486
2, APRIL 2000
An Adaptive PMU Based Fault Detection/Location
Technique for Transmission Lines
Part I: Theory and Algorithms
Joe-Air Jiang, Jun-Zhe Yang, Ying-Hong Lin, Chih-Wen Liu, Member, IEEE, and Jih-Chen Ma
Abstract—An adaptive fault detection/location technique based
on Phasor Measurement Unit (PMU) for an EHV/UHV transmis-
sion line is presented in this two paper set. This paper is Part I of
this set. A fault detection/location index in terms of Clarke com-
ponents of the synchronized voltage and current phasors is de-
rived. The line parameter estimation algorithm is also developed to
solve the uncertainty of parameters caused by aging of transmis-
sion lines. This paper also proposes a new Discrete Fourier Trans-
form (DFT) based algorithm (termed as Smart Discrete Fourier
Transform, SDFT) to eliminate system noise and measurement er-
rors such that extremely accurate fundamental frequency compo-
nents can be extracted for calculation of fault detection/location
index. The EMTP was used to simulate a high voltage transmission
line with faults at various locations. To simulate errors involved
in measurements, Gaussian-type noise has been added to the raw
output data generated by EMTP. Results have shown that the new
DFT based method can extract exact phasors in the presence of
frequency deviation and harmonics. The parameter estimation al-
gorithm can also trace exact parameters very well. The accuracy of
both new DFT based method and parameter estimation algorithm
can achieve even up to 99.999% and 99.99% respectively, and will
be presented in Part II. The accuracy of fault location estimation
by the proposed technique can achieve even up to 99.9% in the per-
formance evaluation, which is also presented in Part II.
Index Terms—Fault Detection/Location Index, Discrete Fourier
Transforms (DFT), Phasor Measurement Unit (PMU).
I. INTRODUCTION
I N POWER systems, high voltage transmission and distri-
bution lines are vital links that achieve the essential con-
tinuity of service from the generating plants to the end users.
From views of economics and quality of power feeding, the im-
portance of fault detection/location of transmission lines is in-
creasingly. The facts are well illustrated in the history of elec-
trical power industry all of the world. The occurrences of fault in
transmission line often engender some severe economical losses
and social problems, such as producing a lots of toxic pollution
in an emergency shutdown of semiconductor production line,
creating a city commotion at night, etc. The more accuracy of
fault detection and location has been obtained the easier task for
inspection, maintenance, and repair of the line we can achieve.
For a long transmission line, for example, an error in fault loca-
tion of a few kilometers may still be acceptable for a relaying de-
cision but would represent a long and hard walk in rough terrain
Manuscript received February 5, 1999.
The authors are with the Department of Electrical Engineering, National
Taiwan University, Taipei, Taiwan.
Publisher Item Identifier S 0885-8977(00)03460-9.
0885–8977/00$10.00 © 2000 IEEE
IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO.
for repair. Rapid restoration of service could reduce customer
complaints, outage time, loss of revenue, and crew repair ex-
pense. All of these factors are increasingly importance to re-
searchers and utilities facing challenges in competitive electric
power industry.
Several investigations have examined the fault detection and
location in transmission lines. Among these investigations,
T. Takagi et al. [1], [2] applied the superposition principle
to estimate the fault location of transmission lines. The au-
thors proposed the Fourier transform and Laplace transform
methods which both are based on single-end measured data.
These approaches introduce several assumptions regarding
fault resistance and current ratios. The accuracy of the results
obtained depends on whether the assumptions are met or not.
Also, those algorithms that based on the single-end data will
seriously suffer from the errors due to the variations of source
impedance, fault incidence angle, line asymmetry, and loading
conditions. The only merit of the single-end methods is that
they need just one terminal data.
To improve the computation of fault location, many authors
suggested utilizing the information from both line ends. Present
communications technology allows for use of data from both
ends of transmission lines. To obtain the most accurate estima-
tion of fault location for transmission lines, a variety of such
two-ended data based algorithms have been presented, but only
a few of the many references will be included in the references.
M. Kezunovć et.al. [3] proposed a fault location algorithm based
on synchronized sampling technique. They adopted a time do-
main model of a transmission line as a basis for the algorithm
development. The accuracy of the proposed algorithm are well
within 1% error, however, the data must be acquired at a suffi-
ciently high sampling rate to provide adequate approximation of
the derivatives, which depend heavily on the selection of the line
model and the system itself. A. A. Girgis et.al. [4] proposed an-
other fault location algorithm based upon two-end data. While
Girgis’ paper considered the synchronization errors in sampling
the voltage and current waveforms at different line terminals,
the line models employed are lump-types, which can’t really
reflect the effect of inter-phase coupling phenomenon. There
are two-terminal fault location estimation approaches based on
unsynchronized data have also been proposed by D. Novosel
et. al. [5]. The authors utilized the lump-model to represent the
short line and made compensation for long lines. The described
schemes suit only off-line post-fault analysis.
It is worthy to note that numerical methods adopted in the
previous papers are either least-square estimation method or
JIANG et al.: AN ADAPTIVE PMU BASED FAULT DETECTION/LOCATION TECHNIQUE 487

