Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

www.ietdl.

org
Published in IET Generation, Transmission & Distribution
Received on 18th November 2010
Revised on 25th April 2011
doi: 10.1049/iet-gtd.2011.0081

ISSN 1751-8687

Phasor-measurement-unit-based transmission line


fault location estimator under dynamic conditions
Z.Y. He1 R.K. Mai1,2 W. He1 Q.Q. Qian1
1
Department of Electrical Engineering, Southwest Jiaotong University, Sichuan Province, 610031,
People’s Republic of China
2
East Power Co., Ltd, Guangdong Province, 610031, People’s Republic of China
E-mail: hezy@swjtu.cn

Abstract: When power system is under dynamic condition such as power oscillation, the fault voltage and current cannot be expressed
as a standard sinusoidal function, so error, in the form of oscillation of estimate, will arise as conventional fault location algorithms
ignore the dynamic characteristics of supplied signals. A novel algorithm extended from conventional fault location algorithms is
proposed in this study. This approach not only can express the spatial characteristic of supplied signals but also considers them as
time-variant signals whose magnitude and frequency are changing against time. Therefore it possesses the ability of describing
spatial property and time property of signals, simultaneously. By using the adjoining phasor estimate from synchrophasor phasor
measurement units (PMUs) to express the dynamic characteristics in terms of derivatives, an accurate estimate of fault location
can be attained via Newton iteration method. The performance of the proposed algorithm is evaluated by computer Power System
Computer Aided Design/Electromagnetic Transient including DC (PSCAD/EMTDC)-generated signals under various situations.
The evaluation results indicate that the proposed technique yields an accurate fault location estimate under dynamic conditions
independent of fault types, fault impedance and fault locations. The maximal estimate error of the proposed algorithm is less than
one-tenth of that of the conventional fault location algorithm in most cases.

Nomenclature Although travelling-wave methods have been proven to


have many excellent characteristics such as immunity to
Some notations used in this paper are shown as follows: current transformer saturation, insensitiveness to the
R series resistance per unit length parameters of power system and fault conditions [1, 2], it
has some undeniable limitation like requirement of high
L series inductance per unit length sampling rate, discrimination between travelling wave from
G shunt conductance per unit length the fault or reflected one from remote end. The choice of
C shunt capacitance per unit length mother wavelet and the presence of noise need to be
l length of transmission line considered with care as well.
Intelligent computational techniques involve knowledge
Re(A) real component of A from human for modelling of complex systems, are capable
Im(A) imaginary component of A of handling uncertainties and non-linearities of power
systems. However, performance of fault location algorithms
based on these techniques depends severely on the amount
1 Introduction of training samples [7, 8, 11], which are often not available
in practice.
Fault location techniques on transmission lines are of Fundamental impedance-based techniques mainly employ
importance for power utilities to speed up line restoration, the fundamental frequency phasor of voltage and current
reduce the outage time and prevent power systems from signal to form a set of formulae, and then utilise some
entering unstable state. Various algorithms have been curve-fitting methods like least-square method or Newton
developed in the past years to address fault location of iteration to calculate the fault location. The principle of this
transmission lines. The fault location methods can be method is proposed in [12, 13], which attains fault location
classified into three fundamental categories: (i) travelling- with voltage and current phasors from both ends and
wave method [1 – 6] based on the analysis of fault- presumes that the phasors are completely synchronised.
originated travelling wave; (ii) intelligent computational However, phasors from both ends sometimes are
techniques [7 – 11] based on modelling of qualitative asynchronous because of different sampling rates or phase
knowledge of human and (iii) fundamental impedance- shift caused by the difference of transformers and phasor
based techniques. measurement unit (PMU) devices, therefore some

