You are on page 1of 14

1356 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO.

3, JUNE 2022

An Improved Noniterative Parameter-Free Fault


Location Method on Untransposed Transmission
Lines Using Multi-Section Models
Dian Lu , Student Member, IEEE, Yu Liu , Senior Member, IEEE, Shi Chen,
Binglin Wang , Student Member, IEEE, and Dayou Lu , Student Member, IEEE

Abstract—This paper proposes an improved noniterative fault the electricity recovery process, and therefore improves the
location method on untransposed transmission lines without uti- power supply reliability of the power system. A large number
lizing line parameters. Two-end three phase voltage and current of researchers proposed extensive fault location techniques on
phasor measurements before and during the fault are typically
required. First, the parameter-free fault location problem is for- transmission lines in the literatures [1]–[5]. Existing transmis-
mulated through multi-section transmission line models, and the sion line fault location methods can be mainly classified into fun-
necessary condition of noniterative solutions is carefully investi- damental frequency phasor based methods, traveling wave based
gated to determine the maximum possible section number of the methods, artificial intelligence based methods, etc. Specifically,
multi-section line model. Afterwards, with the determined section the traveling wave based methods locate faults by detecting
number, the analytical solution of the fault location is derived with
full utilization of the inherent characteristics of the line parameter the arrival time of wavefronts at line terminals. However, the
matrices. Instead of solving all the parameters, two key variables traveling wave intensity generated by the fault could be too
inside the parameter matrices are extracted and the fault location low for reliable detection of wavefronts if the fault inception
is obtained by analytically solving a polynomial equation. The angle is close to zero, and accurate fault location typically
method provides a closed-form analytical solution, and therefore requires high sampling rates and additional hardware costs
avoids convergence issues of iterative algorithms. In addition, line
asymmetry of untransposed transmission lines are fully considered [6]–[7]. Artificial intelligence based methods (or data-driven
to minimize fault location errors. Extensive numerical experiments methods) locate faults considering the information embedded
show that the proposed method has improved fault location accu- inside the existing measurement data during faults. Nevertheless,
racy compared to the existing method, with different fault types, the methods generally need lots of high-quality training data
locations and impedances. that can represent complete fault characteristics in the practical
Index Terms—Fault location, noniterative method, parameter- transmission line of interest; these high-quality data may not be
free, untransposed transmission lines. available in practice [8]–[9]. In comparison, fundamental fre-
I. INTRODUCTION quency phasor based methods first build equations that describe
the relationship among voltage phasor measurements, current
ELIABLE operation of transmission lines is particularly
R vital to transmit power over long distances in modern
power systems. Faults may occur on the transmission lines.
phasor measurements and the unknown fault location, and then
solve the equations to achieve fault location. These methods are
the most widely adopted fault location methods in practice, and
Accurate fault location technology shortens the time for utility are usually compatible with available measurement devices in
crews to find the location of fault within the line, speeds up substations. Therefore, this paper will focus on the fundamental
frequency phasor based methods.
Manuscript received September 26, 2020; revised January 23, 2021 and March
27, 2021; accepted May 24, 2021. Date of publication June 1, 2021; date of
Fundamental frequency phasor based methods mainly include
current version May 24, 2022. This work was supported by the National Natural one-end and two-end fault location methods. One-end fault lo-
Science Foundation of China under Grant 51807119. Paper no. TPWRD-01444- cation methods usually calculate the location of fault according
2020. (Corresponding author: Yu Liu.)
Dian Lu is with the School of Information Science and Technology, Shang-
to only the local voltage and current phasors. Although the
haiTech University, Shanghai 201210, China, and also with the Shanghai Ad- one-end fault location methods do not require remote side infor-
vanced Research Institute, Chinese Academy of Sciences; Shanghai Institute mation, the fault location accuracy is usually affected by source
of Microsystem and Information Technology, Chinese Academy of Sciences,
University of the Chinese Academy of Sciences, Beijing 100049, China (e-mail:
impedances, loading conditions and fault impedances [10]. To
ludian@shanghaitech.edu.cn). solve these limitations, two-end fault location methods were
Yu Liu, Binglin Wang, and Dayou Lu are with the School of Informa- proposed. These methods utilize voltage and current phasors
tion Science and Technology, ShanghaiTech University, Shanghai 201210,
China (e-mail: liuyu.shanghaitech@gmail.com; wangbl@shanghaitech.edu.cn;
at both ends of the line to locate faults, with the requirement of
ludy@shanghaitech.edu.cn). communication channels between terminals of the line. These
Shi Chen is with the State Grid Jiangsu Electric Power Company Research methods can be classified according to the availability of volt-
Institute, Nanjing 211103, China (e-mail: chenshi9004@126.com).
Color versions of one or more figures in this article are available at https:
age/current measurements [11]–[13], or different models of the
//doi.org/10.1109/TPWRD.2021.3085746. transmission line such as lumped parameter model with or with-
Digital Object Identifier 10.1109/TPWRD.2021.3085746 out consideration of capacitance [10], or distributed parameter
0885-8977 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
LU et al.: IMPROVED NONITERATIVE PARAMETER-FREE FAULT LOCATION METHOD ON UNTRANSPOSED 1357

models [14]–[15]. In addition, untransposed transmission lines


are also adopted in practical power systems and the asymmetry
of untransposed transmission lines will also affect the fault lo-
cation accuracy [16]. To solve these issues, literatures [17]–[18]
proposed fault location methods considering line asymmetry
with iterative calculations.
In fact, aforementioned fundamental frequency phasor based
fault location methods require known and accurate transmission
line parameters. However, the line parameters may not be ac-
curate in the utility dataset, or could even fluctuate due to the Fig. 1. Example test system and typical tower structure in [25]. (a) Example
test system. (b) Typical tower structure.
weather and loading conditions. To overcome this limitation,
researchers proposed the parameter-free fault location methods
[19]–[25]. Literature [19] proposed an iterative parameter-free
fault location method for ideally transposed line neglecting the derivation, instead of directly solving the unknown line param-
shunt admittance of the line. Literature [20] also proposed an eters within the highly nonlinear and complicated problem, two
iterative parameter-free method which improves [19] by con- “key variables” embedded inside the parameter matrices are
sidering the fully distributed parameter line model for ideally extracted by observing the specific characteristics of the line
transposed lines. Literature [21] proposed an iterative parameter- parameter matrices. Afterwards, the problem is much simplified
free method for untransposed lines, by neglecting off-diagonal and the fault location can be analytically expressed by solving
elements of the shunt admittance matrix. In fact, most existing a 4th order polynomial equation. The main contributions of the
parameter-free methods are iterative methods. The main lim- paper are summarized as follows:
itation of iterative methods is that they are sensitive towards r The physical laws of the transmission line before and dur-
initial values and may encounter convergence issues. Therefore, ing the fault are accurately described using multi-section
researchers also proposed noniterative parameter-free fault loca- line model, with fully consideration of line asymmetry and
tion methods. Literature [22] proposed a noniterative parameter- good approximation to distributed line parameters;
free fault location method on ideally transposed transmission r To guarantee noniterative solutions of the fault location,
lines, with lumped impedance line model. Literatures [23]–[24] the maximum possible section number of the multi-section
proposed parameter-free fault location methods with nonitera- line model is mathematically derived; this ensures that the
tive solutions as well, to also consider asynchronization between adopted line model is as accurate as possible;
terminals; the methods are also designed for ideally transposed r The fault location is obtained without solving all the
transmission lines. To consider line asymmetry for untransposed unknown line parameters; key variables are extracted by
lines (which result in additional unknowns in the parameter observing the inherent characteristics of the parameter
matrices), literature [25] proposed a noniterative parameter-free matrices, and the fault location problem is much simplified;
method on untransposed transmission lines, with one-section r Compared to the state-of-the-art existing noniterative
π line model. However, one-section π line model is not quite method, the proposed method models the transmission
accurate and may still compromise fault location accuracy. line more accurately with less assumptions, and therefore
The aim of this paper is to propose a noniterative algorithm demonstrates improved fault location accuracy.
that can accurately locate faults on untransposed transmission The remainder of the paper is organized as follows. Part II
lines without knowing transmission line parameters. The chal- reviews the state-of-the-art existing noniterative parameter-free
lenge of this problem comes from the following fact. On one method on untransposed lines. Part III proposes the new nonit-
hand, the modeling of the transmission line should be accurate erative parameter-free fault location methodology. Part IV com-
enough to consider shunt capacitances (preferably distributed pares the performance of the proposed method to the existing
parameters) as well as asymmetry of the untransposed line. method through numerical experiments. Part V makes further
On the other hand, noniterative analytical solutions usually discussions. Part VI concludes the paper.
require relatively simple transmission line models. Therefore,
this paper proposes a noniterative parameter-free fault location II. REVIEW OF THE EXISTING NONITERATIVE
algorithm on untransposed transmission lines. The proposed PARAMETER-FREE FAULT LOCATION METHOD ON
method can be considered as an extension to [25]. First, the UNTRANSPOSED LINES
multi-section transmission line model with n sections is adopted
to well approximate the fully distributed parameters line model, This part briefly reviews the state-of-the-art noniterative
and the parameter-free fault location problem is formulated by parameter-free fault location method on untransposed transmis-
observing the physical laws that the line should obey before and sion lines [25]. Fig. 1(a) shows the example test system. Three
during the fault. Afterwards, the necessary condition to have phase (subscript ‘abc’) voltage (variable Ṽ ) and current (vari-
noniterative solutions is carefully derived, to provide guidance ˜ phasor measurements before (superscript ‘P’) and during
able I)
on the range of section number n. Finally, the maximum possible (superscript ‘F’) the fault are available at the sending (superscript
value of n (n = 2) is selected and in this case the noniterative ‘s’) and receiving (superscript ‘r’) end of the transmission line:
Ṽ abc , Ṽ abc , I˜abc , I˜abc , Ṽ abc , Ṽ abc , I˜abc and I˜abc . The typical
s,P r,P s,P r,P s,F r,F s,F r,F
parameter free fault location method is derived. During the

