Sunteți pe pagina 1din 9

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.2016.2601379, IEEE
Transactions on Smart Grid

Wide-area Fault Location of Transmission Lines by


Hybrid Synchronized/Unsynchronized Voltage
Measurements
Ahmad Salehi Dobakhshari

Abstract—Global positioning system (GPS) has been deployed


in phasor measurement units (PMUs) and protective relays to I+f Positive-sequence fault current at fault point.
provide synchronized measurements across the grid. Synchro- I +,init
f Initialized positive-sequence fault current at fault
nized measurements, however, may not be available throughout point.
the network either due to GPS signal loss (GSL) or basically
V i,+
p Positive-sequence pre-fault voltage measurement at
lack of required infrastructure for synchronized sampling. This
paper presents a novel method for wide-area fault location bus p.
(WAFL) incorporating both synchronized and unsynchronized V f,+
p Positive-sequence post-fault voltage measurement at
voltage measurements. Complex calculus is utilized to develop bus p.

a system of equations based on phase angles of synchronized
∆V + Magnitude of difference of pre- and post-fault volt-

p
measurements and magnitudes of both synchronized and un-
∆V
synchronized measurements. The distributed-parameter model of age phasors at bus p.
the transmission lines are considered and therefore the obtained 6 ∆V +
p Phase angle of difference of pre- and post-fault
results are also valid for long transmission lines. Moreover, the voltage phasors at bus p.
proposed formulation locates the fault irrespective of fault type. Z ip i-p element of the positive-sequence bus impedance
Electromagnetic transient simulations for WSCC 9-bus and a
matrix of the network.
22-bus sub-network of the IEEE 118-bus test system reveal that
in the face of GSL, unsynchronized measurements significantly Ẑ pf p-f element of the modified positive-sequence bus
improve fault-location estimation when the synchronized mea- impedance matrix of the network including virtual
surements are sparse. bus f .
Index Terms—Bus impedance matrix, Complex calculus, Fault
location, Global Positioning System, Unsynchronized measure- II. I NTRODUCTION
ments.

I. N OMENCLATURE R APID development of information and communication


technologies (ICT) in substations parallels recent invest-
ments in transmission lines [1]. To bridge the gap between ICT
f Index of fault point.
i, j Indices of faulted line terminals. and power transmission infrastructure in smart transmission
p Index of bus whose voltage is measured. grids, wide-area fault location (WAFL), utilizing sparse GPS-
m Number of buses whose voltages are measured. synchronized measurements, has been put forth recently [2].
n Number of synchronized voltage measurements (n WAFL can generally be categorized into wide-area traveling
≤ m). wave-based [3]–[7] and impedance-based [8]–[19] methods.
x Unknown fault location in per unit measured from While wide-area traveling-wave-based methods may yield
bus i. more accurate results, they need very high sampling rates for
N Total number of network buses. measurements [20]. Impedance-based methods, on the other
lij Length of the faulted line. hand, rely on GPS-synchronized measurements obtained from
γ ij Propagation constant of the faulted line. sparse phasor measurement units (PMUs), whose sampling
Zc Characteristic impedance of the faulted line. rates are much lower. Using bus impedance matrix (BIM),
X State variables. Liao [8] presented the pioneering work in WAFL, which
M Measurements Vector. needed only two voltage measurements, taken from remote
F Vector function of state variables. substations. The idea has been expanded by Kang et al.
J Jacobian matrix. to incorporate multiple synchronized voltage [9] and current
[10] measurements in WAFL formulation. The foregoing for-
mulation cancels complex fault current variable by dividing
The author is with the Department of Electrical Engineering, Faculty
of Engineering, University of Guilan, Rasht 41996-13776 Iran (e-mail: available voltage/current measurements. Jiang et al. [11] and
salehi ahmad@guilan.ac.ir). Dobakhshari et al. [12] use iterative optimization-based meth-

1949-3053 (c) 2016 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.2016.2601379, IEEE
Transactions on Smart Grid

ods for fault location, where the estimation best explains the x 1–x
measured values. A different approach by Azizi et al. [13] c-phase
and Dobakhshari et al. [14] formulates the problem as an b-phase