Newton-Raphson iterative method, and such numerical tech-

niques possess some errors inherently. Moreover, those papers
described above haven’t considered the errors involved in mea-
surements, e.g., the nonlinear characteristics of CT and CCVT
itself, the setting errors of data-processing units, the synchro-
nization errors of timing clock in measurements, etc. A.T. Johns
et. al. [6], [7] proposed exact formulae for calculating fault lo-
cation based on using post-fault voltage and current at both line
ends. While the performance of the author’s simulation results
has been excellent, the proposed approach still does not consider
the synchronization configuration of phasor measurements and
are not adaptive in nature. The papers of D. J. Lawrence et. al.
[8], [9] discussed the errors involved in the measurement, the ef-
fect of parameter variation are on line model, and had field test
results. However, there are some errors in fault location estima-
tion since the authors did not utilize the parameter estimation
technique to resolve the uncertainty of line parameters due to the
factors such as temperature variation, asymmetry of lines, varia-
tion of power flow, etc. The performance of all algorithms men-
tioned so far gives satisfactory results in simulation, but have
not been addressed two main factors that affect the accuracy
of described approaches in detail yet. They are line parameter
uncertainty and accurate calculation of fundamental phasors in
the presence of harmonics, noise and frequency deviation from
Fig. 1. Configuration of adaptive PMU based fault detection/location system
nominal frequency. The way of solving such problems will be for EHV/UHV transmission line.
discussed in this paper.
Because of the need to clear faults on major transmission
Phasor Measurement Unit (PUM) techniques. The paper is
lines very rapidly, several protective schemes utilizing informa-
organized into five sections, the first of which is introduction.
tion from travelling waves launched by disturbances have been
In the section II an adaptive fault detection/location algorithm
suggested for a long period of time. The high frequency com-
based on PMU technique is presented. In the third section,
ponents in fault waveform present undesirable effects to most
we derive fault detection/location index for a single-phase
fault location algorithms and are essential for accurate fault de-
transmission line, which then is utilized as a prototype formula
tection and location [10], [11]. A fault detection/location al-
in three-phase case. The parameter estimation algorithm is
gorithm that derives from travelling wave principles can cope
described in section IV. The new DFT based method for funda-
with high frequency transients since these basically depend on
mental phasor calculation is also described in section IV. The
travelling wave phenomenon. The travelling wave schemes de-
design and implementation of phasor measurement unit (PMU)
pend on identifying particular voltage and current increments
and performance evaluation of the proposed algorithms will be
from all those which arrive at the relaying point. The identifi-
presented in the Part Il. Finally, the fifth is closing remarks.
cation of the desired signal becomes the essential problem of
the usual kind of travelling wave protection scheme. L. Jie et al.
II. AN ADAPTIVE PMU BASED FAULT DETECTION/LOCATION
[12] proposed an adaptive travelling wave protection algorithm
TECHNIQUE
using two cross-correlation functions to aid effectively in the
identification of travelling wave. Fault location methods uti- The overall diagram of the adaptive PMU based fault detec-
lizing travelling waves are independent of network configura- tion/location technique is shown in Fig. 1. The phasor measure-
tions and devices installed in the network. However, travelling ment units are installed at both ends of the transmission line.
wave algorithm is somewhat frequency dependent, unless auxil- The three phase voltages and three phase currents are measured
iary cross-correlation function is utilized [12]. Besides, it is not by PMU’s located at both ends of line simultaneously Since
easy to identify the local maximum of cross-correlation function the ‘Global Synchronism Clock Generator (GSCG)’ (this device
for calculating fault location. The other limitations of travelling will be discussed in Part II) has been equipped in PMU to pro-
wave techniques are that they require a very high sampling rate vide an extremely accurate and reliable external reference clock
and their implementations are more costly than implementations signal, it can guarantee sampling synchronization to an accuracy
of impedance techniques. of better than 1 -sec. The performance of the proposed PMU
To develop fault detection/location techniques that cope unit has been on-line demonstrated very well in the 161 kV sub-
with various variation factors associated with accuracy of stations of Taipower system. The proposed algorithms are tested
fault locators, an adaptive fault detection/location algorithm with the data generated from EMTP.
is presented in this paper. Our aim is to perform on-line the The EMTP simulated voltage and current waveforms were
fault detection and location for any transposed three-phase directly taken as the synchronized sampled data (voltages
line by employing the Global Positioning System (GPS) based and currents) from substations. The simulation examples are
488 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 2, APRIL 2000