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
1358 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 3, JUNE 2022

tower structure of the untransposed transmission line is provided III. PROPOSED NONITERATIVE PARAMETER-FREE FAULT
in Fig. 1(b). This tower structure is widely adopted in EHV LOCATION METHODOLOGY
transmission circuits [25]. Due to geometric characteristics of In this part, the necessity of proposing a new noniterative
the tower, Zaa = Zcc , Zab = Zbc , Yaa = Ycc , and Yab = Ybc . An
parameter-free fault location method is firstly discussed. After-
additional approximation Yac = Yab /2 is made to further reduce
wards, the possibility of noniterative parameter-free fault loca-
the number of parameters. The three phase series impedance and tion method with n-section transmission line models is explored.
shunt admittance matrices of the entire transmission line Zabc
Finally, the proposed noniterative parameter-free fault location
and Yabc are given in (1).
method is derived.
⎡ ⎤ ⎡ ⎤
Zaa Zab Zac Yaa Yab 1/2 · Yab
A. Necessity of a New Parameter-Free Fault Location Method
Z abc = ⎣ Zab Zbb Zab ⎦ Y abc = ⎣ Yab Ybb Yab ⎦
Zac Zab Zaa 1/2 · Yab Yab Yaa The existing noniterative parameter-free method has several
(1) limitations that may compromise its fault location accuracy.
First, the existing method constructs the lumped impedance Firstly, during the derivation of the analytical solutions, the ex-
line model without the line shunt admittance. Equation (2) and isting parameter-free method directly substitute (5) into the term
(I˜abc − xY abc /2 · Ṽ abc ) and [I˜abc − (1 − x)Y abc /2 · Ṽ abc ]
s,F s,F r,F r,F
(3) describe the physical laws before and during the fault, where
x denotes the location of fault (0 ≤ x ≤ 1), in (7), where equation (5) is obtained by the transmission
line model without the line shunt admittance. Secondly, an
Ṽ abc − Z abc I˜abc = Ṽ abc
s,P s,P r,P
(2)
assumption Yac = Yab /2 is made in the existing parameter-free
method, which are not accurate with general tower structures.
Thirdly, the existing parameter-free method only uses one sec-
Ṽ abc − xZ abc I˜abc = Ṽ abc − (1 − x)Z abc I˜abc
s,F s,F r,F r,F
(3)
tion π-equivalent circuit transmission line model, which may not
be quite accurate especially when dealing with relatively long
After substituting (1) into (2), one can express Zabc using Zab , transmission lines.
In fact, the difficulty of proposing an accurate noniterative
Z abc = A
⎡ ⎤ ⎡ ⎤ parameter-free fault location algorithm comes from the fol-
A11 0 A13 B11 1 B11 lowing facts. On one hand, if the transmission line model is
+ Zab B = ⎣ 0 A22 0 ⎦ + Zab ⎣ 1 B22 1 ⎦ (4) relatively simplified (eg. model without shunt admittance, or
A13 0 A11 B11 1 B11 single-section π model), the line parameters can be solved
directly from the measurements, and the noniterative solutions
where matrices A and B are functions of available measurements,
can be obtained. For example in the existing method, Yabc
independent of the line parameter matrices. From (3) and (4),
matrix can be directly solved from (6) because the model is only
the location of fault can be approximated as,
one-section π and also additional assumption of matrix Yabc is
x = (M41 − M43 )/(M11 − M13 ) (5) made (otherwise if the number of unknown variables in Yabc
is larger than 3, the existing method will fail). However, these
where M11 , M13 , M41 and M43 are also functions of available
simplified models and assumptions may generate fault location
measurements, independent of the line parameter matrices.
errors. On the other hand, if the transmission line model is
The second step of the existing method is to construct the
rather complicated (for example multi-section π model or fully
one-section π transmission line model considering the line shunt
distributed model), the parameter matrices Zabc , Yabc and the
admittance. Similarly, equations (6) and (7) can be derived from
fault location x are strongly coupled together, and the derivation
the transmission line model before and during the fault,
of noniterative solutions of either the line parameters or the fault
I˜abc − Y abc /2·(Ṽ abc + Ṽ abc ) + I˜abc = 0
s,P s,P r,P r,P
(6) location are very difficult.
Therefore in the following parts, the authors will select the
Ṽ abc − xZ abc (I˜abc − xY abc /2 · Ṽ abc )
s,F s,F s,F
multi-section π-equivalent circuit transmission line models as
examples, and explore the possibility (necessary condition)
= Ṽ abc − (1 − x)Z abc [I˜abc − (1 − x)Y abc /2 · Ṽ abc ]
r,F r,F r,F
of noniterative parameter-free fault location algorithms with
(7) different selection of section number n. Afterwards, with the
From (1) and (6), one can observe that there are 3 unknowns in above observations, the authors will propose a new noniterative
Yabc and 3 equations. Therefore, the matrix Yabc can be directly parameter-free fault location algorithm.
solved. In addition, x in the terms (I˜abc − xY abc /2 · Ṽ abc ) and
s,F s,F

B. Exploring Noniterative Solutions With


[I˜abc − (1 − x)Y abc /2 · Ṽ abc ] of (7) can be approximated as
r,F r,F
n-Section Line Model
(5). Therefore, the fault location x in (7) can be solved as,
Multi-section transmission line model is a good approxima-
x = (N41 − N43 )/(N11 − N13 ) (8)
tion to the fully distributed parameter line model. Here the
where N11 , N13 , N41 and N43 are also functions of available multi-section line model with n-section π-equivalent circuits
measurements, independent of the line parameter matrices. is adopted. Specifically, the transmission line before the fault

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
LU et al.: IMPROVED NONITERATIVE PARAMETER-FREE FAULT LOCATION METHOD ON UNTRANSPOSED 1359