estimation problem, where fault location can be obtained in


a-phase f
closed-form and measurement errors can be taken into account. Bus i Bus j
However, when a PMU loses GPS signal, the corresponding
measurement may not contribute to the WAFL formulation. The Remaining Transmission Network
GSL has a more severe impact on the wide-area back-up
protection scheme suggested by Neyestanaki et al. [15], where (a)
loss of synchronization makes the method futile. On the one •• sinh(••• ••• •) Bus f •• sinh[••• ••• (1– •)]
hand, many commercial PMUs experience GSL on a daily ba-
2 ••• ••• • 2 ••• ••• (1– •)
sis [21]. On the other hand, GPS-synchronized measurements tanh( ) •• tanh[ ]
•• 2 • •• 2
Bus i Bus j
are not still ubiquitous in many power systems, where slow
communication links are utilized for measurements [22]. To The Remaining Positive-Sequence Network
overcome this issue, Dobakhshari et al. [16] present a WAFL
formulation using only unsynchronized voltage measurement.
(b)
However, valuable phase-angle measurements are dismissed in
this latter formulation. Fig. 1. Faulted network (a) Three-phase representation (b) Positive-Sequence
This paper presents a novel method to solve WAFL using circuit model.
hybrid synchronized/unsynchronized voltage measurements.
is modeled by a current source I + f . Applying superposition
While the magnitude and phase angle of synchronized mea-
principle to this circuit model yields [25]
surements are available, the unsynchronized measurements are
incorporated in the proposed formulation by their magnitudes. V f,+ = V i,+ +
p p − Ẑ pf I f (1)
The reason is that their phase angles are absent due to GSL.
Utilizing BIM, the problem is formulated as a during-fault where V f,+p and V i,+
p are the positive-sequence post- and
state estimation [23] of the system. In this estimation problem, pre- fault voltage phasors measured at bus p, respectively. I + f
the measured values are sparse synchronized/unsynchronized is the positive-sequence fault current at fault point. Ẑ pf is the
measurements and the unknowns are fault location and positive-sequence p-f element of the modified bus-impedance
positive-sequence fault current at fault point. The phase angle matrix given by (2) shown at the bottom of the next page [8],
and magnitude of each voltage measurement are expressed as [14].
functions of fault location by complex calculus [24]. Since the Ẑ pf is a function of fault location (x). Therefore, (1) can
proposed formulation is nonlinear and solved iteratively by be re-written for voltage measurement at bus p as
Newton-Raphson method, a procedure for estimating initial
values of unknown phase angle and magnitude of positive- ∆V + I+
p = Ẑ pf (x)I f (3)
sequence fault current is introduced. The rest of the paper
is organized as follows. In Section III, the application of It is shown in [8] and proved in [14] that the elements of
BIM for formulating WAFL in symmetrical components is Ẑ pf (x) can be expressed in terms of the elements of the
demonstrated. Sections IV and V respectively express the original BIM as shown in (2), where complex values are
magnitudes and phase angles of voltage measurements in printed in bold.
terms of fault location and total fault current. Section VI
provides initial values for unknown variables in the proposed IV. U SING M AGNITUDE OF S YNCHRONIZED AND
iterative-based formulation. Section VII is devoted to case U NSYNCHRONIZED VOLTAGE M EASUREMENTS
studies, followed by the conclusion in Section VIII.
When the voltage measurements are not synchronized, e.g.
III. FAULTED N ETWORK IN S YMMETRICAL C OMPONENTS when a slow communication link is used for fault location or
GSL occurs, only the magnitudes of voltage measurements
Any fault along a transmission line may be modeled by are available. Getting the magnitude of two sides of (3)
three independent networks of symmetrical components [25]. results in the following set of equations for unsynchronized
Fig. 1(a) shows a faulted network, where a single-line-to- measurements.
ground fault has occurred at point f along line i-j. Fig. 1(b)
depicts the equivalent positive-sequence circuit model of the
+
∆V +
p = Ẑ pf (x) I
If (4)

same faulted network in which fault current at the fault point
∆V

1949-3053 (c) 2016 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.2016.2601379, IEEE
Transactions on Smart Grid

V. U SING P HASE A NGLE OF S YNCHRONIZED VOLTAGE


M EASUREMENTS ∂F
J= , (10)
∂X X=Xt
Thus far, the phase-angle data of synchronized measure-
ments have not been included in WAFL formulation in (4). The iterative process (8)-(10) is terminated when ∆Xt
In what follows, these valuable data are also integrated within reaches a specified threshold for the estimate accuracy. Refer-
the formulation of fault location so that a more accurate result ring to (6), we can calculate the non-zero elements of J related
can be obtained by fully achieving the potential of phase-angle to voltage magnitude measurements by partial derivatives of
measurements. To deploy the phase-angle measurements of n F as follows.
synchronized voltage measurements in our formulation, the
∆V +
∂ ∆V ∂ Ẑ (x)

following can be deduced from (3). p pf
Jp,1 = = II +
f (11)
+
∂x ∂x
6 ∆V + 6 6
p = Ẑ pf (x) + I f (5)

∆V +
∂ ∆V

In this way, phase-angle measurements of synchronized p

Jp,2 = = Ẑ (x) (12)

pf
measurements can be integrated in (5) to fully utilize the

+
∂ II f
potential of hybrid synchronized and unsynchronized measure-
ments as for p=1, ..., n, n+1, ..., m. The only problematic elements
    of J above are those given by (11), whose calculation requires
∆V +
|∆V 1| Ẑ 1f (x) II +

expanding the magnitude of Ẑ pf in (2). Nevertheless, utilizing
f 

 .
..
 
  ..  a novel formula proved in Appendix A, (11) can be calculated
   . 
    as follows.
 ∆V +   Ẑ mf (x) II +
   
f 

 m = (6) ( )

+ 
 
 6 Ẑ (x) + 6 I + 
 ∂Ẑ pf (x) ∗
 6
 ∆V 1   Re Ẑ pf (x)

1f f 
∆V +
∂ ∆V ∂x

.. p

 ..    +

.
 
.
 = II f , (13)
    ∂x Ẑ pf (x)

6 ∆V + 6 Ẑ nf (x) + 6 I +
n f

∆V + ∆V + + + T where (.)∗ denotes the conjugate of the complex argument. As


1 |, ..., |∆V m |, ∆V 1 , ..., ∆V n ] denote
Let M = [|∆V 6 6

the measurement vector and X = [x, |II + + T can be seen from (13), these elements of Jacobian matrix can
f |, I f ]
6 the un-
known variable vector. Equation (6) can be re-written in a easily be calculated without the need for separating the real
compact form as and imaginary parts of the complex variable Ẑ pf in (2).
Equations (11)-(12) give elements of Jacobian related to
M − F (X) = 0, (7)
the voltage magnitude measurements. The remaining non-zero
where F is a vector function relating the unknown variables to elements of J corresponding to phase-angle measurements can
the measurements as given by the right side of (6). If there are be calculated by partial derivatives of F as follows.
two or more voltage phasor measurements, an overdetermined
∂ 6 ∆V +
p ∂ 6 Ẑ pf (x)
system of equations is resulted from (5), which can be solved Jp,1 = = , (14)
∂x ∂x
by the Newton-Raphson method [25], iteratively, as follows

Xt+1 = Xt + ∆Xt , (8) ∂ 6 ∆V + p


Jp,3 = = 1, (15)
∂6 I +
f
where subscript t denotes the iteration number. The incremen-
tal change in each iteration is obtained by: for p=m+1, ..., n. It must be noted that since Z ip and Z jp , p=1,
..., n, as well as γ ij , are complex quantities, the calculation of
−1 
phase-angle function in (5) in terms of x is cumbersome; it is