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191 1183
doi: 10.1049/iet-gtd.2011.0081 & The Institution of Engineering and Technology 2011
www.ietdl.org
approaches are presented to get fault location by solving a where the constants AU and AI are the magnitudes of voltage
non-linear equation. In [14], the authors proposed a new and current signals, respectively. v is the angular velocity.
approach to alleviate the effect of current-transformer Substituting (3) and (4) into (1) and (2) yields
saturations by utilising voltage phasors from both ends.
As power systems are often operated close to the stability ∂U
boundary for economic purposes, some operations or severe = IR + jvLI (5)
∂x
disturbances may lead power systems into rapid power
oscillation conditions on a long transmission line [15 – 17]. ∂I
= GU + jvCU (6)
The frequency or/and magnitude of voltage and current ∂x
signals are changing relatively fast during this dynamic
process. The methods mentioned above, however, do not Consider a single-phase transmission line between terminals S
appropriately consider these dynamic characteristics of and R as shown in Fig. 2, where UR and IR are the voltage and
power systems, which significantly affect the accuracy of current phasor measurements from terminal R, whereas US
fault localisation. and IS are the voltage and current phasor measurements
This paper presents a new fault location algorithm. This from terminal S. As long as the voltage phasor UR and
method is able to express the dynamic characteristics of current phasor IR are available, it is possible to obtain the
power systems in spatial span as well as in time span, and general solution of (5) and (6) for voltage and current at a
gives accurate estimates of fault location under different distance x from the terminal R as shown in (7) and (8)
dynamic conditions. The rest of this paper is organised as
follows: Section 3 presents a fault location estimator UR − IR ZC gx UR + IR ZC −gx
U= e + e (7)
considering dynamic characteristics of power systems; 2 2
Section 4 presents some simulation results with PSCAD/ UR /ZC − IR gx UR /ZC + IR −gx
EMTDC-generated signals based on a faulted transmission I= e − e (8)
line. Conclusions are drawn in Section 5. 2 2
√
where ZC = Z/Y , the characteristic impedance per unit
√
2 Proposed fault location estimator length and g = ZY , the propagation constant per unit
considering dynamic characteristics of length.
power systems Suppose that a midway fault occurred at the point F which
is x km away from terminal R on a transmission line as shown
2.1 Single-phase case in Fig. 2. With the knowledge of phasor measurements of
voltage and current signals from both terminals measured
2.1.1 Static-fault-location algorithm: Fig. 1 depicts by PMUs located at the transmission line, the phasors of
voltage and current relationship among single-phase voltage at the fault location derived from both terminals are
transmission line in terms of the distributed line parameter given by
at an infinitesimal distance (∂x). Based on Fig. 1, the
differential voltage (∂U ) and differential current (∂I ) across UR − IR ZC gx UR + IR ZC −gx
the elemental length at a distance x from terminal R are URF = e + e (9)
2 2
derived [18]
US − IS ZC g(l−x) US + IS ZC −g(l−x)
USF = e + e (10)
∂U ∂I 2 2
= IR + L (1)
∂x ∂t
where URF and USF are the voltage estimates at point F,
∂I ∂U respectively.
= UG + C (2) The fault location is determined by (11) as the estimates of
∂x ∂t
voltage at the fault point derived from both terminals are
A few fault-location algorithms assume that magnitude and identical
fundamental frequency of voltage and current signals are  
constant within a short period −1 (UR + IR Zc )/2 − (US − IS ZC )egl /2
x = (2g) ln (11)
(US + IS Zc )e−gl /2 − (UR − IR ZC )/2
U = AU ejvt (3)
2.1.2 Dynamic fault location estimator: Given accurate
I = AI ejvt (4) parameters of transmission line and the phasor measurement

Fig. 1 Voltage and current relationship of a distributed parameter


line Fig. 2 Single-phase transmission line with fault at F

1184 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2011.0081
www.ietdl.org
from PMU, we can have a good fault-location estimate under impedance and equivalent shunt admittance to estimate the
the condition that the frequency and magnitude of voltage and fault location, so that estimate errors arise.
current signals are constant. However, they may change quite With the assumption that the dynamic characteristics
rapidly under fast power oscillation condition. That will lead among the transmission line from one terminal to the fault
to some error in the fault-location estimate because of the location are the same as that at the measurement point
dynamic characteristics. Therefore an extended model is xm [ {0, l}, we can express the dynamic characteristics at
needed to express the dynamic characteristics by current report time t1 ¼ 0 within a short period via two
considering the changing of magnitude and frequency phasor measurements at adjoining report time as shown in
against time. The phasors of voltage and current signals are (18) and (19)
expressed by the products of a band-limited phasor and a
constantly rotating phasor ejv0 t as shown in (12) and (13) ∂[i(x, 0)] i(xm , t) − i(xm , − t)
= (18)
∂t 2t
U (x, t) = u(x, t)ejv0 t (12) ∂[u(x, 0)] u(xm , t) − u(xm , − t)
= (19)
∂t 2t
I(x, t) = i(x, t)ejv0 t (13)
where t is the report time interval of phasor measurement.
where u(x, t) and i(x, t) are the band-limited phasors of Fig. 3 shows the relationship between report time and band-
voltage and current signals, respectively; v0 is angular limited phasors during the report time interval t.
velocity of the constantly rotating phasor. Substituting (12), (13), (18) and (19) into (16) and (17)
By substituting (12) and (13) into (1) and (2), we have yields the equivalent series impedance and equivalent shunt
admittance derived from terminal R(xm ¼ 0)
∂U ∂I
= IR + L
∂x ∂t Z̃ R = R + jv0 L
jv i(x, t)ejv0 t ∂t + ejv0 t ∂[i(x, t)] L IR (0, t)e−jv0 t − IR (0, − t)ejv0 t
= IR + L 0 + (20)
∂t IR (0, 0) 2t
∂[i(x, t)]
= IR + jv0 LI + L i(x, t)ejv0 t
i(x, t)∂t Ỹ R = G + jv0 C
 