For ith section during the fault (within the line at the left or
right side of the fault), the KCL and KVL can be expressed as,
si ,F si ,F si ,F si+1 ,F
Ṽ abc − xZ abc /n · (Ĩ abc − xY abc /2n · Ṽ abc ) = Ṽ abc
(12)

si ,F si ,F si+1 ,F si+1 ,F
Ĩ abc − xY abc /2n · Ṽ abc − xY abc /2n · Ṽ abc = Ĩ abc
Fig. 2. The n-section line model before the fault.
(13)

ri ,F ri ,F ri ,F
Ṽ abc − (1 − x)Z abc /n · (Ĩ abc − (1 − x)Y abc /2n · Ṽ abc )
ri+1 ,F
= Ṽ abc (14)

ri ,F ri ,F ri+1 ,F
Ĩ abc − (1 − x)Y abc /2n · Ṽ abc − (1 − x)Y abc /2n · Ṽ abc
ri+1 ,F
= Ĩ abc (15)

Therefore, before the fault, with (10) and (11), one can express
si+1 ,P si+1 ,P si ,P si ,P
Ṽ abc and Ĩ abc using Ṽ abc and Ĩ abc . After substitutions,
r,P sn+1 ,P r,P sn+1 ,P
Fig. 3. The n-section line model during the fault. Ṽ abc ( =Ṽ abc ) and Ĩ abc ( =Ĩ abc ) can be expressed as
s,P s1 ,P s,P s1 ,P
functions of Ṽ abc ( =Ṽ abc ) and Ĩ abc ( =Ĩ abc ),

fV (Ṽ abc , I˜abc , Z abc , Y abc ) = Ṽ abc


(n) s,P s,P r,P
is equally divided into n sections, as shown in Fig. 2. For (16)
transmission line during the fault, the line at the left side of
the fault is equally divided into n sections (same for the line at
fI (Ṽ abc , I˜abc , Z abc , Y abc ) = I˜abc
(n) s,P s,P r,P
si ,P si+1 ,P
the right side of the fault), as shown in Fig. 3. Ṽ abc , Ṽ abc , (17)
I˜abc and I˜abc are the three phase voltage and current phasors
si ,P si+1 ,P (n) (n)
where fV and fI are corresponding functions, and the su-
at the left and right side of the ith section before the fault (i = 1, perscript n denotes the model with n sections.
, I˜abc , I˜abc
endi ,F endi+1 ,F endi ,F endi+1 ,F
2, …, n). Ṽ abc , Ṽ abc are the three Similarly, during the fault, with (12) to (15), one can express
endi+1 ,F endi+1 ,F endi ,F endi ,F
phase voltage and current phasors at the left and right side of the Ṽ abc and Ĩ abc using Ṽ abc and Ĩ abc (end = s,
ith section during the fault (i = 1, 2, …, n), where end (end = s, sn+1 ,F
r). After substitutions, Ṽ abc can be expressed as functions of
r) corresponds to the variables within the line at the left (‘s’) or s,F s1 ,F s,F s1 ,F rn+1 ,F
Ṽ abc ( =Ṽ abc ) and Ĩ abc ( =Ĩ abc ); Ṽ abc can be expressed
the right side (‘r’) of the fault. r,F r1 ,F r,F r1 ,F
The tower structure is consistent with the typical structure as functions of Ṽ abc ( =Ṽ abc ) and Ĩ abc ( =Ĩ abc ). Since
sn+1 ,F rn+1 ,F
described in [25], as shown in Fig. 1(b). To be more general, Ṽ abc = Ṽ abc , the following equation holds,
here the line model is without the assumption of Yac = Yab /2, to (n) s,F s,F
ensure modeling accuracy. Here Zabc and Yabc are redefined as, gV (Ṽ abc , Ĩ abc , Z abc , Y abc , x)
⎡ ⎤ (n) r,F r,F
Zaa Zab Zac = gV (Ṽ abc , Ĩ abc , Z abc , Y abc , 1 − x) (18)
Z abc = ⎣ Zab Zbb Zab ⎦ (9)
(n)
Zac Zab Zaa where gV is the corresponding function, and the superscript n
denotes the model with n sections. Note that from the symmetry
⎡ ⎤
Yaa Yab Yac of the problem, the functions at the left and the right side of the
Y abc = ⎣ Yab Ybb Yab ⎦ equation are the same but with different input variables.
Yac Yab Yaa Next, the possibility of deriving a noniterative analytical so-
lution of fault location x from equation (16) to (18) is discussed.
For ith section before the fault, the Kirchhoff’s voltage laws First, one can observe that there are 9 complex equations, 8
(KVLs) and current laws (KCLs) can be expressed as, complex unknown variables (4 in Zabc and 4 in Yabc , respec-
si ,P si ,P si ,P si+1 ,P tively) and 1 real unknown variable x. If we rewrite the complex
Ṽ abc − Z abc /n · (Ĩ abc − Y abc /2n · Ṽ abc ) = Ṽ abc
equations into real equations by separating real and imaginary
(10)
parts, there are 18 equations and 17 unknowns. Also, since
the rows of equations are generally independent, theoretically
si ,P si ,P si+1 ,P si+1 ,P
Ĩ abc − Y abc /2n · Ṽ abc − Y abc /2n · Ṽ abc = Ĩ abc speaking one can use equation (16) to (18) to calculate all
(11) the unknown parameters and the fault location x via iterative

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
1360 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 3, JUNE 2022