∆X t = JT J JT M − F (Xt ) ,

(9)
required to calculate the real and imaginary parts of RHS of
where superscript T denotes the transpose operator and J is (2) in terms of x and then obtain the phase angle by inverse
the Jacobian matrix calculated as follows tangent function. Instead, as proved in Appendix B, RHS of

Z ip Z jp
+
sinh(γγ ij lij x) sinh(γγ ij lij (1 − x))
Ẑ pf (x) =     (2)
1 1 γ ij lij γ ij lij
+ + tanh x + tanh (1 − x)
sinh(γγ ij lij x) sinh(γγ ij lij (1 − x)) 2 2

1949-3053 (c) 2016 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.2016.2601379, IEEE
Transactions on Smart Grid

sequence qu
(14) can be calculated by (B.4) in Appendix B, where Z (x) cannot be lo
should be replaced with Ẑ pf . Nevertheless
( ) type, in pract
∂Ẑ pf (x) ∗
Re Ẑ pf (x) Ẑ pf (x)
∂ Ẑ pf (x) ∂x
− 2
∂x

F
Ẑ (x)
+

∂ 6 ∆V p pf
= −j INCLUDE IN
∂x Ẑ pf (x)
(16)
Actual
To analyze the computational burden of the proposed Faulted
method, it should be noted that the Jacobian matrix J is a
2(m + n) by 3 matrix, and thus JT J inverted in (9), is a 3 by
3 matrix. Therefore, matrix inversion in (9) is computationally
inexpensive as a 3 by 3 matrix is only needed to be inverted.
This is not the case for some of the previously reported wide-
area fault-location methods [9], [10].

VI. I NITIALIZING N EWTON -R APHSON M ETHOD


In order to initialize the unknown variable vector X = Fig. 2. Single-line-diagram of the 9-bus WSCC System.
[x, |II + 6 + T
f |, I f ] , it is reasonable to initialize fault point at the Fig. 2. Single-line diagram of the WSCC 9-bus system.
midpoint of the line, i.e. x = 0.5 pu. Next, each measurement Different types of faults at different points of each system
at bus p can be used to initialize |II + +
f | and I f . In order to
6
are located using the proposed WAFL method.
+
obtain |II f |, it is sufficient to use x = 0.5 in (4) and simplify In practice, an analog anti-aliasing filter is first used for
(2) to obtain: rejecting high-frequency harmonics of measurements. To sim-

γ ij lij
ulate the analog anti-aliasing filter, the signals are first sam-
2 cosh( ) ∆V +

p
2
∆V
pled with a high sampling rate (100 KHz) and then passed

+,init
II f,p = , (17)

Z ip + Z jp through a digital anti-aliasing filter, which is a second-order
Butterworth filter with a cut-off frequency of 350 Hz. The
where II +,init
+
is initial estimate of II f obtained by voltage

f,p output of the anti-aliasing filter is next downsampled to 4800
measurement at bus p. Now, all measurements at buses 1, ..., m Hz (96 samples per cycle), which is considered to be PMU
can be utilized to estimate the magnitude of total fault current: sampling rate. Finally, the discrete Fourier transform (DFT)

γ ij lij +
is applied to extract the fundamental component of the input
m cosh( ) ∆V p
2 X 2 signal. The phasor values used in the proposed algorithm are
∆V
+,init
I
I f = . (18)
m obtained by averaging the magnitude and phase angle of time-
Z ip + Z jp
p=1
varying phasors over an interval of one cycle starting two
Likewise, (5) is utilized to initialize the phase angle of total cycles after fault inception. The simulated voltage phasors are
fault current using voltage phasor measurement at bus p as input into MATLAB, where the proposed algorithm has been
γ ij lij implemented.
cosh( ∆V +
)∆V p
6 I f,p +,init
=6 2 . (19) Relative difference between actual and estimated fault loca-
Z ip + Z jp tion is the measure of fault-location accuracy, defined by
If all of the measurements at buses 1, ..., n are utilized, the Estimated F L −Actual F L
angle of total fault current can be estimated by F L Error(%) = ×100. (21)
Line Length
γ ij lij The iterative process (7)-(9) will be terminated when the
1 X cosh( 2 )∆V
n ∆V +
p
6 I +,init = 6 . (20) update of (9) for fault location is less than 1e-6 pu.
f
n p=1 Z ip + Z jp
A. Fault Location on WSCC 9-bus Test System
VII. S IMULATION R ESULTS The WSCC 9-bus system, whose parameters can be found
To demonstrate the efficacy of the proposed method, elec- in [26], comprises of 6 transmission lines, three power sources
tromagnetic transient simulations for the western systems and three PMUs, as shown in Fig. 2.
coordinating council (WSCC) 9-bus system and a 22-bus sub- Table I reflects the results of fault location estimation for
network of the IEEE 118 bus test system are carried out. different types of low- and high-resistance faults at different

1949-3053 (c) 2016 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.2016.2601379, IEEE
Transactions on Smart Grid

TABLE I TABLE II
FAULT L OCATION ON THE 9-B US T EST S YSTEM D IFFERENT P RE -FAULT O PERATING C ONDITIONS IN THE 9-B US S YSTEM

FL estimation error (%) Operating


Faulted G1 G2 G3 L5 L7 L9
d (pu) RF (Ω) Fault Type Condition
Line
3PH LL SLG DLG OC1 +10% +10% +10% +10% +10% +10%
0 0.0145 0.0282 0.6596 -0.0219 OC2 -20% -20% +50% +10% 0% 0%
0.1
50 -0.0174 0.0177 0.3146 -0.0922 OC3 +5% +15% -30% -50% +20% +20%
4-5
0 -0.0047 -0.0128 -0.0312 -0.0314
0.5
50 0.0048 -0.0341 0.1273 -0.0496
0 0.0354 -0.0278 0.4597 -0.146
0.9 0.3
50 0.0913 -0.0452 0.2975 -0.1852 OC1
OC2
0 0.0339 0.0162 0.4406 -0.0216 0.25
0.1 OC3
50 -0.0184 0.0015 0.0915 -0.1302
4-9 0.2
0 0.0015 -0.0204 -0.5214 -0.098
0.5
50 -0.0693 -0.0303 -1.3174 -0.1556 0.15