L ∂[i(x, t)] C UR (0, t)e−jv0 t − UR (0, − t)ejv0 t
= I R + jv0 L + + (21)
i(x, t) ∂t UR (0, 0) 2t
= I Z̃ (14) The voltage phasor at fault location derived from terminal R is
∂I ∂U given by
= UG + C
∂x ∂t UR − IR Z̃ RC g̃R x UR + IR Z̃ RC −g̃R x
URF = e + e (22)
∂t + e ∂[u(x, t)]
jv0 u(x, t)e jv0 t jv0 t 2 
 2

= UG + C
∂t where Z̃ RC = Z̃ R /Ỹ R and g̃R = Z̃ R Ỹ R .
∂[u(x, t)] Similarly, the voltage phasor at fault location derived from
= UG + jv0 CU + C u(x, t)ejv0 t
u(x, t)∂t terminal S(xm ¼ l ) is given by
 
C ∂[u(x, t)]
= U G + jv 0 C + US − IS Z̃ SC g̃S (l−x) US + IS Z̃ SC −g̃S (l−x)
u(x, t) ∂t USF = e + e (23)
2 2
= U Ỹ (15)  
where Z̃ SC = Z̃ S /Ỹ S , g̃S = Z̃ S Ỹ S and
where the equivalent series impedance and equivalent shunt
admittance are Z̃ S = R + jv0 L
L IS (l, t)e−jv0 t − IS (l, − t)ejv0 t
Z̃ = R + jv0 L +
L ∂[i(x, t)]
(16) + (24)
i(x, t) ∂t IS (l, 0) 2t

C ∂[u(x, t)] Ỹ S = G + jv0 C


Ỹ = G + jv0 C + (17)
u(x, t) ∂t C US (l, t)e−jv0 t − US (l, − t)ejv0 t
+ (25)
US (l, 0) 2t
In static condition, the magnitude and fundamental frequency
of voltage and current remain constant. In other words,
derivatives of the band-limited phasor [u(x, t) and i(x, t)]
are equal to zero. In that situation, this dynamic model will
degrade to be the same as the static one. However, the
conventional fault location estimators completely ignore the
dynamic characteristics and consider (∂[i(x, t)])/∂t and
(∂[u(x, t)])/∂t to be zero even when they are under dynamic
condition like power oscillation. The conventional fault Fig. 3 Report time and band-limited phasors relationship during
location estimators apply inaccurate equivalent series report time interval t

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191 1185
doi: 10.1049/iet-gtd.2011.0081 & The Institution of Engineering and Technology 2011
www.ietdl.org
In ideal condition, the voltage phasors of fault location Up to now, sequence components decoupled from three-
derived from both terminals should be identical and the phase components are available. What needs to do is to use
fault location is determined by sequence parameters of transmission lines and to replace
phase components (UR , IR and US , IS) with sequence
URF = USF (26) components (UR1 , IR1 and US1 , IS1). UR1 and IR1 are
positive components of voltage and current from terminal
Since it is hard to solve the non-linear equation directly, R, whereas US1 and IS1 are positive components of voltage
Newton iteration is employed and (26) is rewritten as and current from terminal S.
The process to attain the estimate of fault location is given
Fdis (x) = USF − URF below and the maximum number of iterations is set to 5.

US − IS Z̃ SC g̃S (l−x) US + IS Z̃ SC −g̃S (l−x) Step 1: Obtain the positive sequence components of current
= e + e and voltage signals by (31) and (32) after getting the phasor
2 2
  estimates from PMUs.
UR − IR Z̃ RC g̃R x UR + IR Z̃ RC −g̃R x Step 2: Obtain the initial fault-location estimate (x0) with the
− e + e (27)
2 2 assumption that no dynamic characteristics exist. In other
words, we have Z̃ = Z̃ S1 = Z̃ R1 and g̃ = g̃S1 = g̃R1 . Then,
The first-order derivative of Fdis(x) with respect to x can be x0 can be obtained through (11).
determined as follows Step 3: After substituting x0 into (27) and (28), Fdis(x0) and
(∂Fdis(x0))/∂x are given.
∂Fdis (x) U − IS Z̃ SC g̃S (l−x) U + IS Z̃ SC −g̃S (l−x) Step 4: Employ Fdis(x0) and (∂Fdis(x0))/∂x to calculate the next
= −g̃S S e + g̃S S e estimate of fault location by
∂x 2 2  
  ∂Fdis (x0 )
U − IR Z̃ RC g̃R x U + IR Z̃ RC −g̃R x x1 = x0 − Fdis (x0 )/ (33)
− g̃R R e − g̃R R e ∂x
2 2
(28) Step 5: Check the difference between x0 and x1 . If the
difference is less than the specified tolerance or the number
When the initial value of x is available, the fault location of iterations is greater than maximum iteration time, the
estimate is attained by procedure of Newton iteration as absolute value of x1 is output as the fault-location estimate,
described in the next section. or else set x0 ¼ x1 and go back to Step 3.