methods. However, similar as existing iterative methods, since First, with n = 2, equation (16) to (18) can be rewritten as,
these equations are highly nonlinear, iterative methods will be
Ṽ abc − Z abc /2·(I˜abc − Y abc /4·Ṽ abc )
s,P s,P s,P
sensitive to initial values and possibly encounter convergence (19)
= Ṽ abc − Z abc /2·(I˜abc − Y abc /4·Ṽ abc )
r,P r,P r,P
issues. Therefore, noniterative analytical solutions are preferable
in this case.
Nevertheless, since these equations are complicated and Before the fault, for (19) and (20) as shown at the bottom of
highly nonlinear, noniterative analytical solutions are generally this page, the difference between the 1st and the 3rd row can be
very difficult to be obtained with an arbitrarily given section rewritten as,
number n. In fact, from (12) and (13), the highest order of x is
si ,F si ,F si+1 ,F Zaa − Zac
2 (or 3) if we utilize Ṽ abc and Ĩ abc to express Ṽ abc (or
si+1 ,F
Ĩ abc ) (similarly for equation 14 and 15). Recursively, one can = 2(J1 − J2 )/[(J3 − J4 ) − (Yaa − Yac )(J1 − J2 )/4] (22)
observe that the highest order of x is 2n (or 2n+1) if we utilize
end1 ,F end1 ,F endn+1 ,F endn+1 ,F
Ṽ abc and Ĩ abc to express Ṽ abc (or Ĩ abc ) (end
Zaa − Zac
= s, r). Therefore, from (18), even we aggressively assume that
we somehow solve all the parameters of Zabc and Yabc from (Yaa − Yac )(J1 + J2 )/2 − (J3 + J4 )
=
(16) to (18) (which is not quite likely), one can observe that the (Yaa − Yac )(J3 + J4 )/8 − (Yaa − Yac )2 (J1 + J2 )/32
highest order of x in (18) is 2n. Since there is no analytical solu- (23)
tion (or mathematically speaking “no solution in radicals”) to a
general polynomial equation of degree 5 or higher, the necessary where J1 = Ṽas,P − Ṽcs,P , J2 = Ṽar,P − Ṽcr,P , J3 = I˜as,P −
condition to have a noniterative solution is to select n such that 2n I˜cs,P , and J4 = I˜ar,P − I˜cr,P .
< 5. Note that here larger section number corresponds to more One can observe that from (22) and (23), Ykey = Yaa −Yac
accurate transmission line model, and therefore the maximum and Zkey = Zaa −Zac are two “key variables” of the parameter
possible value n = 2 is selected. Also, here n = 2 satisfies the matrices Zabc and Yabc . From (22) and (23), Ykey and Zkey can
necessary condition but the condition itself is not sufficient: there be analytically solved. Substitute (22) into (23),
is no guarantee of noniterative solutions even with n = 2, and
H1 Ykey 2 + H2 Ykey + H3 = 0 (24)
therefore careful derivations are still required.
where H1 = (J1 2 − J2 2 )/16, H2 = −(J1 + J2 )(J3 − J4 )/2 ,
H3 = J3 2 − J4 2 .
C. The Proposed Parameter-Free Fault Location Method The solutions to (24) is,
Next, the noniterative solution with n = 2 is mathematically  
derived. Note that the derivation of the noniterative solution Ykey = (−H2 ± H2 2 − 4H1 H3 ) (2H1 ) (25)
with n = 2 is much more difficult than that of the existing
method, for the following reasons. First, unlike the existing Note that there are two solutions of Ykey . The correct solution
method which directly solves the matrix Yabc from (6) with can be obtained by checking the order of magnitude and the
3 complex equations and 3 complex unknowns, there are 6 polarity (positive/negative) of the value. Afterwards, Zkey can
equations and 8 knowns in (16) and (17), and matrices Zabc and also be calculated by substituting Ykey into (22). The calculation
Yabc are strongly coupled together. Therefore, direct solutions of of the two “key variables” only requires available measurements,
Zabc and Yabc are extremely difficult. In addition, if we solve the and does not require line parameter matrices.
fault location x together with Zabc and Yabc in (18), the highest During the fault, for (21) as shown at the bottom of the next
order of the equation could be easily higher than 4, resulting in page, similarly, the difference between the 1st and the 3rd row
no noniterative solutions. can be rewritten as,
With above challenges, in this part we carefully derive the ax4 + bx3 + cx2 + dx + e = 0 (26)
noniterative solution by taking full use of the inherent geometric
characteristics of the untransposed transmission line. Instead of where a = Zkey 2 Ykey 2 (K2 − K1 )/32, b = Zkey 2 Ykey (K4 −
solving the entire Zabc and Yabc , we first extract two “key vari- Ykey K2 +K3 )/8, c=Zkey Ykey (K2 −K1 )/2−3/8 · Zkey 2 Ykey
ables” embedded inside the parameter matrices of the untrans- K4 +3/16 · Zkey 2 Ykey 2 K2 , d = Zkey (K4 + K3 ) − Zkey Ykey
posed transmission line. After obtaining the two key variables K2 + Zkey 2 Ykey [3/8 · K4 − 1/8 · Ykey K2 ], e = K2 − K1 −
from the available measurements, the fault location x can be Zkey K4 +Zkey Ykey K2 /2−Zkey 2 Ykey K4 /8+Zkey 2 Ykey 2 K2 /
solved. Through this routine, the complicated nonlinearity of 32, K1 = Ṽas,F − Ṽcs,F , K2 = Ṽar,F − Ṽcr,F , K3 = I˜as,F −
the problem can be much mitigated. I˜cs,F , K4 = I˜ar,F − I˜cr,F .

I˜abc − Y abc /4 · Ṽ abc − Y abc /4 · [Ṽ abc − Z abc /2 · (I˜abc − Y abc /4 · Ṽ abc )]
s,P s,P s,P s,P s,P

= −{I˜abc − Y abc /4 · Ṽ abc − Y abc /4 · [Ṽ abc − Z abc /2 · (I˜abc − Y abc /4 · Ṽ abc )]}
r,P r,P r,P r,P r,P
(20)

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
LU et al.: IMPROVED NONITERATIVE PARAMETER-FREE FAULT LOCATION METHOD ON UNTRANSPOSED 1361

One can observe that, due to the inherent geometric character- in PSCAD/EMTDC, as shown in Fig. 1(a). The system is a
istics of the untransposed transmission line, the equation consists three phase system with the rated voltage of 500 kV and the
of Ykey , Zkey , x, and other available measurements, with x as the nominal frequency of 50 Hz. The source impedances per phase
only unknown variable. The highest order of x in (26) is 4 = 2n, at the sending and the receiving end are 10∠80◦ Ω and 15∠75◦ Ω,
which is consistent with the results in part III.B. The analytical respectively. The phase angles of the power sources at the
solutions to (26) are, sending and the receiving end are 30° and 0°, respectively. The
transmission line of interest is with the length of 200 km and
b
x1,2 = − with the tower structure in Fig. 1(b), where h = 30 m, h1 =
4a
  2 m and d = 10 m. The frequency dependent model (phase)
1 b2 2c 1 b2 4c of the transmission line is adopted during PSCAD/EMTDC
− 2
− +Δ± − − Δ − Δ3 (27)
2 4a 3a 2 2a2 3a simulation to ensure accuracy. Three phase voltage and current
measurements are installed at both terminals of the transmission
line. The instantaneous voltages and currents are first captured
b
x3,4 = − using 4000 samples/second (80 samples/cycle) sampling rate
4a according to IEC61850-9-2LE standard [26]. Afterwards, the
 
1 b2 2c 1 b2 4c phasors are calculated according to IEEE C37.118 standard
+ 2
− +Δ± − − Δ + Δ3 (28)
2 4a 3a 2 2a2 3a [27]. Note that the voltage level of untransposed lines can vary
within a large range (eg. 345kV [16], 400kV [25], 500kV [29],
where Δ1 = c2 − 3bd + 12ae ,  Δ2 = 2c3 − 9bcd + 27ad2 + 765kV [30], 1000kV [31], etc.). Here a 500 kV untransposed

27b2 e − 72ace, Δ = 2Δ1 /(3a Δ2 + −4Δ1 3 + Δ2 2 ) +
3 3
line is selected as an example, however, the applicability of the
 √ proposed method is similar for other voltage levels.
Δ2 + −4Δ1 3 +Δ2 2 /(3 3 2a), and Δ3 =(−b3 /a3 +4bc/a2
3

Next the proposed method is validated via 5 test cases with


− 8d/a)/(4 b2 /4a2 − 2c/3a + Δ). different fault types and fault impedances. In each test case,
Here there are four solutions of fault location x. The correct different fault locations through the entire line (21 fault loca-
solution can be obtained by checking whether the fault location tions, every 10 km) and 4 different fault impedances are studied,
is within the range 0 ≤ x ≤ 1. resulting in 84 fault events. The absolute fault location errors (in
It is worth noting that similar ideas can be applied when the percentage) of the proposed noniterative fault location method
section number n is selected as 3, as shown in the Appendix. are compared to those of the existing state-of-the-art noniterative
However, one can observe from equation (A7) that the highest fault location method in [25]. The absolute fault location error
order of x in the corresponding polynomial is 6 = 2n, which is (in percentage) is defined as,
also consistent with the results in part III.B. This polynomial
does not have noniterative solutions. This further proves that
the maximum possible value of n should indeed be 2 to ensure Error(%)
noniterative solutions to the problem. Estimated location − Actual location
To sum up, the proposed fault location method does not need to = × 100% (29)
Line length
calculate all the parameters of the transmission line of interest.
With the observation of inherent geometric characteristics of
the untransposed lines, the proposed method first extracts and A. Test Case 1: Single Phase to Ground Faults
solves two key variables. Afterwards, the fault location can be
This test case studies low impedance A-G faults with different
accurately calculated with a noniterative analytical solution.
fault locations and fault impedances. The fault impedances are
The modeling accuracy as well as the fault location accuracy
chosen as 0.01 ohm, 1 ohm, 5 ohm and 10 ohm. The fault
of the proposed noniterative fault location method is improved
location errors of the existing and the proposed method are
compared to the existing noniterative method.
depicted in Fig. 4 and Table I. The maximum fault location
errors of the existing and the proposed method are 1.8427%
IV. NUMERICAL EXPERIMENTS
and 0.0537%, respectively. The results demonstrate improved
In order to verify the performance of the proposed method, fault location accuracy of the proposed method compared to the
a two-terminal untransposed transmission line system is built existing method.