propagation constant of the transmission line respectively. Con-

Fig. 2. A single-phase transmission line.
intentionally chosen as same with a practical 345 kV trans-
mission line of Taipower system. In order to truly include the
errors involved in the measurement procedure of phasors, the
Gaussian-type [13] noise signals were added into the EMTP
generated data before they are processed in the algorithm. Next
, the SDFT method is used to extract close-in fundamental
phasors. These phasors are then transformed by Clarke trans-
formation. The estimation of line parameters is proceeded in
the next step. At this stage, the technique has generated the
L)
correct modal components that can be utilized to calculate the
index. This technique possesses both abilities for detecting and
locating the fault of transmission line. It is worthy to note that
L(VS 0 ZC IS ) exp(0DL) (3)
the proposed technique is recursive. Hence, the proposed tech-
stants A and B can be determined by boundary conditions at
sending and receiving ends.
Suppose that a midway fault occurred at the point F which is
x = DL km away from receiving end R on a transmission line
SR shown in Fig. 3. L is the total length of the transmission
line, and D is the per unit distance from receiving end to the
fault and is also used as a fault detection/location index. In case
of a fault occurrence at the point F , the transmission line is thus
divided into two homogeneous parts. One is line section SF ,
the other is line section F R. These two line sections still can
be regarded as perfect transmission lines. This means that the
voltages at any point on the two line sections can be expressed in
terms of the voltages and currents measured on both ends S and
R. Moreover, at fault point F the voltages expressed in terms of
these two data sets (VS ; IS ) and (VR ; IR ) are identical. We can
visualize that there is an pre-fault transmission line as shown
in Fig. 2. If we put the two sets of measured data (VS ; IS ) and
(VR ; IR ) on the same reference position x = 0 and take into
account the boundary condition on both ends, then the voltage
on the fault position x = DL km can be expressed as:
VF
1 (V + Z I ) exp(DL)
= 2 exp(
S C S
+ 12 exp(

= VR +2ZC IR exp(DL) + VR 02ZC IR exp(0DL)

nique is very suitable for the on-line monitoring applications.
VF (4)
III. THE PROPOSED FAULT DETECTION/LOCATION INDEX Eq. (3) and (4) represent the post-fault voltages at fault point
Single-phase Case: which are expressed in terms of measured sending and re-
The index using the synchronized voltage and current sam- ceiving data (VS ; IS ) and (VR ; IR ), respectively. Since VF is
ples at both ends of a transmission line to calculate the location the voltage at the same point F , equating the Eq. (3) and (4)
and solving the fault detection/location index D yields:
of the fault is presented in this section.
Consider an un-faulted single-phase (two-conductors in free- D = ln(N=M )2 L (5)
space) transmission line shown in Fig. 2. Then both voltage and
current measured at a distance x km away from receiving end where
obey the partial differential equations: N = VR 02ZC IR 0 VS 02ZC IS exp( L) (6-1)
@v @i
= Ri + L @t and
= VS +2ZC IS exp(0 L) 0 VR +2ZC IR
(1-1)
@x
@i M (6-2)
@x
= Gv + C @v
@t
(1-2)
It is worthy to note that no assumptions in the procedure of
where R; L; G, and C are resistance, inductance, conductance derivation for the fault detection/location index D are made.
and capacitance of the transmission line per unit length, re- Hence, the index D is very robust. It will be clearly shown in
spectively. Under sinusoidal steady state condition, the oper- simulation results that the index D is hardly affected by the vari-
ator @=@t in Eq.(1) can be substituted by j! , and therefore ations of source impedance, loading change, fault impedance,
Eq.(1) can be converted into two linear, second-order, homoge- fault inception angle and fault type. Simulations have shown that
neous differential equations (original lowercase variable change this proposed index can provide extremely accurate fault loca-
to upper case ones which represent the phasor quantities), whose tion, for example, the error is well within 300 m (about span
solution are of two transmission line towers). Moreover, the index D incor-
porates with the computed absolute values M and N also can
= Ae+ x + Be0 x
V (2-1) be utilized as fault detector. In the case of single-phase line, the
+I x 0
= (Ae + Be )=ZC x (2-2) computed absolute values of M and N (Eq. (6)) will all be held
p
where ZC = (R + j!L)=(G + j!C ), whose units are
, is
at zero before the occurrence of a fault. This can be proved by
substituting the measured data (VS ; IS ) and (VR ; IR ), which
p
called the characteristic impedance (or surge impedance), and
= (R + j!L)=(G + j!C ), whose units are m01 , is called
are used to compute M and N in Eq. (6), into the formulae
of M and N . Since such measured components all satisfy the
JIANG et al.: AN ADAPTIVE PMU BASED FAULT DETECTION/LOCATION TECHNIQUE 489

Fig. 3. A single-phase faulted transmission line.