2.2 Three-phase case 2.3 On-line parameter estimator


A single-phase model for fault-location estimate is The parameters of transmission lines may change because of
established in the last subsection. In practice, three-phase aging and changing of surrounding environment. Therefore
line and double-circuit transmission line are commonly used the true value of parameters will deviate from the nominal
so that the coupling effect of the inter-phase is introduced. ones. Consider that a power system is staying in static
The three-phase model cannot be applied to estimate the condition or a relatively slow changing condition, the
fault location before decoupling. Taking three-phase line for dynamic characteristics are ignorable. Therefore it is
instance, symmetrical component method [19] is applied to possible to obtain an accurate estimate of parameters of
eliminate the coupling effect of the inter-phase as follows transmission lines by using the phasor measurements of
⎡ ⎤ ⎡ ⎤⎡ ⎤ both terminals from PMUs as shown in (34) and (35)
U1
1 1 a a2 Ua
⎣ U2 ⎦ = ⎣ 1 a2 a ⎦⎣ Ub ⎦ (29) UR1 − IR1 ZC1 g1 x UR1 + IR1 ZC1 −g1 x
3 US1 = e + e (34)
U0 1 1 1 Uc 2 2
⎡ ⎤ ⎡ ⎤⎡ ⎤ UR1 /ZC1 − IR1 g1 x UR1 /ZC1 + IR1 −g1 x
I1 1 a a2 Ia IS1 = e − e (35)
1
⎣ I2 ⎦ = ⎣ 1 a2 a ⎦⎣ Ib ⎦ 2 2
(30)
3
I0 1 1 1 Ic where ZC1 and g1 are the positive characteristic impedance
√ and propagation constant, respectively.
where a = −(1/2) + j( 3/2) is a phase rotation operator. After solving (34) and (35), ZC1 and g1 are given
Ua , Ub , Uc and Ia , Ib , Ic represent three-phase components

of the voltages and currents, respectively; U1 , U2 , U0 and 2
US1 − UR1
2
I1 , I2 , I0 denote sequence components of the voltages and ZC1 = 2 − I2
(36)
currents. No matter what kind of fault happens in IS1 R1
transmission lines, positive components exist. Therefore  
U + ZC1 IS1
only the positive components of voltages and currents are g1 = l −1 ln S1 (37)
employed to estimate the fault location in this paper UR1 − ZC1 IR1
 √  √ Series impedance and shunt admittance are given by
1 1 3 1 3
U1 = Ua + − + j Ub + − − j Uc (31)
3 2 2 2 2 g1
 √  √ Y1 = (38)
ZC1
1 1 3 1 3
I1 = Ia + − + j Ib + − − j I (32)
3 2 2 2 2 c Z1 = ZC1 · g1 (39)

1186 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2011.0081
www.ietdl.org
The series resistance, series inductance, shunt conductance are 10% of total length of transmission line away from the
and shunt capacitance are then obtained as follows next one. A-phase voltage and current signals under the Ag
fault (fault inception angle: 08; fault resistance: 20 V; fault
R1 = Re(Z1 ) (40) distance: D4) are shown in Fig. 5. It is clear that there are
oscillations before and after occurrence of the fault. The
Im(Z1 ) fault-location estimate of SFLA and DFLE is given in
L1 = (41)
v0 Fig. 6. At the early beginning of fault occurrence, the
errors of both algorithms are relatively large because of the
G1 = Re(Y1 ) (42) inaccuracy of phasor measurements from PMUs as the data
Im(Y1 ) window of Discrete Fourier Transform contains both the
C1 = (43) pre-fault data and post-fault data. Even though the fault-
v0 location estimate of both algorithms appears oscillating
where v0 is given by PMUs’ estimate.

3 Performance evaluation
In order to verify the performance of the proposed algorithm,
PSCAD/EMTDC-generated signals are utilised. In all the
cases, the sampling frequency is set to 2400 Hz (i.e. 48
samples per cycle in a 50 Hz system) and one cycle
rectangular window is applied. We employed the Dynamic
Phasor Estimation Algorithm [20] to estimate the phasors of
voltage and current signals. The report rate is set to 50 Hz.
A simulation model of a typical 230 kV transmission
system was set up (Fig. 4). The model system consists of an
800-MVA generator supplying power to an infinite bus. The
occurrence of a three-phase short circuit occurs at F for a
period of 100 ms near source N creates a disturbance,
which forces a power system to appear power oscillation.
Then, different types of fault, different fault resistances,
various fault locations and inception angles of the fault
occurrence were applied during the following power
oscillation Steady Fault Location Algorithm (SFLA) is
employed as a comparison algorithm to evaluate the
performance of Dynamic Fault Location Estimator (DFLE).
In this study, the percentage error used to indicate the
performance of algorithms is calculated as follows
x − xTheoretical
Eest = × 100% (44)
l
where xTheoretical is the theoretical value of fault location.