Estimation Error (%)


0 0.0525 -0.0253 -0.592 -0.056
0.9 0.1
50 0.1077 -0.0345 -0.7181 -0.1041
0 -0.0082 -0.0204 -0.1575 -0.0343 0.05
0.1
50 0.0055 -0.0221 -0.077 -0.0668
5-6 0
0 0.0089 -0.0004 0.7894 0.009
0.5
50 0.0036 -0.0027 0.6643 0.021 −0.05
0 0.0448 -0.0021 0.0363 0.0439
0.9 −0.1
50 0.066 0.0019 -0.0032 0.1129
0 -0.0076 -0.0014 -0.2332 0.0084 −0.15
0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
50 -0.0103 -0.0101 -0.7687 -0.0606 Fault Distance
6-7
0 -0.0056 0.0023 0.0783 -0.0114
0.5
50 0.0019 0.0054 -0.0297 -0.022
0 -5E-05 0.0081 -0.2402 0.0312 Fig. 3. Estimation error for different pre-fault operating conditions
0.9
50 0.0009 0.0156 -0.5137 0.0434
0 -0.0013 0.0131 0.2194 0.0224 TABLE III
0.1
50 0.0131 0.0279 -0.1978 0.02117 C OMPARISON OF D IFFERENT WAFL M ETHODS IN FACE OF GSL FOR THE
7-8
0 0.0103 0.0119 0.3859 0.0102 9-B US S YSTEM
0.5
50 0.0228 0.0294 0.2025 0.0158
0 -0.0143 -0.0009 0.1587 0.0567 FL estimation error (%)
0.9
50 0.0273 0.0198 0.3948 0.0275 No GSL GSL at PMU 6
Network WAFL
0 0.0228 0.0124 0.466 0.0157 Case 1 Case 2 Case 3
0.1 Parameters Method
50 0.002 0.0103 0.1886 -0.04 Using ∆V 4 , Using Using ∆V 4 ,
8-9
0 0.007 -0.0027 -0.0094 -0.0043 ∆V 8
∆V 6 ,∆V ∆V 8
∆V 4 ,∆V ∆V 6 |
∆V 8 ,|∆V
0.5
50 -0.005 -0.0021 -0.0804 0.0229 [13] -0.029 -0.08 -0.08
No Error in
0 0.0103 0.0132 0.0854 -0.0041 [15] -0.142 —- —-
0.9 Parameters
50 0.0126 0.0109 -0.083 -0.0444 Proposed -0.022 -0.063 -0.028
[13] 1.195 4.163 4.163
5% Error in
[15] -2.076 —- —-
Parameters
points of the network. As can be seen, all fault cases are pin- Proposed 0.997 3.34 1.192
pointed accurately, using synchronized voltage measurements
at buses 4, 6 and 8. OCs, successfully. The reason is that full capacitance of the
In order to study the effect of pre-fault operating condition line has been taken into account in the formulation, as shown
(OC) on the algorithm accuracy, different load-generation in Fig. 1(b).
patterns in the 9-bus test system are simulated. It is intended
to show that the proposed method successfully locates faults B. Impact of GPS Signal Loss on 9-Bus Test System
regardless of the pre-fault OC. Table II presents the examined Most of previous WAFL methods require all the mea-
OCs deviated from OC of Table I, which is taken from [26]. surements to be synchronized [11], [13]–[15], [18], [19].
Simulation results, as reflected in Fig. 3, for faults along the Otherwise, the unsynchronized measurements have to be left
longest line of the 9-bus test system show that the method out and fault location should be carried out by the remaining
is successful in both high- and low-load profiles at the fault synchronized measurements. In the proposed method, in con-
instant. OC1 represents high-load profile, where the pre-fault trast, magnitudes of unsynchronized measurements can also
network is more stressed than the normal condition. OC2 contribute to fault location. To illustrate the advantage of
represents a large phase-angle difference at the terminals of the proposed method over previous WAFL methods, a DLG
line 5-6, before fault inception. The reason is the increase in fault at 10% of line 4-5 is studied using different available
the generation of G3 as well as load growth at L5. OC3, in measurements, as shown in Table III. The method presented
contrast, simulates low-loading condition for line 5-6. As can in [13], as the typical of WAFL based on synchronized voltage
be seen, the method locates the fault along the line for various measurements, and another WAFL method in [15], which

1949-3053 (c) 2016 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.2016.2601379, IEEE
Transactions on Smart Grid

TABLE IV
~
S ENSITIVITY A NALYSIS OF E STIMATION E RROR WITH I NEXACT
N ETWORK PARAMETERS (9-B US S YSTEM ) 4

FL estimation error (%)


Fault d
Method Error in Network Parameters

10

9
Type (pu) 5 3
2% 4% 6% 8%
~

G
[13] -1.515 -3.197 -4.821 -6.389 1
0.1 8
Proposed 0.151 0.187 0.219 0.248
SLG
[13] -1.267 -2.8 -4.279 -5.706

29
28
0.5

27
Proposed -0.812 -1.684 -2.647 -3.707 ~ 2
[13] -1.005 -2.344 -3.635 -4.880 6
C
0.9
Proposed -0.502 -1.874 -3.293 -4.751
[13] -1.606 -3.288 -4.911 -6.478
0.1
Proposed -0.385 -0.797 -1.195 -1.582

31
~
LL
[13] -1.442 -2.974 -4.451 -5.878 13 7 9
0.5 22
Proposed -0.437 -0.920 -1.401 -1.882
[13] -1.256 -2.595 -3.885 -5.129