Fig. 4. A three-phase transposed transmission line.

transmission line equation, i.e. Eq. (2), after algebraic manipu-
lation, we can prove that M and N are indeed identical to zero,
and hence the algorithm will give an indefinite value of fault
location index through Equation (5). However, as soon as the
post-fault measured data been input into the algorithm through
As in the case of single-phase transmission lines, the voltages
and currents at a distance x km away from receiving end are
related through partial differential equations:
@v

@x
= RI + L
@i

@t
(7-1)
the moving data window (described in Part II), the computed
values of M and N abruptly deviated from zero, and hence the
@i

@x
= Gv + C
@v

@t
(7-2)
fault location index D also quickly converged to the interval of
(0, 1). These facts imply that M and N incorporated with loca- where both v and i are 3 2 1 vectors, R; L; G, and C are all
tion index D could be employed as indicators of various fault 3 2 3 transposed line parameters matrice whose forms are all
events. Such performance will be explained in the simulation like as:
section (see Part II). 2 3
Ls Lm Lm
Three-phase Case: L = 4 Lm Ls Lm 5 [H/km] (8)
Our main concern in this subsection is to extend the tech-
L m Lm Ls

niques of the previous subsection to the three-phase line. Up to Under sinusoidal steady-state condition, Eq. (7) can be changed
now we have proposed a fault detection/location approach for a into the form
single-phase two-wire transmission line. Unfortunately, the ap-
proach can’t be applied straightforwardly to a three-phase trans- @V
= ZI (9-1)
mission line due to the coupling effect of the inter-phase. @x

Consider an arbitrary three-phase transmission line illustrated @I

@x
= YV (9-2)
in Fig. 4, which consists of phase a; b; c and a ground system
consisting of earth and ground wires. Both ends of transmission In order to de-couple phase quantities, a suitable transformation,
line of interest are labeled by S and R. The system shown in referred to as the modal (or called Clarke) transformation, is
Fig. 4 is divided into two parts. The part of transmission line given as following:
is plotted in three-wire form for emphasizing the transposition 2 3 2 3
configuration of the line, the other parts of sources are shown Va V0
in one-line diagram for simplicity. The quantities on both ends 4 Vb 5 = T 4 V 5 (10-1)
shown in Fig. 4 are all vectors of phase voltages and currents. V c V 
490 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 2, APRIL 2000

2 3 2 3
Ia I0 TABLE I
4 Ib 5 = T 4 I 5 (10-2) D
A LIST OF CORRECTNESS OF FAULT LOCATION INDEX i FOR DIFFERENT
Ic I FAULT TYPES UNDER TRANSPOSED LINE CASE.