3.1 Cases with different fault location and


fault resistances
Fig. 5 A-phase voltage signal under an Ag fault (fault inception
An Ag fault was applied at t ¼ 0 ms on Circuit 2 with various angle: 08; fault resistance: 20 V; fault distance: D4)
fault resistances (0 V, 5 V, 20 V, 50 V) and two different a Voltage signal
angles (0 and 908) at different locations (D0 – D9), which b Current signal

Fig. 6 Fault location estimate of two algorithms under an Ag fault


Fig. 4 Model power system and its parameters (fault inception angle: 08; fault resistance: 20 V; fault distance: D4)

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191 1187
doi: 10.1049/iet-gtd.2011.0081 & The Institution of Engineering and Technology 2011
www.ietdl.org
after a short period, the estimate of SFLA deviates is applied as shown in Tables 1 – 3. The largest average of
considerably from the theoretical fault location in the form estimate errors of the proposed algorithm stays within
of oscillation compared to DFLE because of the dynamic 0. 1769%, that of SFLA is up to 0.5405%. It is also found
characteristics of voltage and current signals. The that the estimation errors of both algorithms under different
convergence time of DFLE is about 71 ms as shown in fault inception angles are nearly the same. Besides, some
the amplified figure of Fig. 6. The error estimate of DFLE simulations such as faults near to buses are done, and the
is close to zero but some small oscillation exists in the simulation results with fault within 5% of the line away
estimate because of Taylor error of the signal model. Of from bus M are shown in Table 4.
course, the higher order the signal model is of, the more As performance of both algorithms has little to do with
accuracy of estimate can be achieved. However, more fault inception angle, only the maximum error of estimate
computation power is needed and it will take longer to under Ag fault (fault inception angle: 908; fault resistance:
converge to the theoretical value as more phasors are used. 0 – 50 V; fault distance: 0.0 – 1.0 pu) is shown in Fig. 7. It
Tables 1 – 3 summarise the mean errors of estimate of two can be seen that the performance of DFLE is much better
algorithms under Ag fault (fault resistance: 0 –50 V; fault than that of SFLA. The maximum estimate error of DFLE
distance: 0.0– 1.0 pu) for three different fault inception can be kept within 0.41544% whereas that of SFLA
angles (0 and 908), respectively. The estimates between increases up to 1.33789%. It is clear that the max errors of
time span from 50 to 1000 ms are considered. As expected, both algorithms increase as the fault resistance increases as
DFLE gives relatively small estimate error than SFLA as mentioned before.
DFLE has taken full consideration of dynamic The mean error of estimate of two algorithms under
characteristics than SFLA. When a small fault resistance is Ag fault (fault inception angle: 08; fault distance: D4) for
applied to a transmission line, the dynamic characteristics high resistive phase to ground fault (50 – 2000 V) is shown
caused by the oscillation of two end generators become as in Fig. 8. It is clear that the performance of DFLE is
smaller than that with a large one. Therefore both much better than that of SFLA and the mean errors of both
algorithms give better estimates when small fault resistance algorithms increase as the fault resistance increases.

Table 1 Mean errors of estimate under Ag fault with fault inception angle: 08; fault resistance: 0–50 V; fault distance: 0.0–1.0 pu

Fault distance, pu Eest , %

SFLA, V DFLE, V

0 5 20 50 0 5 20 50

0.0 0.0001 0.0166 0.0649 0.1622 0.0002 0.0151 0.0599 0.1497


0.1 0.0457 0.0998 0.1581 0.2170 0.0030 0.0103 0.0485 0.1204
0.2 0.0816 0.1383 0.2506 0.3069 0.0048 0.0050 0.0358 0.0905
0.3 0.1073 0.1658 0.3160 0.3909 0.0068 0.0049 0.0196 0.0590
0.4 0.1214 0.1821 0.3575 0.4543 0.0096 0.0076 0.0103 0.0243
0.5 0.1247 0.1854 0.3826 0.5021 0.0141 0.0136 0.0187 0.0220
0.6 0.1196 0.1797 0.3959 0.5288 0.0209 0.0255 0.0333 0.0538
0.7 0.1055 0.1651 0.3989 0.5375 0.0286 0.0349 0.0527 0.0904
0.8 0.0971 0.1438 0.3893 0.5273 0.0403 0.0485 0.0739 0.1275
0.9 0.1379 0.1331 0.3724 0.4913 0.0560 0.0610 0.0914 0.1577
1.0 0.1910 0.1735 0.3420 0.4305 0.0681 0.0790 0.1028 0.1765

Table 2 Mean errors of estimate under Ag fault with fault inception angle: 458; fault resistance: 0–50 V; fault distance: 0.0– 1.0 pu