G
0.9
39

Proposed -0.474 -1.024 -1.58 -2.1426


[13] -1.603 -3.284 -4.907 -6.474 10
32

0.1
Proposed -0.350 -0.731 -1.202 -1.462
DLG
[13] -1.532 -3.064 -4.541 -5.968
40

0.5
25

Proposed -0.501 -1.005 -1.512 -2.023 14


C

11
26

[13] -1.288 -2.627 -3.917 -5.162


0.9 ~
Proposed -0.612 -1.211 -1.820 -2.439 15
G

C
[13] -1.635 -3.317 -4.940 -6.507
G

~
0.1
Proposed -0.425 -0.869 -1.297 -1.710 20
3PH ~
[13] -1.480 -3.011 -4.489 -5.915 12
0.5
21

Proposed -0.421 -0.868 -1.306 -1.735


23

20
22

~
[13] -1.285 -2.623 -3.913 -5.158
0.9 16
Proposed -0.396 -0.822 -1.244 -1.663 19
~
18
avoids using external network parameters, are compared with
21
the proposed method. 17
Case 1 assumes that synchronized voltage measurements
are available at buses 4, 6 and 8. Case 2 considers GSL at
bus 6, and therefore, previous WAFL methods have to use only Fig. 4. Single-line diagram of the 22-bus test system.
two voltage synchrophasors at buses 4 and 8, the latter being
as can be seen in Case 3 of Table III. Sensitivity analysis of
far from the fault point. Case 3 demonstrates the advantage
fault-location estimation on this line, as reflected in Table IV,
of the proposed method in the face of GSL. Although phase-
reveals that the proposed method outperforms [13] for different
angle of voltage measurement at bus 6 is not available, its
fault types along this line, when network parameters vary by
magnitude is incorporated in the fault-location procedure by
2% to 8%.
the proposed formulation. Fault-location results in Table III
It is worth noting that there are many other WAFL methods
show that when network parameters are 100% accurate, all
proposed in the literature. In [8], only two voltage measure-
methods present accurate results for all measurement cases.
ments are utilized, and therefore the method is prone to GSL.
It should be noted that the proposed method in [15] requires
The proposed method in [9] involves estimating phase-angle
synchronized measurements at all border buses 4, 6 and 8, and
values of measurements, and hence needs more computational
is paralyzed by GSL at bus 6. As network parameters vary
effort. Bad data detection and identification may also become
in practice according to weather and loading conditions [27],
problematic, since voltage measurements are divided by each
simulations are repeated considering an additional situation
other. Comparison of fault-generated waveforms with those
of 5% error in network parameters. Fault-location accuracy
calculated and recorded offline is another approach pursued
deteriorates with error in network parameters. This inaccuracy
in [28]–[31]. This approach does not lend itself to large-scale
is pronounced when the unsynchronized measurement at bus 6
power systems, where faults of different types, locations and
is dismissed in Case 2. When voltage measurement at bus 6 is
resistance values have to be simulated and stored in a database.
not utilized due to GSL, the method in [15] is inapplicable as
before and the method in [13] leads to less accurate results
due to error in network parameters. Using unsynchronized C. Fault Location on 22-Bus Test System
voltage measurement at bus 6, the proposed method yields To examine the application of the proposed method in a
more accurate results than those of previous WAFL methods, large system, a 22-bus test system, whose single line diagram

1949-3053 (c) 2016 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.2016.2601379, IEEE
Transactions on Smart Grid