where 0,  and  represent the Clarke components of the volt-

ages and currents, and Clarke transformation matrix is chosen
herein as:
2 3
1 1 2p p 0p 5
T = p 41 01=p2 p3=p2 (11)
3 1 01= 2 0 3= 2
Substitute Eq. (10) and Eq. (11) into Eq. (9), after some alge-
braic manipulations, Eq. (9) can then be solved as:
Vm = exp(0x)A + exp(00x)B (12-1) if we utilize Clarke transformation matrix to de-couple the
0
Im = exp(0x)ZC A 0 exp(00x)ZC B (12-2)
1 0 1 phase components of signals into the modal waves, not all
modal indices D , in the formulae derived above could provide
where Vm and Im are both 3 2 1 vectors whose entries are accurate performance of fault location and fault detection. In
modal components of signals, and subscript in represents 0,  fact, it would launch different modal waves depending upon
and  -mode of the transmission line respectively. 0 and ZC are various fault types. When the fault is an a-phase earth fault,
modal propagation constant matrix and modal surge impedance for example, there is no  -modal wave on the line, and thus
of the line respectively, whose forms are as follows: just only two modal index D0 and D can give the accurate
p results. On the contrary, the three modal index D0 , D , and
0 = pT 01 ZY T (13-1)
D all can be utilized to provide the extremely accurate results
ZC = T ZY T01 01 (13-2) in the event that the fault can launch 0,  and  -modal waves
Note that 0 and ZC both are 3 2 3 diagonal matrices. Since we simultaneously. In Table I, a list is summarized to indicate
have assumed that the transmission line is transposed ones, the the conditions under which the modal indices Di , i = O , ,
commutative properties of matrix multiplication have been em-
ployed in the above procedure of derivation. As same treatment
with single-phase case, we could still take the voltages at fault
point F which locate at x = DL km away from receiving and
x = 0 as boundary condition, the fault location index are there-
fore solved as
lnf[A(i) 0 C (i)]=[E (i) 0 B (i)]g
 , can yield the accurate fault detection/location
p results on
a three-phase transposed lines. The ‘ ’ mark represents the
modal index Di can give accurate results, but the ‘x’ mark
represents the modal index can not give the correct results.
It should be noted that, although the characteristic impedance
and propagation constant of the three modes are determined by
the line parameter matrices of Eq. (13), the definitions of the
Di = i  0; ;  three modes derived by Clarke transformation are not unique.
20(i; i)L Other possibilities exist, for example, we also can get another
(14)
where A(i); B (i), C (i) and E (i) are the entries of 3 2 1 vec-
common alternative definition by utilizing Karrenbauer trans-
tors A; B; C , and E respectively, and i = 1; 2; 3 are utilized to
formation [14]. Therefore, it would provide another form of
represent 0,  and  -modal components of signals. 0(i; i) rep-
Table I.
resents diagonal entries of the 3 2 3 modal propagation constant
We have only considered the ideal transposed lines so far,
however, the transmission lines usually are un-transposed due
matrix. Those signals quantities utilized in the above equation
to effects of parameters aging and asymmetry of the lines. With
can be expressed as following:
the advent of synchronized phasor measurement techniques, we
1
A(i) = [VRm (i) + ZC (i)IRm (i)] (15-1) can estimate the line parameters with very high accuracy, and
2 therefore can find out a suitable transformation matrix by the
1
B (i) = [VRm (i) 0 ZC (i)IRm (i)] i  0; ;  (15-2) perturbation method (see [15]) to solve the problem mentioned
2 above.
1
C (i) = exp[00(i; i)L][VSm (i) + ZC (i)ISm (i)] (15-3)
2
1 IV. ON-LINE PARAMETER ESTIMATION ALGORITHM AND THE
E (i) = exp[00(i; i)L][VSm (i) 0 ZC (i)ISm (i)] (15-4) NEW DFT BASED ALGORITHM
2
where ZC (i; i) represents the diagonal entries of 3 2 3 modal On-line Parameter Estimation Algorithm
surge impedance matrix, and L is the total length of the trans- Almost all of the existing fault detection/location techniques
mission line. It should be noted that the quantities utilized in did not consider uncertainty about the line parameters, partic-
Eq. (15) could also be pure-fault data. Therefore, the effects of ularly zero-mode impedance. It can be seen from our simula-
line loading on the accuracy the index can be eliminated. tion results (see Fig. 13 in the companion paper) that if there
is a 620% error in line parameters can result in significant er-
Mode Selection rors in fault location estimation (even up to 6.6% error approx-
Each fault launches a set of 0,  and  -modal waves imately under b-phase ground fault). This fact has also been
propagating on the lines. Under the case of transposed lines, demonstrated by several papers [6]–[9], [16]. With the advent
JIANG et al.: AN ADAPTIVE PMU BASED FAULT DETECTION/LOCATION TECHNIQUE 491

of synchronized phasor measurement technique, one can ob- Then x (t) can be expressed as
tain extremely accurate phasors. Consider the three-phase trans-
xej!t + x3 e0j!t
x(t) =
posed transmission line shown in Fig. 4, using the technique
proposed, we can on-line monitor the status of transmission 2 (22)