Ṽ abc − Z abc x/2 · (I˜abc − Y abc x/4 · Ṽ abc ) − Z abc x/2 · {I˜abc
s,F s,F s,F s,F

−Y abc x/4 · Ṽ abc − Y abc x/2 · [Ṽ abc − Z abc x/2 · (I˜abc − Y abc x/4
s,F s,F s,F

·Ṽ abc )]} = Ṽ abc − Z abc (1 − x)/2 · [I˜abc − Y abc (1 − x)/4 · Ṽ abc ]
s,F r,F r,F r,F

−Z abc (1 − x)/2 · {I˜abc − Y abc (1 − x)/4 · Ṽ abc − Y abc (1 − x)/2


r,F r,F

·{Ṽ abc − Z abc (1 − x)/2 · [I˜abc − Y abc (1 − x)/4 · Ṽ abc ]}}


r,F r,F r,F
(21)

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
1362 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 3, JUNE 2022

Fig. 4. Absolute errors of low impedance A-G faults with different fault Fig. 6. Absolute errors of low impedance BC-G faults with different fault
locations and impedances. locations and impedances.

TABLE I TABLE III


AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF LOW IMPEDANCE A-G AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF LOW IMPEDANCE BC-G
FAULTS WITH DIFFERENT FAULT LOCATIONS AND IMPEDANCES FAULTS WITH DIFFERENT FAULT LOCATIONS AND IMPEDANCES

Fig. 5. Absolute errors of low impedance B-C faults with different fault Fig. 7 Absolute errors of low impedance three phase faults with different fault
locations and impedances. locations and impedances.

TABLE II
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF LOW IMPEDANCE B-C C. Test Case 3: Double Phase to Ground Faults
FAULTS WITH DIFFERENT FAULT LOCATIONS AND IMPEDANCES
This test case studies low impedance BC-G faults with differ-
ent fault locations and fault impedances. The fault impedances
are chosen as 0.01 ohm, 1 ohm, 5 ohm and 10 ohm. The fault
location errors of the existing and the proposed method are
depicted in Fig. 6 and Table III. The maximum fault location
errors of the existing and the proposed method are 1.5556%
and 0.0223%, respectively. The results demonstrate improved
fault location accuracy of the proposed method compared to the
B. Test Case 2: Phase to Phase Faults existing method.
This test case studies low impedance B-C faults with different
D. Test Case 4: Three Phase Faults
fault locations and fault impedances. The fault impedances are
chosen as 0.01 ohm, 1 ohm, 5 ohm and 10 ohm. The fault This test case studies low impedance three phase faults
location errors of the existing and the proposed method are with different fault locations and fault impedances. The fault
depicted in Fig. 5 and Table II. The maximum fault location impedances are chosen as 0.01 ohm, 1 ohm, 5 ohm and 10 ohm.
errors of the existing and the proposed method are 1.0954% The fault location errors of the existing and the proposed method
and 0.0273%, respectively. The results demonstrate improved are depicted in Fig. 7 and Table IV. The maximum fault location
fault location accuracy of the proposed method compared to the errors of the existing and the proposed method are 0.8085%
existing method. and 0.0260%, respectively. The results demonstrate improved

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
LU et al.: IMPROVED NONITERATIVE PARAMETER-FREE FAULT LOCATION METHOD ON UNTRANSPOSED 1363

TABLE IV shunt admittances during the derivation) and the third limitation
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF LOW IMPEDANCE THREE
PHASE FAULTS WITH DIFFERENT FAULT LOCATIONS AND IMPEDANCES
(single section π model) are the primary reasons for the improved
fault location results.

V. DISCUSSIONS
In this part, further discussions are made to ensure practica-
bility of the proposed method, including pseudo solutions of
the algorithm, effect of loading conditions, effect of measure-
ment errors, effect of line parameters, effect of unsynchronized
measurements, comparison to other existing methods and impor-
tance to model line asymmetry for untransposed lines, and ap-
plicability to transposed lines. For part V.A to V.E, 100 ohm A-G
faults are selected as examples. For part V.F and V.G, A-G faults
with different low and high fault impedances are selected as
examples. Note that the proposed method demonstrates similar
advantages towards existing methods for other fault scenarios.

A. Pseudo Solutions of the Algorithm

Fig. 8. Absolute errors of high impedance A-G faults with different fault
One can observe from (25), (27) and (28) that there could
locations and impedances. be more than one solution during the proposed fault location
procedure. Among these solutions, only one solution is the
correct solution while others should be neglected (here we refer
TABLE V
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF HIGH IMPEDANCE A-G them as ‘pseudo solutions’).
FAULTS WITH DIFFERENT FAULT LOCATIONS AND IMPEDANCES First, equation (25) will result in 2 solutions of Ykey . As
mentioned after (25) in part III.C, the pseudo solutions can be
neglected by checking the order of magnitude and the polarity
of the value. For a typical 200 km overhead transmission line,
Ykey should be a positive number, with the order of magnitude
of 10−4 ∼10−3 mho. Note that since the calculation of Ykey
only requires measurements prior to the fault, the fault types,
locations and impedances will not influence the calculation
results of Ykey . The two solutions of Ykey are (2.0125×10−6 +
fault location accuracy of the proposed method compared to the j9.0063×10−4 mho) and (0.0075 – j0.3151 mho) with the magni-
existing method. tude of 9.01×10−4 mho and 3.15×10−1 mho, respectively. It can
be observed that only one solution is with a positive imaginary
E. Test Case 5: High Impedance Faults part, and falls into the correct order of magnitude (10−4 ∼10−3
This test case studies high impedance A-G faults with differ- mho). The other solution is with a negative imaginary part, and
ent fault locations and fault impedances. The fault impedances the magnitude of the solution is around 350 times larger than the
are chosen as 100 ohm, 200 ohm, 300 ohm and 500 ohm. The correct solution. This proves that the pseudo solutions of Ykey
fault location errors of the existing and the proposed method can be clearly identified and neglected.
are depicted in Fig. 8 and Table V. The maximum fault location Second, equation (27) and (28) will result in 4 solutions of
errors of the existing and the proposed method are 28.8406% fault location x. As mentioned after (28) in part III.C, the pseudo
and 0.8080%, respectively. The results demonstrate improved solutions can be neglected by observing that the correct fault
fault location accuracy of the proposed method compared to the location is within the range 0 ≤ x ≤ 1. Here the results of a
existing method. group of A-G faults with 100 ohm fault impedance and different
fault locations through the line are shown as examples. The four
F. Summary roots of x are depicted in Fig. 9. One can clearly observe that
only one solution falls within the range of 0 ≤ x ≤ 1, while
From test cases 1 to 5, one can observe that the proposed other solutions are way beyond the range. Therefore, the pseudo
method has improved fault location accuracy compared to solutions of x can also be clearly identified and neglected.
the existing methods, regardless of fault types, locations and
impedances. It is worth noting that the improved accuracy of
the proposed method comes from the fact that it overcomes B. Effect of Loading Conditions
the three limitations as mentioned in the first paragraph of part This test case studies 100 ohm A-G faults with different fault
III.A. Specifically for the example test system, among those locations and loading conditions (various phase angle differ-
limitations, the first limitation (utilization of line model without ences between the two sources). The fault location errors of

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
1364 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 3, JUNE 2022

Fig. 11. Absolute errors of 100 ohm A-G faults with different fault locations
Fig. 9. Four solutions of calculated fault location x for 100 ohm A-G faults
and measurement conditions.
with different fault locations.