TABLE V TABLE VI
FAULT L OCATION ON L ONG T RANSMISSION LINES OF THE 22-B US T EST W IDE A REA FAULT L OCATION OF L INE 16-18 IN 22-B US T EST S YSTEM
S YSTEM
Network WAFL FL estimation error (%)
FL estimation error (%) Parameters Method No GSL GSL at PMU 19
Faulted Fault
Fault Distance (pu) No Error in [13] 0.016 -0.072
Line Type
0.1 0.3 0.5 0.7 0.9 Parameters Proposed 0.007 0.015
3PH 0.0135 0.0024 -0.0021 0.0061 0.0078 5% Error in [13] -0.014 11.249
LL 0.0099 -0.0027 0.0397 0.0034 0.0047 Parameters Proposed -0.068 0.395
16-18
SLG 0.0267 0.0166 0.0766 0.0395 0.1853
DLG 0.0073 0.0106 0.0117 0.0298 0.0355
3PH 0.0009 0.0037 0.0013 0.008 0.0052
at bus 19, there will be no measurement to give information on
LL -0.0096 -0.0049 -0.0038 0.0023 -0.0023 bus 18. This can also be figured out from Fig. 1(b), where fault
10-12
SLG -0.0562 -0.098 0.1122 -0.093 0.2418 current is expected to be supplied from both ends of the line.
DLG -0.0151 -0.006 0.0024 0.0077 -0.0236
When all parameters are exact this problem does not appear as
3PH 0.004 0.0085 0.0035 0.0074 -0.0042
LL -0.0055 -0.0004 -0.002 0.0014 -0.0123 the weak interconnection gives exact value of the fault current
11-12
SLG 0.0923 0.1014 0.1118 0.1046 0.0913 from bus 18. However, as shown in Table VI, when there is
DLG -0.0177 -0.0069 -0.0011 0.0117 0.0049
only 5% error in network parameters, fault current cannot be
3PH 0.0074 0.0085 0.0057 0.0099 0.0112
LL -0.0045 -0.0015 -0.002 0.001 0.0058 estimated accurately and hence fault location will have more
13-14
SLG 0.029 0.1309 0.1793 0.2126 0.3342 than 11% error if unsynchronized voltage measurements are
DLG -0.0186 -0.0076 -0.0024 0.0074 0.0122
not utilized, as is the case in previous WAFL methods such as
3PH 0.0024 -0.0006 0.0018 0.0031 0.0068
LL -0.0063 -0.0125 -0.0072 -0.0063 0.0031 [13]. It can be seen from Table VI that when there is no GSL,
3-9
SLG -0.0369 0.0066 -0.0589 -0.1137 -0.102 even with 5% error in network parameters both [13] and the
DLG -0.0417 -0.0458 -0.0377 -0.0245 -0.0284
proposed method show quite accurate results. It has already
3PH -0.0192 -0.0093 0.0157 0.0304 0.0219
LL -0.0217 -0.0093 0.0086 0.01 0.0056 been observed in [16], [35] that in large-scale networks,
17-21
SLG 0.3451 0.3432 0.4126 0.528 0.7342 more PMUs lead to more accurate fault-location estimation.
DLG 0.0074 0.0168 0.0415 0.0552 0.0619 However, the absence of synchronized measurements at one
3PH 0.0264 0.0527 0.0366 0.0101 -0.0115
LL 0.0133 0.0395 0.0276 -0.0021 -0.0213 end of the line makes previous WAFL methods significantly
12-17
SLG 0.0599 0.1238 0.1459 0.3769 0.3267 vulnerable to even a slight error in network parameters. The
DLG -0.0054 0.0405 0.0423 0.0187 0.0063 proposed method, in contrast, shows a good robustness against
this error in network parameters by using the magnitude of
is shown in Fig. 4, is adopted for simulations. Based on observ-
measured voltage at bus 19, i.e. |∆V∆V 19 |.
ability criteria [32]–[34], synchronized voltage measurements
Table VII reflects sensitivity analysis of fault location on
at buses 1, 5, 9, 12, 16 and 19 are assumed to be available.
this line, revealing that the proposed method outperforms the
Table V shows fault-location results for faults of different
previous method for different fault types along different points
types on different points of long transmission lines of the
of the line, when network parameters are not exact. The
network. It can be seen from Table V that fault-location
advantage of the proposed method, therefore, becomes clear
accuracy stays well beneath 1% for all the fault cases. The
when network parameters are inexact, as is the case in practice
reason is taking full capacitance of the line into account in
due to loading and weather conditions [27].
the proposed formulation, as depicted in Fig 1.
Table VIII shows the convergence process for a bolted SLG
It should be noted that many fault-location methods require fault at 10% of line 16-18. The initial values for |II + +
f | and I f
6
fault type identification and/or fault resistance estimation. are calculated by (18) and (20), respectively. This expedites
The proposed method, however, locates the faults, accurately, the iterative process as can be seen in Table VIII. A similar
irrespective of fault type and resistance. This can clearly be behavior is observed for all simulation cases, where at most
observed from Tables I and V. five iterations are carried out. The average elapsed time for
fault location is less than 10 ms, on a 2.1 GHz dual-core
D. Impact of GPS Signal Loss on 22-Bus Test System processor with 4 GB of RAM.
The influence of the fault resistance for different faults
It has been proved in [14] that if all of the measurements
along this line is reflected in Table IX. It is observed that
are taken from one of the terminals of the faulted line, fault
the proposed method is robust against fault resistance, thanks
location is impossible to carry out. This implies that when
to considering the full capacitance of the line.
the parallel path to the faulted line is a weak interconnection,
WAFL will be unsuccessful if no measurement is utilized from
the far end of the line. This situation can be experienced in 22- VIII. C ONCLUSION
Bus test system when there is a DLG fault at 10% of line 16- A novel method for wide-area fault location of transmission
18. The parallel path to this line is line 12-17; if GLS occurs lines by both synchronized and unsynchronized measurements

1949-3053 (c) 2016 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.2016.2601379, IEEE
Transactions on Smart Grid

TABLE VII
S ENSITIVITY A NALYSIS OF E STIMATION E RROR WITH I NEXACT results have shown that faults at different points of the network
N ETWORK PARAMETERS (22-B US S YSTEM ) can be located regardless of fault type. Even for transmission
lines not equipped with any synchronized or unsynchronized
FL estimation error (%)
Fault d measuring device, the estimation error is far below 1% for
Method Error in Network Parameters
Type (pu)
2% 4% 6% almost all faults.
[13] -4.744 -9.199 -13.389 It has been observed that WAFL methods using sparse syn-
0.1
Proposed -0.025 -0.037 -0.053
SLG chronized measurements can be significantly vulnerable when
[13] -3.585 -7.042 -10.28
0.5 there is a slight error in network parameters. The proposed
Proposed 0.367 0.734 1.081
[13] -2.479 -4.906 -7.163 method, however, integrates unsynchronized measurements,
0.9
Proposed 0.846 1.609 2.335
and hence, provides useful information from the part of
[13] -4.684 -9.143 -13.337
0.1 the network with no available synchronized measurements.
Proposed -0.016 -0.023 -0.036
LL
0.5
[13] -3.782 -7.234 -10.466 This has resulted in decreasing the error of fault-location
Proposed 0.422 0.825 1.206
estimation for a long transmission line from 11% in a previous
[13] -2.709 -5.130 -7.381
0.9 method, which used only synchronized measurements, to less
Proposed 0.854 1.665 2.437
0.1
[13] -4.673 -9.132 -13.325 than 0.5%, thanks to utilizing an additional unsynchronized
Proposed -0.016 -0.022 -0.034
DLG measurement in the proposed method.
[13] -3.717 -7.171 -10.405
0.5
Proposed 0.420 0.830 1.218
[13] -2.616 -5.041 -7.295 A PPENDIX A
0.9
Proposed 0.856 1.690 2.485
PARTIAL D ERIVATIVE OF M AGNITUDE F UNCTION
[13] -4.659 -9.119 -13.314
0.1
Proposed -0.015 -0.019 -0.03 Expanding Z (x) in terms of its real and imaginary parts
3PH
[13] -3.789 -7.241 -10.473
0.5
Proposed 0.437 0.855 1.250
yields
[13] -2.736 -5.156 -7.405 Z (x) = P (x) + jQ(x) (A.1)
0.9
Proposed 0.899 1.746 2.553
Derivating the magnitudes of functions at two sides of (A.1)
TABLE VIII leads to
C ONVERGENCE B EHAVIOR OF THE P ROPOSED M ETHOD
∂ Z (x)
= ∂ P 2 (x) + Q2 (x)
p
Iteration Estimated x Estimated Estimated ∂x ∂x
Number (pu) |II +
f | (pu)
6 I+f (Deg)
∂P (x) ∂Q(x) (A.2)
0 0.5 5.0968 -76.5
P (x) + Q(x)
1 0.0966728 5.0337 -78.34
= p ∂x ∂x
2 0.1002947 5.0398 -78.12 P 2 (x) + Q2 (x)
3 0.1002928 5.0396 -78.16
4 0.1002928 5.0396 -78.16 On the other hand, derivating both sides of (A.1) yields
Total Time 7.917 ms Z (x)
∂Z ∂P (x) ∂Q(x)
= +j (A.3)
TABLE IX
∂x ∂x ∂x
I NFLUENCE OF FAULT R ESISTANCE ON E STIMATION ACCURACY Getting the conjugate of (A.1) and multiplying it by (A.3)
leads to
FL estimation error (%)
Fault d  
Fault Resistance (Ω) ∂ZZ (x) ∂P (x) ∂Q(x)
Z (x)∗ =