Fault distance, pu Eest , %

SFLA, V DFLE, V

0 5 20 50 0 5 20 50

0.0 0.0001 0.0166 0.0648 0.1623 0.0002 0.0152 0.0600 0.1498


0.1 0.0457 0.1001 0.1580 0.2163 0.0029 0.0103 0.0487 0.1203
0.2 0.0814 0.1391 0.2526 0.3091 0.0049 0.0056 0.0355 0.0918
0.3 0.1064 0.1668 0.3182 0.3920 0.0070 0.0046 0.0203 0.0578
0.4 0.1197 0.1823 0.3604 0.4580 0.0103 0.0082 0.0092 0.0243
0.5 0.1217 0.1866 0.3854 0.5045 0.0144 0.0149 0.0181 0.0224
0.6 0.1129 0.1807 0.3986 0.5328 0.0209 0.0244 0.0338 0.0558
0.7 0.0954 0.1653 0.4001 0.5405 0.0292 0.0368 0.0539 0.0907
0.8 0.0784 0.1428 0.3930 0.5302 0.0407 0.0483 0.0734 0.1279
0.9 0.1040 0.1287 0.3744 0.4951 0.0565 0.0619 0.0915 0.1586
1.0 0.2201 0.1526 0.3434 0.4329 0.0852 0.0790 0.1027 0.1769

1188 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2011.0081
www.ietdl.org
Table 3 Mean errors of estimate under Ag fault with fault inception angle: 908; fault resistance: 0–50 V; fault distance: 0.0– 1.0 pu

Fault distance, pu Eest , %

SFLA, V DFLE, V

0 5 20 50 0 5 20 50

0.0 0.0001 0.0164 0.0652 0.1622 0.0002 0.0150 0.0603 0.1497


0.1 0.0455 0.1002 0.1590 0.2170 0.0029 0.0102 0.0489 0.1204
0.2 0.0809 0.1395 0.2532 0.3069 0.0052 0.0051 0.0351 0.0905
0.3 0.1057 0.1674 0.3185 0.3909 0.0070 0.0054 0.0200 0.0590
0.4 0.1189 0.1838 0.3607 0.4543 0.0096 0.0077 0.0097 0.0243
0.5 0.1209 0.1875 0.3875 0.5021 0.0146 0.0150 0.0173 0.0220
0.6 0.1122 0.1821 0.4007 0.5288 0.0204 0.0244 0.0333 0.0538
0.7 0.0971 0.1670 0.4017 0.5375 0.0291 0.0360 0.0546 0.0904
0.8 0.0822 0.1456 0.3948 0.5273 0.0411 0.0477 0.0740 0.1275
0.9 0.1117 0.1342 0.3761 0.4913 0.0571 0.0605 0.0911 0.1577
1.0 0.2161 0.1643 0.3452 0.4305 0.0854 0.0790 0.1020 0.1765

Table 4 Mean errors of estimate under Ag fault with fault inception angle: 08; fault resistance: 0–50 V; fault distance: 0.0–0.05 pu

Fault distance, pu Eest , %

SFLA, V DFLE, V

0 5 20 50 0 5 20 50

0.0 0.0001 0.0166 0.0649 0.1622 0.0002 0.0151 0.0599 0.1497


0.01 0.0026 0.0194 0.0681 0.1650 0.0003 0.0139 0.0556 0.1451
0.02 0.0090 0.0242 0.0729 0.1686 0.0005 0.0132 0.0543 0.1418
0.03 0.0103 0.0299 0.0814 0.1728 0.0008 0.0126 0.0527 0.1374
0.04 0.0158 0.0351 0.0922 0.1782 0.0013 0.0118 0.0513 0.1338
0.05 0.0261 0.0403 0.1016 0.1855 0.0019 0.0115 0.0505 0.1307

Fig. 8 Mean errors of estimate of two algorithms under Ag fault


with fault inception angle: 08; fault resistance: 50– 2000 V; fault
Fig. 7 Maximum errors of estimate of two algorithms under Ag
distance: D4
fault with fault inception angle: 908; fault resistance: 0– 50 V;
fault distance: 0.0– 1.0 pu

estimate errors of both algorithms under the same fault


3.2 Cases with different fault type and fault type increase as the fault resistance increases. The
resistances maximum estimate error of DFLE is less than one-tenth
of that SFLA except an Ag fault with 0 V fault resistance.
Table 5 shows the maximum and mean estimate errors of The reason why some cases like AB, ABg and ABCg
both algorithms under different fault types (fault inception faults with 0 and/or 5 V fault resistance are not listed in
angle: 08; fault resistance: 0 –50 V; fault distance: D4). Table 5 is that these faults create vital disturbance to the
The estimate errors under different fault types have similar power system so that the power system becomes unstable
patterns as given in Table 5. The maximum and mean soon after fault.