TABLE VII
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF 100 OHM A-G FAULTS WITH
DIFFERENT FAULT LOCATIONS AND MEASUREMENT CONDITIONS

TABLE VIII
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF 100 OHM A-G FAULTS WITH
Fig. 10. Absolute errors of 100 ohm A-G faults with different fault locations DIFFERENT FAULT LOCATIONS AND LINE PARAMETERS
and loading conditions.

TABLE VI
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF 100 OHM A-G FAULTS WITH
DIFFERENT FAULT LOCATIONS AND LOADING CONDITIONS

C. Effect of Measurement Errors


This test case studies 100 ohm A-G faults with different
fault locations and different measurement errors. Here Gaussian
distributed errors with different standard deviations are added
the existing method [25] and the proposed method are shown to the instantaneous measurements. The fault location errors of
in Fig. 10 and Table VI. For those results, the phase angle the existing method [25] and the proposed method are shown in
of the sending end is leading that of the receiving end. The Fig. 11 and Table VII. The maximum fault location errors of the
maximum fault location errors of the existing and the proposed existing and the proposed method are 4.2747% and 1.6192%,
method are 6.8082% and 0.2906%, respectively. The results respectively. The results demonstrate improved fault location
demonstrate improved fault location accuracy of the proposed accuracy of the proposed method compared to the existing
method compared to the existing method with different loading method with different measurement errors.
conditions.
From Fig. 10, the curve of the existing method becomes less
“symmetrical” (the axis of symmetry is the midpoint of the D. Effect of Line Parameters
line where fault location = 100 km) with higher phase angle This test case studies 100 ohm A-G faults with different fault
differences. In fact, one can also observe from Fig. 4 to Fig. 8 that locations and different line parameters. Here 5 parameters are
the curve of the existing method becomes less “symmetrical” considered: C1 represents the distance between phase conduc-
with higher fault impedances. The main reason is that the load tors (d in Fig. 1(b)); C2 represents the height of phase conductors
current itself is “asymmetrical” (in this case the load current (h in Fig. 1(b)). C3 represents the diameter of phase conductors;
is flowing from the sending end to the receiving end). Within C4 represents the DC resistance of phase conductors; and C5
the measured fault current, the percentage of the load current represents the resistivity of the soil. The fault location errors
(or percentage of the “asymmetry”) is higher with larger phase of the proposed method with parameter variations (+50% and
angle differences or higher fault impedances. −50%) are shown in Fig. 12 and Table VIII (the maximum and

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
LU et al.: IMPROVED NONITERATIVE PARAMETER-FREE FAULT LOCATION METHOD ON UNTRANSPOSED 1365

TABLE IX
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF 100 OHM A-G FAULTS WITH
DIFFERENT FAULT LOCATIONS AND SYNCHRONIZATION ERRORS

Fig. 12. Absolute errors of 100 ohm A-G faults with different fault locations
the fault location errors are mainly caused by the synchronization
and line parameters. error; in this case the performances of the two methods are
comparable, and the fault location errors are relatively large.
Therefore, noniterative parameter-free fault location method on
untransposed lines with unsynchronized measurements may still
require further investigations in the future.

F. Comparison to Other Existing Methods and Importance to


Model Line Asymmetry for Untransposed Lines
In fact, many parameter-free fault location methods in the
existing literatures are with the assumption that the transmission
line is ideally-transposed. However, applying these methods to
untransposed transmission lines will cause fault location errors
since the asymmetry of the transmission line is not considered.
Fig. 13. Absolute errors of 100 ohm A-G faults with different fault locations
and synchronization errors. This test case first studies A-G faults on the untransposed trans-
mission line with different fault locations and different low fault
impedances (0.01 ohm, 1 ohm, 5 ohm and 10 ohm). To further
validate the effectiveness of the proposed method, here we select
the average absolute fault location errors with original parame-
two existing methods [20], [24] (with the assumption of ideally
ters can be found in Table V). The maximum fault location errors
transposed lines) and one existing method [28] (considering
with parameter variations and original parameters are 0.2158%
asymmetry of untransposed lines) as examples. Method [20]
and 0.1916%, respectively. The results show that the proposed
is an iterative parameter-free fault location method for ideally
method is insensitive towards different line parameters.
transposed lines, with fully distributed parameter model. Here
to ensure best performance of this iterative method, the initial
E. Effect of Unsynchronized Measurements
values are selected to be very close to the actual parameters
In this paper, best implementation of the proposed fault lo- and fault location (in practice inappropriate selection of initial
cation method requires synchronized measurements, which can values could possible cause convergence issues). Method [24] is
be achieved via GPS synchronization. Nevertheless, for cases a noniterative parameter-free fault location method for ideally
where accurate GPS synchronization is not available (loss of transposed lines, with lumped impedance line model. Method
GPS signal, GPS spoofing attack, GPS not installed, etc.), if the [28] is a noniterative parameter-free fault location method for
synchronization error cannot be corrected, there will be addi- untransposed lines, with lumped impedance line model that con-
tional fault location errors, since the derivation of the proposed siders line asymmetry. The fault location errors of the existing
method assumes synchronized measurements at terminals (same methods [20], [24], [28] and the proposed method are shown in
for the existing method [25]). Fig. 14 and Table X. The maximum fault location errors of the
To examine the effect of unsynchronized measurements, a existing methods [20], [24], [28] and the proposed method are
group of 100 ohm phase A to ground faults are studied, with 1.6462%, 9.8295%, 1.8574% and 0.0537%, respectively. One
different fault locations and different synchronization errors. can observe that the lumped impedance line model in [24] brings
The example synchronization errors are 0, 1, 2 and 5 degrees, relatively large error for long transmission lines. In addition,
respectively. The fault location errors of the existing method [25] even with fully distributed parameter model in [20], the fault
and the proposed method are shown in Fig. 13 and Table IX. location accuracy could still be compromised since line asym-
One can observe that the fault location error increases with metry is not considered. Furthermore, although method [28] only
larger synchronization error. If the synchronization error is rather utilizes the lumped impedance line model, the accuracy of fault
small, the proposed method still demonstrates improved fault location is still comparable to method [20] (fully distributed
location accuracy compared to the existing method, and the fault parameter model) due to the consideration of line asymmetry.
location accuracy remains within an acceptable range. However, These results demonstrate that the proposed method presents
when the synchronization error is relatively large (eg. 5 degrees), higher accuracy compared to other existing methods, and it is

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
1366 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 3, JUNE 2022

TABLE XI
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF HIGH IMPEDANCE A-G
FAULTS WITH DIFFERENT FAULT LOCATIONS AND IMPEDANCES,
METHOD [20], [24], [28] AND PROPOSED METHOD

Fig. 14. Absolute errors of low impedance A-G faults with different fault
locations and impedances, method [20], [24], [28] and proposed method.

TABLE X
AVERAGE AND MAXIMUM ABSOLUTE ERRORS OF LOW IMPEDANCE A-G
FAULTS WITH DIFFERENT FAULT LOCATIONS AND IMPEDANCES, METHOD [20],
[24], [28] AND PROPOSED METHOD

Fig. 16. Absolute errors of A-G faults with different fault locations and low
fault impedances on ideally transposed lines.

method [20] and the proposed method are relatively small,


while the rest of the two methods have much larger errors. The
proposed method still presents higher fault location accuracy
compared to the existing methods. Note that method [20] is an
iterative method and may encounter convergence issues without
proper initial values, while the proposed method is noniterative
with analytical solutions.