Type (pu) +j P (x) −jQ(x)
0 50 100 200 ∂x
 ∂x ∂x

0.1 -0.0292 -0.019 -0.0279 -0.1222 ∂P (x) ∂Q(x)
SLG 0.5 -0.0884 -0.0476 -0.0015 0.071516 = P (x) + Q(x) (A.4)
0.9 -0.1786 -0.2 -0.2139 -0.22415  ∂x ∂x 
∂P (x) ∂Q(x)
0.1 -0.0098 -0.008 -0.0019 0.0054 −j Q(x) + P (x)
LL 0.5 0.0055 -0.0135 -0.0154 -0.0111 ∂x ∂x
0.9 -0.0036 -0.0328 -0.0395 -0.0376
Using (A.4) we can rewrite (A.2) as
0.1 -0.0164 -0.014 -0.041 -0.0045
 
DLG 0.5 -0.0039 -0.0097 0.0029 0.0283 Z (x)
∂Z ∗
0.9 -0.0124 -0.0243 -0.0161 -7.7e-5
Re Z (x)
∂ Z (x) ∂x
0.1 -0.0142 -0.0115 -0.0088 -0.0045 = (A.5)
3PH 0.5 0.0057 0.0014 0.0034 0.003 ∂x Z (x)
0.9 -0.0031 -0.0113 -0.015 -0.0255
A PPENDIX B
has been presented in this paper. Only positive-sequence PARTIAL D ERIVATIVE OF P HASE -A NGLE F UNCTION
voltage measurements are used in the method and therefore
Writing (A.1) in polar form results in
it is inherently immune against CT saturation as well as
unreliable zero-sequence parameters of the line. Simulation Z (x) = R(x)ej δ(x) (B.1)

1949-3053 (c) 2016 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.2016.2601379, IEEE
Transactions on Smart Grid

Derivating both sides of (B.1) results in [14] A. S. Dobakhshari and A. M. Ranjbar, “A novel method for fault location
of transmission lines by wide-area voltage measurements considering
Z (x)
∂Z ∂R(x) j δ(x) ∂δ(x) measurement errors,” IEEE Trans. Smart Grid, vol. 6, no. 2, pp. 874–
= e +j Z (x) (B.2)
∂x ∂x ∂x 884, 2015.
[15] M. K. Neyestanaki and A. M. Ranjbar, “An adaptive PMU-based wide
Therefore, the derivative of phase angle of Z (x) with respect area backup protection scheme for power transmission lines,” IEEE
to x can be written as Trans. Smart Grid, vol. 6, no. 3, pp. 1550–1559, 2015.
[16] A. S. Dobakhshari and A. M. Ranjbar, “A wide-area scheme for power
∂R(x) Z (x) system fault location incorporating bad data detection,” IEEE Trans.
∂ZZ (x)
− ∂x Power Del., vol. 30, no. 2, pp. 800–808, 2015.
∂δ(x) ∂x R(x)
= −j (B.3) [17] A. Esmaeilian and M. Kezunovic, “Fault location using sparse syn-
∂x Z (x) chrophasor measurement of electromechanical wave oscillations,” IEEE
Trans. Power Del., vol. PP, no. 99, pp. 1–1, 2016.
Recall from (B.1) that R(x) = Z (x) and hence substituting [18] G. Feng and A. Abur, “Fault location using wide-area measurements
(A.5) into (B.3) yields and sparse estimation,” IEEE Trans. Power Syst., vol. PP, no. 99, pp.
  1–8, 2015.
Z (x)
∂Z ∗ [19] M. Majidi, M. Etezadi-Amoli, and M. S. Fadali, “A sparse-data-driven
Re Z (x) Z (x) approach for fault location in transmission networks,” IEEE Trans. Smart
Z (x)
∂Z ∂x
− Grid, vol. PP, no. 99, pp. 1–9, 2015.
∂x Z (x) 2