line, and extract the true phasor measurement data VS ; IS ( ) where 3 denotes complex conjugate. Moreover, the fundamental
( )
and VR ; IR from both line ends by the new DFT filtering ap- ()
frequency (60Hz) component of DFT of fx k g is given by
proach. These data can be utilized to derive the characteristic
impedance and propagation constant of transmission line in- x^r =
X
2 N 01 x(k + r)e0j 2Nk (23)
versely. Such line parameters thus can be respectively expressed N k=0
as following:
Combing Eq. (22) and Eq. (23) and taking frequency deviation
V 2 0 VRm2 ( = 2 (60 + 1 ))
!  f into consideration, at last, we obtain:
ZCm = Sm
2 2 m = 0; ; 
ISm 0 IRm
(16)
N1
x sin 2 j 60N (1f (2r+N 01)+120r)
and x^ = e
N sin 1
m = cosh01(Km )=L m = 0; ;  (17) 2
N2
x3 sin 2 0j 60N (1f (2r+N 01)+120(r+N 01))
where
VSm ISm + VRm IRm + N 2 e
Km = m = 0; ;  (24)
VSM IRm + VRm ISm sin
(18)
2
The entire quantities mentioned above all are the modal com-
1f
21f , and 2 = 0 2(2+ 60 ) . If we define Ar , and
ponents of measured signals obtained by Clarke transforma-
tion, where subscript m denote 0, ;  -mode, respectively. The where 1 = 60N N
subscripts ‘S ’ and ‘R’ represent the data of sending and re- Br as
ceiving ends respectively. It should be noted that in order to N1
x sin 2 j 60N (1f (2r+N 01)1 20r)
Ar =
avoid zero-mode effects and uncertainties in zero-mode parame-
e (25)
ters, the zero-mode calculation is not recommended. It has been N sin 1
verified in field-test that the proposed on-line parameter esti-
mation technique is extremely accurate for high voltage distri-
2
N2
x3 sin 2 0j 60N (1f (2r+N 01)+120(r+N 01))
bution lines. Such a parameter estimation technique can also
Br =
be suitable for EHV/UHV transposed transmission case. The
e (26)
performance has been demonstrated by EMTP and will be dis- N sin 2
cussed in the Part II in detail. 2
Then Eq. (24) can be expressed as
The New DFT Based Algorithm
In order to achieve a high degree of accuracy in the location of x^r Ar + Br (27)
a fault, it is vitally important to be able to accurately extract the
Actually, the first half development of the algorithm of SDFT
phasor quantities of fundamental frequency from the measured
is the same as the conventional DFT method. So the SDFT
waveform. A new digital algorithm based on Discrete Fourier
can keep all advantages of DFT such as recursive computing
Transform, terms as Smart Discrete Fourier Transform (SDFT),
manner. In the conventional DFT, it assumes that the frequency
is presented to meet the purpose. The SDFT not only keeps all of
the advantages of DFT but also smartly take frequency deviation
^
deviation where is small enough to be ignored, and xr  Ar .
Therefore,
from nominal frequency and harmonics into consideration. This
section presents the algorithm of the SDFT. r = arctan(imag(^xr )=real(^xr )) (28)
Consider a sinusoidal signal of frequency ! =2 f given by:
x(t) = X cos(!t + ) (19)
r 0 r01
f = 60 +
where 2 2 60N (29)
X : the amplitude of the voltage/current signal, Conventional DFT methods incur errors in estimating fre-
: the phase angle of the voltage/current signal
()
Suppose that x t is sampled with a sampling rate (603N ) Hz quency and phasor when frequency deviates from nominal fre-
quency (60 Hz). However, in the SDFT we take Br into consid-
()
waveform to produce the sample set fx k g eration. So we define
k
x(k) = X cos(!
60N + ); k = 0; 1; 1 1 1 ; N 0 1 (20) a = ej ( 60N (21f +120))

(30)
The signal x(t) is conventionally represented by a phasor (a And from Eq. (25) and Eq. (26), we will find the following re-
complex number) x lations
x = Xej = X cos  + jX sin  (21) Ar+1 = A3r a (31)
492 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 2, APRIL 2000

Br+1 = Br3a01 (32) 2(m 0 1 + m601f ) 2(0m 0 1 + m601f )

Then 3 = N
and 4 = N
:

^
xr+1 = Ar+1 + Br+1 = A3r a + Br3 a01 (33) There are five unknown variables in Eq. (41), hence, we
need five equations to solve this problem. So using xr ; xr+1 ,
xr+2 ; xr+3 and xr+4 , we can get exact frequency and phasor
^
xr+2 = Ar+2 + Br+2 = A3r+1a + Br3+1 a01 = A3r a2 + Br3 a02 in presence of one harmonic. We use SDFT, denote that rd 3
(34) harmonic has been taken into consideration to distinguish from
There are three unknown variables in Eq. (27), Eq. (33) and basic SDFT. Of course, any other integral order harmonic still
Eq. (34), and after some algebraic manipulations we obtain: can be taken into consideration too, for example: SDFT35 and
x3r+1 a2
^ 0 (^xr + x^r+2 )3 a + x^r+1 = 0 (35)
3 5 3 5 7
SDFT357 take rd , th harmonic and rd , th , th harmonic
into consideration, respectively. Similarly, nonintegral har-
monics also can be developed. We use SDFTn to denote that
Solve Eq. (35) to obtain that
q
(^xr + x^r+2 ) 6 (^xr + x^r+2)2 0 4^x2r+1
nonintegral harmonics has been taken into consideration.
a=
2^xr+1 :
V. CLOSING REMARKS
Then from the definition of ‘a’ in (30), we can get the exact In this paper, we propose an adaptive fault detection/location
solution of the frequency. technique. Specifically, a fault detection/location index in terms
f = 60 + 1f = cos01(Re(a))3 602N (36)
of Clarke components of synchronized phasor measurements is
derived. The parameter estimation algorithm and SDFT method
Moreover, we can compute phasor after getting exact ‘f ’ by the have also been developed in this paper. With the advent of these
following equations: algorithms the proposed technique can achieve an excellent per-