G. Applicability to Transposed Lines


Although the proposed method is designed for untransposed
lines, the proposed fault location method still works equally well
on ideally transposed transmission lines. For ideally transposed
lines, the parameter matrices have the following structures: for
Fig. 15. Absolute errors of high impedance A-G faults with different fault either Zabc or Yabc matrix, the diagonal elements are the same
locations and impedances, method [20], [24], [28] and proposed method. (Zs or Ys ), and the off-diagonal elements are the same (Zm or
Ym ). One can observe that these structures also satisfy (9), and
therefore the method is also be applicable to ideally transposed
essential to model line asymmetry for untransposed transmission lines (ideally transposed lines can be considered as a special case
lines. of untransposed lines).
To examine the performances of those methods during high To validate the applicability of the proposed method to
impedance faults, a group of high impedance A-G faults on the ideally transposed lines, the line of interest in the example
untransposed line is studied, with fault impedances of 100 ohm, test system is updated as an ideally transposed line (the rest
200 ohm, 300 ohm and 500 ohm. The results are provided in of the test system remains unchanged). A group of phase A
Fig. 15 and Table XI. The maximum fault location errors of to ground faults are studied with different fault locations and
the existing methods [20], [24], [28] and the proposed method impedances (0.01 ohm, 1 ohm, 5 ohm,10 ohm, 100 ohm, 200
are 1.6462%, 164.7048%, 28.3840% and 0.8080%, respectively. ohm, 300 ohm and 500 ohm). Fig. 16 and Fig. 17 show the results
One can observe that the fault location errors of the existing of the proposed fault location method on ideally transposed

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
LU et al.: IMPROVED NONITERATIVE PARAMETER-FREE FAULT LOCATION METHOD ON UNTRANSPOSED 1367

VI. CONCLUSION
This paper proposes a noniterative parameter-free fault
location method on untransposed transmission lines, which
provides a closed-form analytical fault location solution with
full consideration of line asymmetry. To ensure model accuracy,
the paper first explores possibilities of noniterative fault location
solutions with multi-section line model, resulting in maximum
possible selection of the section number. Next, with the selected
section number, the analytical solution of the fault location is
carefully derived by observing the special characteristics of
Fig. 17. Absolute errors of A-G faults with different fault locations and high
fault impedances on ideally transposed lines. the line parameter matrices. In this case, instead of solving
all the line parameters, the method only needs to solve two
key variables and the solution to the fault location problem is
greatly simplified. In the numerical experiments, the results
verify that the proposed method demonstrates higher accuracy
transmission lines. The maximum fault location errors with than the existing method, regardless of different locations, types
0.01 ohm, 1 ohm, 5 ohm, 10 ohm, 100 ohm, 200 ohm, 300 and impedances of faults. The method also shows robustness
ohm and 500 ohm fault impedances are 0.0509%, 0.0461%, against different loading conditions, measurement errors and
0.0461%, 0.0552%, 0.2067%, 0.3729%, 0.5329% and 0.8713% parameter variations, and is also applicable to transposed lines.
respectively. In comparison, the maximum fault location errors
of the proposed method on untransposed transmission lines APPENDIX
with corresponding fault impedances are comparable: 0.0490%,
0.0464%, 0.0458%, 0.0537%, 0.1916%, 0.3451%, 0.4995% This Appendix provides the derivation of the parameter-free
and 0.8080% respectively, as shown in Table I and Table V. fault location method with section number n = 3, which is an
Therefore, the proposed fault location method is applicable to example to verify that there is indeed no noniterative solution
not only untransposed lines but also ideally transposed lines. when the section number is n ≥ 3. The idea of the parameter-free

I˜abc − Y abc Ṽ abc 6 − Y abc [Ṽ abc − Z abc (I˜abc − Y abc Ṽ abc 6) 3] 3
s,P s,P s,P s,P s,P

(A1)
= −{(I˜abc − Y abc Ṽ abc 6 − Y abc [Ṽ abc − Z abc (I˜abc − Y abc Ṽ abc 6) 3] 3}
r,P r,P r,P r,P r,P

Ṽ abc − Z abc (I˜abc − Y abc Ṽ abc 6) 3 − Z abc /3·{I˜abc − Y abc Ṽ abc 6 − Y abc /3
s,P s,P s,P s,P s,P

(A2)
[Ṽ abc − Z abc (I˜abc − Y abc Ṽ abc 6) 3] = Ṽ abc − Z abc (I˜abc − Y abc Ṽ abc 6) 3
s,P s,P s,P r,P r,P r,P

Ṽ abc − xZ abc /3·(I˜abc − xY abc /6·Ṽ abc ) − xZ abc /3·(I˜abc − xY abc /6·Ṽ abc −
s,F s,F s,F s,F s,F

xY abc /3·(Ṽ abc − xZ abc /3·(I˜abc − xY abc /6·Ṽ abc ))) − xZ abc /3 · (I˜abc − xY abc
s,F s,F s,F s,F

/6 · Ṽ abc − xY abc /3·(Ṽ abc − xZ abc /3·(I˜abc − xY abc /6·Ṽ abc )) − xY abc /3
s,F s,F s,F s,F

(Ṽ abc − xZ abc /3·(I˜abc − xY abc /6·Ṽ abc ) − xZ abc /3·(I˜abc − xY abc /6·Ṽ abc
s,F s,F s,F s,F s,F

−xY abc /3·(Ṽ abc − xZ abc /3·(I˜abc − xY abc /6·Ṽ abc ))))) = Ṽ abc − (1 − x)Z abc
s,F s,F s,F r,F

/3 · (I˜abc − (1 − x)Y abc /6·Ṽ abc ) − (1 − x)Z abc /3·(I˜abc − (1 − x)Y abc /6·Ṽ abc
r,F r,F r,F r,F
(A3)
− (1 − x)Z abc /3·(I˜abc − (1 − x)Y abc /6·Ṽ abc ))) − (1 − x)
r,F r,F r,F
−(1 − x)Y abc /3·(Ṽ abc
Z abc /3·(I˜abc − (1 − x)Y abc /6·Ṽ abc − (1 − x)Y abc /3·[Ṽ abc − (1 − x)Z abc /3·
r,F r,F r,F

(I˜abc − (1 − x)Y abc /6·Ṽ abc )] − (1 − x)Y abc /3·(Ṽ abc − (1−x)Z abc /3·(I˜abc
r,F r,F r,F r,F

−(1 − x)Y abc /6·Ṽ abc ) − (1 − x)Z abc /3·(I˜abc − (1 − x)Y abc /6·Ṽ abc −
r,F r,F r,F

(1 − x)Y abc /3·(Ṽ abc − (1 − x)Z abc /3·(I˜abc − (1 − x)Y abc /6·Ṽ abc )))))
r,F r,F r,F

(J3 + J4 ) − (Yaa − Yac )(J1 + J2 )/2


Zaa − Zac = (A4)
(Yaa − Yac )2 (J1 + J2 )/54 − (Yaa − Yac )(J3 + J4 )/9
0 = (J1 − J2 ) − (Zaa − Zac )(2J3 − J4 )/3 + (Zaa − Zac )(Yaa − Yac )/18 · (4J1
(A5)
−J2 ) − (Zaa − Zac )2 (Yaa − Yac )J3 /27 + (Zaa − Zac )2 (Yaa − Yac )2 J1 /162