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191 1189
doi: 10.1049/iet-gtd.2011.0081 & The Institution of Engineering and Technology 2011
www.ietdl.org
Table 5 Mean and maximum errors of estimate under different fault types with fault inception angle: 08; fault resistance: 0–50 V; fault
distance: D4

Fault type Fault resistance, V Eest , %

SFLA DFLE

Maximum error Average error Maximum error Average error

Ag 0 0.2070 0.1214 0.0319 0.0096


5 0.3073 0.1821 0.0218 0.0076
20 0.6570 0.3575 0.0298 0.0103
50 1.0812 0.4543 0.0525 0.0243
AB 20 0.1268 0.0661 0.0163 0.0058
50 0.3732 0.1862 0.0241 0.0092
ABg 20 0.1850 0.1035 0.0124 0.0047
50 0.3587 0.1819 0.0223 0.0110
ABCg 50 0.3401 0.1068 0.0327 0.0101

Table 6 Mean errors of estimate under different noise density with fault type: a– g; fault inception angle: 08; fault resistance: 50 V; fault
distance: 0.0– 1.0 pu

Fault distance, pu Eest , %

SFLA, dB DFLE, dB

40 50 60 40 50 60

0.0 0.17149 0.16381 0.15990 0.15803 0.15100 0.14851


0.1 0.24801 0.22277 0.21798 0.15925 0.13026 0.12233
0.2 0.30185 0.30624 0.30741 0.09947 0.08988 0.08932
0.3 0.39033 0.39852 0.39327 0.11370 0.05946 0.05669
0.4 0.46400 0.45581 0.45356 0.10401 0.03428 0.02471
0.5 0.50716 0.50417 0.50075 0.09293 0.03236 0.02366
0.6 0.54504 0.52514 0.52985 0.09380 0.06556 0.05286
0.7 0.56697 0.53653 0.54239 0.14057 0.09903 0.09090
0.8 0.51737 0.52014 0.52549 0.17246 0.12597 0.12624
0.9 0.51334 0.49414 0.48603 0.18654 0.15931 0.15695
1.0 0.40980 0.43662 0.43220 0.21288 0.17019 0.18224

3.3 Cases with different noise density the situations listed in Table 6, although the accuracy of
both algorithms decreases as signal noise rate increases.
Table 6 shows the mean estimate errors of both algorithms The impact of noise on the error of estimate of SFLA is
under different noise density (fault inception angle: 08; fault limited compared to that of DFLE. SFLA is mainly affected
resistance: 0 – 50 V; fault distance: D4). It is clear that the by dynamic characteristics whereas DFLE has considerably
performance of DFLE is superior to that SFLA under all removed the impact of dynamic characteristics, therefore the

Fig. 9 Mean errors of estimate of two algorithms under different Fig. 10 Maximum errors of estimate of two algorithms under
fault types (fault inception angle: 08; fault resistance: 50 V) different fault types (fault inception angle: 08; V)