∂ 6 Z (x) [20] M. M. Saha, J. Iżykowski, and E. Rosolowski, Fault location on power
= −j (B.4) networks. Springer, 2010.
∂x Z (x)
[21] W. Yao, Y. LIU, D. Zhou, Z. Pan, J. Zhao, M. Till, L. Zhu, L. Zhan,
Q. Tang, and Y. Liu, “Impact of gps signal loss and its mitigation in
power system synchronized measurement devices,” IEEE Transactions
on Smart Grid, vol. Early Access, 2016.
R EFERENCES [22] D. Novosel, D. G. Hart, E. Udren and J. Garitty, “Unsynchronized two-
terminal fault location estimation,” IEEE Trans. Power Del., vol. 11,
[1] J. Pfeifenberger, J. Chang, and J. Tsoukalis. Investment trends no. 1, pp. 130–138, 1996.
and fundamentals in U.S. transmission and electricity infrastructure. [23] A. Abur and A. G. Exposito, Power system state estimation: theory and
[Online]. Available: http://www.brattle.com/system/publications/pdfs/ implementation. CRC Press, 2004.
000/005/190/original/Investment Trends and Fundamentals in US [24] K. Kreutz-Delgado, The Complex Gradient Operator and the CR-
Transmission and Electricity Infrastructure.pdf Calculus. University of California, San Diego, 2009. [Online].
[2] M. Kezunovic, “Smart fault location for smart grids,” IEEE Trans. Smart Available: http://dsp.ucsd.edu/∼kreutz/Publications/kreutz2009complex.
Grid, vol. 2, no. 1, pp. 11–22, 2011. pdf
[3] M. Korkali, H. Lev-Ari, and A. Abur, “Traveling-wave-based fault- [25] H. Saadat, Power system analysis. McGraw-Hill:New York, 1999.
location technique for transmission grids via wide-area synchronized [26] Matpower: A matlab power system simulation package. [Online].
voltage measurements,” IEEE Trans. Power Syst., vol. 27, no. 2, pp. Available: http://www.pserc.cornell.edu/matpower/
1003–1011, 2012. [27] V. Terzija, Z. M. Radojevic, and G. Preston, “Flexible synchronized
[4] M. Korkali and A. Abur, “Optimal deployment of wide-area synchro- measurement technology-based fault locator,” IEEE Trans. Smart Grid,
nized measurements for fault-location observability,” IEEE Trans. Power vol. 6, no. 2, pp. 866–873, 2015.
Syst., vol. 28, no. 1, pp. 482–489, 2013. [28] A. Salehi Dobakhshari and A. Ranjbar, “A circuit approach to fault
[5] S. Azizi, S. Afsharnia, and M. Sanaye-Pasand, “Fault location on multi- diagnosis in power systems by wide area measurement system,” Int.
terminal DC systems using synchronized current measurements,” Int. Trans. Elect. Energy Syst., vol. 23, no. 8, pp. 1272–1288, 2013.
Jour. Elect. Power & Energy Syst., vol. 63, pp. 779–786, 2014. [29] Z. Galijasevic and A. Abur, “Fault location using voltage measurements,”
[6] S. Azizi, M. Sanaye-Pasand, M. Abedini, and A. Hassani, “A traveling- IEEE Trans. Power Del., vol. 17, no. 2, pp. 441–445, 2002.
wave-based methodology for wide-area fault location in multiterminal [30] M. Kezunovic and Y. Liao, “Fault location estimation based on matching
DC systems,” IEEE Trans. Power Del., vol. 29, no. 6, pp. 2552–2560, the simulated and recorded waveforms using genetic algorithms,” in 7th
2014. Int. Conf. Develop. Power Syst.Protect., 2001, pp. 399–402.
[7] Y. Chen, D. Liu, and B. Xu, “Wide-area traveling wave fault location [31] A. S. Dobakhshari and A. M. Ranjbar, “Transmission grid fault diagnosis
system based on IEC61850,” IEEE Trans. Smart Grid, vol. 4, no. 2, pp. by wide area measurement system,” in 3rd IEEE PES Innovative Smart
1207–1215, 2013. Grid Technologies Europe (ISGT Europe), 2012, pp. 1–7.
[8] Y. Liao, “Fault location for single-circuit line based on bus-impedance [32] N. M. Manousakis, G. N. Korres, and P. S. Georgilakis, “Taxonomy
matrix utilizing voltage measurements,” IEEE Trans. Power Del., vol. 23, of PMU placement methodologies,” IEEE Trans. Power Syst., vol. 27,
no. 2, pp. 609–617, 2008. no. 2, pp. 1070–1077, 2012.
[9] N. Kang and Y. Liao, “Double-circuit transmission-line fault location [33] S. Azizi, A. S. Dobakhshari, S. A. Nezam Sarmadi A. M. Ranjbar,
with the availability of limited voltage measurements,” IEEE Trans. “Optimal PMU placement by an equivalent linear formulation for
Power Del., vol. 27, no. 1, pp. 325–336, 2012. exhaustive search,” IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 174–
[10] ——, “Double-circuit transmission-line fault location utilizing synchro- 182, 2012.
nized current phasors,” IEEE Trans. Power Del., vol. 28, no. 2, pp. [34] S. Azizi, A. Salehi Dobakhshari, S. Nezam Sarmadi, A. Ranjbar, and
1040–1047, 2013. G. Gharehpetian, “Optimal multi-stage PMU placement in electric power
[11] Q. Jiang, X. Li, B. Wang, and H. Wang, “PMU-based fault location using systems using boolean algebra,” Int. Trans. Elect. Energy Syst., vol. 24,
voltage measurements in large transmission networks,” IEEE Trans. no. 4, pp. 562–577, 2014.
Power Del., vol. 27, no. 3, pp. 1644–1652, 2012. [35] S. Azizi, M. Sanaye-Pasand, and M. Paolone, “Locating faults on
[12] A. Salehi-Dobakhshari and A. M. Ranjbar, “Application of synchronised untransposed, meshed transmission networks using a limited number
phasor measurements to wide-area fault diagnosis and location,” IET of synchrophasor measurements,” IEEE Trans. Power Syst., vol. Early
Gen. Transm. Distrib., vol. 8, no. 4, pp. 716–729, 2014. Access, 2016.
[13] S. Azizi and M. Sanaye-Pasand, “A straightforward method for wide-
area fault location on transmission networks,” IEEE Trans. Power Del.,
vol. 30, no. 1, pp. 264–272, 2015.

1949-3053 (c) 2016 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.

S-ar putea să vă placă și