= x^r+1 ^r
3 a0x
formance for detecting and locating the fault of transmission
Ar
a 01
line. In this study, we have run a great deal of ENM simulations.
2 (37)
1f
These simulation data are utilized to evaluate the performance
sin( 60N )
N3 of the proposed algorithm under the various fault resistances,
X = abs(Ar )3
1f )
sin( 60
(38) un-synchronization errors, source-impedance variations, system
frequency fluctuations, and line-parameters aging effects. These
simulation results will be presented in the companion paper.
 = angle(Ar ) 0
60N 2 (1f 2 (N 0 1))

(39)
It is observed that SDFT can provide exact phasor and frequency REFERENCES
^ ^ ^
using xr ; xr+1 and xr+2 in the presence of frequency deviation. [1] T. Takagi, Y. Yamakoshi, J. Baba, K. Uemura, and T. Sakaguchi, “A
New Alogrithm of an Accurate Fault Location for EHV/UHV Transmis-
Next, we take harmonics into consideration. Assume a sinu-
soidal signal of frequency ! =2
f with mth harmonic given
sion Lines: Part I- Fourier Transformation Method,” IEEE Transactions
on Power Apparatus and Systems, vol. PAS-100, no. 3, pp. 1316–1323,
by: March 1981.
[2] T. Takagi, Y. Yamakoshi, J. Baba, K. Uemuraand, and T. Sakaguchi, “A
( ) = X1 cos(!t1) + X2 cos(m!t + 2)
x t (40) New Algorithm of an Accurate Fault Location for EHV/UHV Transmis-
sion Lines: Part Il- Laplace Transform Method,” IEEE Transactions on
Power Apparatus and Systems, vol. PAS-101, no. 3, pp. 564–573, March
where X1 ; X2 : the amplitude, 1 ; 2 : the phase angle. Fol- 1982.
lowing similar steps developed previously, we can get: [3] M. Kezunović, J. Mrkić, and B. Perunicić, “ An Accurate Fault Loca-

^ = Ar + Br + Cr + Dr
tion Algorithm Using Synchronized Sampling,” Electric Power Systems
xr (41) Research, vol. 29, pp. 161–169, 1994.
[4] A. A. Girgis, D. G. Hart, and W. L. Peterson, “A New Fault Location
Technique For Two-and Three-Terminal Lines,” IEEE Transactions on

^ = Ar+1 + Br+1 + Cr+1 + Dr+1

Power Delivery, vol. 7, no. 1, pp. 98–107, January 1992.
xr+1 [5] D. Novosel, D. G. Hart, E. Udren, and J. Garitty, “Unsynchronized Two-

= A3r a + Br3 a01 + Cr3am + Dr3a0m (42)

Terminal Fault Location Estimation,” IEEE Transactions on Power De-
livery, vol. 11, no. 1, pp. 130–137, January 1996.
[6] A.T. Johns and S. Jamali, “Accurate Fault Location Technique For
where Power Transmission Lines,” IEE Proceedings, pt. C, vol. 137, no. 6,

sin N23 0j  (m1f (2r+N01)+60(2mr+mN0m0N +1))