1190 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2011.0081
www.ietdl.org
estimate error of DFLE increases from 60 dB (0.02471%) to 3 Borghetti, A., Bosetti, M., Di Silvestro, M., Nucci, C.A., Paolone, M.:
40 dB (0.10401%) when the fault distance is D4. However, ‘Continuous-wavelet transform for fault location in distribution power
networks: definition of mother wavelets inferred from fault originated
estimate error of SFLA has limited increase compared to transients’, IEEE Trans. Power Syst., 2008, 23, pp. 380–388
the error caused by dynamic characteristics. 4 Elkalashy, N.I., Lehtonen, M., Darwish, H.A., Taalab, A.M.I., Izzularab,
M.A.: ‘DWT-based detection and transient power direction-based
3.4 Cases with different fault type location of high-impedance faults due to leaning trees in unearthed
MV networks’, IEEE Trans. Power Deliv., 2008, 23, pp. 94– 101
5 Silva, M.D., Coury, D.V., Oleskovicz, M., Segatto, E.C.: ‘Combined
Figs. 9 and 10 show the mean estimate errors of both solution for fault location in three-terminal lines based on wavelet
algorithms under different fault types (fault inception angle: transforms’, IET Gener. Transm. Distrib., 2010, 4, pp. 94–103
08; fault resistance: 50 V). Although the estimate errors of 6 Reddy, M.J., Mohanta, D.K.: ‘Adaptive-neuro-fuzzy inference system
different fault types varies, but the performance of DFLE is approach for transmission line fault classification and location
much better than that of SFLA as DFLE fully takes the incorporating effects of power swings’, IET Gener. Transm. Distrib.,
2008, 2, pp. 235 –244
dynamic characteristics of a power system into account. 7 Reddy, M., Mohanta, D.K.: ‘Performance evaluation of an adaptive-
network-based fuzzy inference system approach for location of faults
4 Conclusion on transmission lines using Monte Carlo simulation’, IEEE Trans.
Fuzzy Syst., 2008, 16, pp. 909–919
8 Thukaram, D., Khincha, H.P., Vijaynarasimha, H.P.: ‘Artificial neural
A novel algorithm DFLE to estimate the fault location under network and support vector machine approach for locating faults in
dynamic conditions has been proposed in this paper. Not only radial distribution systems’, IEEE Trans. Power Deliv., 2005, 20,
the effect of spatial domain (distributed capacitor of pp. 710–721
transmission line) was taken into account but also the effect 9 Souza, J.C.S., Rodrigues, M.A.P., Schilling, M.T., Do Coutto Filho,
of time domain (the change of fundamental phasors of M.B.: ‘Fault location in electrical power systems using
intelligent systems techniques’, IEEE Trans. Power Deliv., 2001, 16,
voltage and current) was fully considered by DFLE. In pp. 59–67
addition, the parameters of transmission line are estimated 10 Seung-Jae, L., Myeon-Song, C., Sang-Hee, K., et al.: ‘An intelligent and
online to confront with the impact of aging and surrounding efficient fault location and diagnosis scheme for radial distribution
temperature. The algorithm gives better fault localisation systems’, IEEE Trans. Power Deliv., 2004, 19, pp. 524– 532
11 Mora-Florez, J., Barrera-Nuez, V., Carrillo-Caicedo, G.: ‘Fault location
estimate under dynamic condition with comparison to in power distribution systems using a learning algorithm for
conventional fault location algorithms. multivariable data analysis’, IEEE Trans. Power Deliv., 2007, 22,
The performance of the proposed algorithm has been pp. 1715– 1721
evaluated with diverse faults on a 230 kV transmission line 12 Joe-Air, J., Jun-Zhe, Y., Ying-Hong, L., Chih-Wen, L., Jih-Chen, M.:
system. PSCAD/EMTDC was used to generate fault signals ‘An adaptive PMU based fault detection/location technique for
transmission lines. I. Theory and algorithms’, IEEE Trans. Power
for different fault locations, fault types, fault inception Deliv., 2000, 15, pp. 486– 493
angles and fault resistances. The evaluation results have 13 Joe-Air, J., Ying-Hong, L., Jun-Zhe, Y., Tong-Ming, T., Chih-Wen, L.:
shown that the proposed algorithm performs better than the ‘An adaptive PMU based fault detection/location technique for
conventional algorithm. The algorithm considering both transmission lines. II. PMU implementation and performance
space domain and time domain is useful for fault location evaluation’, IEEE Trans. Power Deliv., 2000, 15, pp. 1136–1146
14 Zamora, I., Minambres, J.F., Mazon, A.J., Alvarez-Isasi, R., Lazaro, J.:
under dynamic condition like power oscillation. ‘Fault location on two-terminal transmission lines based on voltages’,
IEE Proc. Gener. Transm. Distrib., 1996, 143, pp. 1– 6
5 Acknowledgments 15 Mechraoui, A., Thomas, D.W.P.: ‘A new principle for high resistance
earth fault detection during fast power swings for distance protection’,
IEEE Trans. Power Deliv., 1997, 12, pp. 1452–1457
This work was supported in part by the Chinese National 16 Karegar, H.K., Mohamedi, B.: ‘A new method for fault detection during
Science Fund: No. 50877068 in part by the Fundamental power swing in distance protection’. Proc. 2009 Sixth Int. Conf. on
Research Funds for the Central Universities: No. Electrical Engineering/Electronics, Computer, Telecommunications
SWJTU112T07 and in part by the Fundamental Research and Information Technology, ECTI-CON 2009, 2009, pp. 230–233
Funds for the Central Universities: No. SWJTU11CX124. 17 Xiangning, L., Zhengtian, L., Shuohao, K., Yan, G.: ‘Theoretical
fundamentals and implementation of novel self-adaptive distance
protection resistant to power swings’, IEEE Trans. Power Deliv.,
6 References 2010, 25, pp. 1372– 1383
18 Kundur, P., Balu, N.J., Lauby, M.G.: ‘Power system stability and
1 Bo, Z.Q., Jiang, F., Chen, Z., Dong, X.Z., Weller, G., Redfern, M.A.: control’ (McGraw-Hill, New York, 1994)
‘Transient based protection for power transmission systems’. Proc. 19 Blackburn, J.L.: ‘Symmetrical components for power systems
2000 Power Engineering Society Winter Meeting, 2000, pp. 1832– 1837 engineering’ (CRC, 1993)
2 Bo, Z.Q., Weller, G., Lomas, T., Redfern, M.A.: ‘Positional protection 20 Mai, R.K., He, Z.Y., Fu, L., He, W., Bo, Z.Q.: ‘Dynamic phasor and
of transmission systems using Global Positioning System’, IEEE frequency estimator for phasor measurement units’, IET Gener.
Trans. Power Deliv., 2000, 15, pp. 1163–1168 Transm. Distrib., 2010, 4, (1), pp. 73–83

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 11, pp. 1183–1191 1191
doi: 10.1049/iet-gtd.2011.0081 & The Institution of Engineering and Technology 2011

You might also like