.K.: Academic Press, 1982. include modern signal processing and their applications in power systems, pro-
[13] D. Wang and D. C. He, “A fast implementation of 1-D grayscale mor- tective relaying, power system control, and analysis.
phological filters,” IEEE Trans. Circuits Syst. II, Analog Digit. Signal
Process., vol. 41, no. 9, pp. 634–636, Sep. 1994.
[14] L. Zou, P. Liu, and Q. Zhao, “A mathematical morphology based phase
selection scheme in digital relaying,” in Proc. Inst. Elect. Eng., Gen. Pei Liu received the M.S. degree from the Huazhong University of Science and
Transm. Distrib., Mar. 2005, vol. 152, no. 2, pp. 157–163. Technology (HUST), Wuhan, China.
[15] Y. Ge, New Types of Principles and Technologies of Protection and Currently, she is a Professor with the College of Electric and Electronic En-
Fault Location (in Chinese). Xi’an, China: Xi’an Jiaotong Univ. gineering with HUST. Her research interests include digital protection relaying,
Press, 1996. power quality analysis, and wavelet applications in power systems.
[16] A. D. Wang, Y. Z. Ge, L. L. Lai, and A. T. Johns, “Application of a
new technique to HV transmission line distance relaying,” in Proc. Int.
Conf. Advances in Power System Control, Operation and Management,
Hong Kong, China, 1991, pp. 449–454.