pp. 395–402, 1990.
x2 [7] P.K. Aggarwal, D.V. Coury, A.T. Johns, and A. Kalam, “A Practical
Cr = e 60N Approach To Accurate Fault Location on Extra High Voltage Teed
3
sin 2
N Feeders,” IEEE Transactions on Power Delivery, vol. 8, no. 3, pp.
874–883, July 1992.
[8] D. J. Lawrence, L. Z. Cabeza, and L. T. Hochberg, “Development of
(43) an Advanced Transmission Line Fault Location System, Part I: Input
N 4
3 sin Transducer Analysis and Requirements,” IEEE Transactions on Power
Dr = xN2 2 e0j 60N (1f (2mr+N01)+60(2mr+mN0m +n01))
 Delivery, vol. 7, no. 4, pp. 1963–1971, October 1992.
4
sin 2
[9] D. J. Lawrence, L. Z. Cabeza, and L. T. Hochberg, “Development of an
Advanced Transmission Line Fault Location System, Part Il: Algorithm
Development and Simulation,” IEEE Transactions on Power Delivery,
(44) vol. 7, no. 4, pp. 1972–1981, October 1992.
JIANG et al.: AN ADAPTIVE PMU BASED FAULT DETECTION/LOCATION TECHNIQUE
[10] A.K.S. Chaudhary, Kwa-Sur Tam, and Anm G. Phadke, “Protection
System Representation in The Electromagnetic Transients Program,”
IEEE Transactions on Power Delivery, vol. 9, no. 2, pp. 700–711, April
1994.
[11] Bao Lian and M. M. A. Salama, “An overview of digital Fault Location
Algorithms for power transmission lines using transient waveforms,”
Electric Power Systems Research, vol. 29, pp. 17–25, 1994.
[12] L. Jie, S. Elangovan, and J. B. X. Devotta, “Adaptive Travelling
Wave Protection Algorithm Using Two Correlation Functions,”,
PE481-PAIRD-0-04-1998.
[13] A. G. Phadke, J. S. Thorp, and K. J. Karimi, “State Estimation with
Phasor Measurements,” IEEE Transactions on Power Systems, vol.
PWRS-1, no. 1, pp. 233–241, February 1986.
[14] A.G. Phadke and J.S. Thorp, Computer Relaying For Power Systems:
John Wiley & Sons, 1988.
[15] J. A. Brandão Faria and J. H. Briceño Mendez, “Modal Analysis of Un-
transposed Bilateral Three-Phase Lines— A Perterbation Approach,”
IEEE Transactions on Power Delivery, vol. 12, no. 1, pp. 497–504, Oc-
tober 1997.
[16] Power Engineering Education Committee and the Power System Re-
laying Committee of the IEEE Power Engineering Society, , “Advance-
ments in Microprocessor Based Protection and Communication,”, IEEE
Tutorial Course, 1997.
Joe-Air Jiang was born in Tainei, Taiwan, in 1963. He graduated from National
Taipei University of Technology in 1983 and received M.S. degree in electrical
engineering from National Taiwan University, Taipei, Taiwan in 1990. Since
1990, he has been with Private Kuan-Wu Institute of Technology and Com-
merce, where he is instructor of electrical engineering. He is currently working
on a Ph.D research project in power system protection and control, also at Na-
tional Taiwan University. His area of interest is in computer relaying and bio-ef-
fects of EM-wave.
493
Jun-Zhe Yang was born in Tainan, Taiwan, in 1971. He received his B.S. degree
in electrical engineering from Tatung Institute of Technology in 1992 and M.S.
degree from National Taiwan University in 1995. He is presently a graduate
student in the electrical engineering department of National Taiwan University,
Taipei, Taiwan.
Ying-Hong Lin was born in Taipei, Taiwan, in 1970. He received his B.S. degree
in electrical engineering from Taiwan University of Technology in 1995 and
is presently working on M.S. degree from National Taiwan University, Taipei,
Taiwan. His interested researches are the application of GPS and PMU in power
system.
Chih-Wen Liu was born in Taiwan in 1964. He received his B.S. degree in
electrical engineering from National Taiwan University in 1987, and M.S. and
Ph.D. degrees in electrical engineering from Cornell University in 1992 and
1994. Since 1994, he has been with National Taiwan University, where he is
associate professor of electrical engineering. He is a member of the IEEE and
serves as a reviewer for IEEE Transactions on Circuits and Systems, Part I. His
main research area is in application of computer technology to power system
monitoring, operation, protection and control. His other research interests in-
clude GPS time transfer and chaotic dynamics and their application to system
problems.
Jih-Chen Ma was born in Chekiang, China in 1928. He received the B.S. degree
in Electrical Engineering from National Taiwan University in 1951. In 1961,
he joined the faculty. In 1963-65, being awarded the Alexander-von-Humboldt
Fellowship, (did his research in Max Planck Institut für Ionosphärenphysik,
Göttingen, W. Germany.) In 1969 as full professor, 1971-1974, as Department
Chairman. (In 1978, as President of CIEE. In 1985, he was awarded the Centen-
nial Memorial Medal from IEEE Society) Now retired as Professor Emeritus of
the University. His current interest is bio-electromagnetism.