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
1368 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 37, NO. 3, JUNE 2022

fault location with n = 3 is the same as n = 2 as shown in part [5] D. Lu, Y. Liu, Q. Liao, B. Wang, W. Huang, and X. Xi, “Time-domain trans-
III.C. First, the physical laws that the line should obey before mission line fault location method with full consideration of distributed
parameters and line asymmetry,” IEEE Trans. Power Del., vol. 35, no. 6,
and during the fault are described. Afterwards, the key variables pp. 2651–2662, Dec. 2020.
are extracted and the polynomial equation that the fault location [6] C. Y. Evrenosoglu and A. Abur, “Traveling wave based fault location for
should satisfy is derived. Note that the definitions of variables teed circuits,” IEEE Trans. Power Del., vol. 20, no. 2, pp. 1115–1121,
Apr. 2005.
are consistent with those in part III.C. [7] F. V. Lopes, “Settings-free traveling-wave-based earth fault location using
With n = 3, equation (16) to (18) can be rewritten as, unsynchronized two-terminal data,” IEEE Trans. Power. Del., vol. 31,
Before the fault, for (A1) and (A2) as shown at the bottom of no. 5, pp. 2296–2298, Oct. 2016.
[8] J. Gracia. A. J. Mazon, and I. Zamora, “Best ANN structure for fault
the previous page, the difference between the 1st and the 3rd row location in single-and double-circuit transmission lines,” IEEE Trans.
can be rewritten as, Power Del., vol. 20, no. 4, pp. 2389–2395, Oct. 2005.
In (A4) and (A5) as shown at the bottom of the previous page, [9] M. Mirzaei, B. Vahidi, and S. H. Hosseinian, “Accurate fault location
and faulted section determination based on deep learning for a parallel-
similarly one can observe that Ykey = Yaa −Yac and Zkey = compensated three-terminal transmission line,” IET Gener. Transmiss.
Zaa −Zac are the two “key variables” of the parameter matrices Distrib., vol. 13, no. 13, pp. 2770–2778, Jul. 2019.
Zabc and Yabc . From (A4) and (A5), Ykey and Zkey can be [10] IEEE Guide for Determining Fault Location on AC Transmission and
Distribution Lines, IEEE Standard C37.114-2014, 2014.
analytically solved. Substitute (A4) into (A5), [11] C. J. Lee, J. B. Park, J. R. Shin, and Z. M. Radojevie, “A new two-terminal
numerical algorithm for fault location, distance protection, and arcing fault
L1 Ykey 3 + L2 Ykey 2 + L3 Ykey + L4 = 0 (A6) recognition,” IEEE Trans. Power Syst., vol. 21, no. 3, pp. 1460–1462,
Aug. 2006.
where L1 = (J1 + J2 )2 [(J1 − J2 )/2916 − (4J1 − J2 )/1944 [12] S. M. Brahma and A. A. Girgis, “Fault location on a transmission line using
+ J1 /648], L2 = [4(4J1 − J2 )(J1 + J2 )(J3 + J4 ) + 3(2J3 − synchronized voltage measurements,” IEEE Trans. Power Del., vol. 19,
no. 4, pp. 1619–1622, Oct. 2004.
J4 )(J1 + J2 )2 ]/972 − (J1 2 −J2 2 )(J3 + J4 )/243 − [J1 · (J1 [13] M. M. Saha, J. Izykowski, and E. Rosolowski, “A two-ended method of
+ J2 )(J3 + J4 )/162 + J3 (J1 + J2 )2 /108], L3 =(J3 +J4 )2 fault location immune to saturation of current transformers,” in Proc. 8th
2
(J1 −J2 )/81−[(4J1 −J2 )(J3 + J4 ) + 4(2J3 − J4 ) (J3 + J4 ) Int. Conf. Develop. Power Syst. Protection, 2004, pp. 172–175.
[14] J. Izykowski, R. Molag, E. Rosolowski, and M. M. Saha, “Accurate
(J1 + J2 )]/162 + [J1 (J3 + J4 )2 /162 + J3 (J1 + J2 )(J3 +J4 ) location of faults on power transmission lines with use of two-end un-
/27], and L4 = (J3 − J4 )(J3 + J4 )2 /27. synchronized measurements,” IEEE Trans. Power Del., vol. 21, no. 2,
Therefore, Ykey can be obtained from (A6) by noniteratively pp. 627–633, Apr. 2006.
[15] J. Izykowski, E. Rosolowski, P. Balcerek, M. Fulczyk, and M. M.
solving the polynomial equation of degree 3. Similarly, the Saha, “Accurate noniterative fault-location algorithm utilizing two-end
correct solution can also be obtained by checking the order of unsynchronized measurements,” IEEE Trans. Power Del., vol. 25, no. 1,
magnitude and the polarity of the value. Afterwards, Zkey can pp. 72–80, Jan. 2010.
[16] S. Azizi, M. Sanaye-Pasand, and M. Paolone, “Locating faults on un-
be calculated by substituting Ykey into (A4). transposed, meshed transmission networks using a limited number of
During the fault, for (A3) as shown at the bottom of the next synchrophasor measurements,” IEEE Trans. Power Syst., vol. 31, no. 6,
page, similarly, the difference between the 1st and the 3rd row pp. 4462–4472, Nov. 2016.
[17] X. Jiao and Y. Liao, “Accurate fault location for untransposed/Transposed
can be rewritten as, transmission lines using sparse wide-area measurements,” IEEE Trans.
Power Del., vol. 31, no. 4, pp. 1797–1805, Aug. 2016.
a1 x6 + b1 x5 + c1 x4 + d1 x3 + e1 x2 + f1 x + g1 = 0 (A7) [18] Y. Liu, B. Wang, X. Zheng, D. Lu, M. Fu, and N. Tai, “Fault loca-
tion algorithm for non-homogeneous transmission lines considering line
where a1 = Zkey 3 Ykey 3 (K1 −K2 )/1458, b1 = Zkey 3 Ykey 2 asymmetry,” IEEE Trans. Power Del., vol. 35, no. 5, pp. 2425–2437,
[Ykey K2 − (K3 + K4 )]/243, c1 = Zkey 2 Ykey 2 (K1 − K2 )/27 Oct. 2020.
[19] Y. Liao and S. Elangovan, “Unsynchronised two-terminal transmission-
+ 5Zkey 3 Ykey 2 (2K4 − Ykey K2 )/486, d1 = 10Zkey 3 Ykey 2 line fault location without using line parameters,” IEE Proc. Gener. Trans-
2
(Ykey K2 − 3K4 )/729 + 4Zkey Ykey (Ykey K2 −K3 −K4 )/27, miss. Distrib., vol. 153, no. 6, pp. 639–643, Nov. 2006.
e1 = Zkey Ykey (K1 − K2 )/2+5Zkey 3 Ykey 2 (4K4 −Ykey K2 )/ [20] Y. Liao and N. Kang, “Fault-location algorithms without utilizing line
parameters based on the distributed parameter line model,” IEEE Trans.
486 + 2Zkey 2 Ykey ·[2K4 − Ykey K2 ]/9, f1 = Zkey 3 Ykey 2 Power Del., vol. 24, no. 2, pp. 579–584, Apr. 2009.
(Ykey K2 − 5K4 )/243 + 4Zkey 2 Ykey (Ykey ·K2 − 3K4 )/27 + [21] C. Apostoloppulos and G. N. Korres, “A novel algorithm for locating
Zkey (Ykey K2 − K3 − K4 ), g1 = K1 − K2 + Zkey 3 Ykey 2 faults on transposed/untransposed transmission lines without utilizing line
parameters,” IEEE Trans. Power Del., vol. 25, no. 4, pp. 2328–2338,
(6K4 −Ykey K2 )/1458 + Zkey 2 Ykey (4K4 − Ykey K2 )/27 − Oct. 2010.
Zkey (Ykey K2 /2 + K4 ). [22] G. Preston, Z. Radojevic, C. H. Kim, and V. Terzija, “New settings-free
fault location algorithm based on synchronised sampling,” IET Gener.
One can observe that the highest order of x in (A7) is 6 = 2n. Transmiss. Distrib., vol. 5, no. 3, pp. 376–383, 2011.
The polynomial is indeed without noniterative solutions. [23] D. A. G. Vieira, D. B. Oliveira, and A. C. Lisboa, “A closed-form
solution for transmission-line fault location without the need of terminal
synchronization or line parameters,” IEEE Trans. Power Del., vol. 28,
REFERENCES no. 2, pp. 1238–1239, Apr. 2013.
[24] V. Terzija, Z. M. Radojevic, and G. Preston, “Flexible synchronized
[1] A. A. Girgis, D. G. Hart, and W. L. Peterson, “A new fault location
measurement technology-based fault locator,” IEEE Trans. Smart Grid,
technique for two- and three-terminal lines,” IEEE Trans. Power Del.,
vol. 6, no. 2, pp. 866–873, Mar. 2015.
vol. 7, no. 1, pp. 98–107, Jan. 1992.
[25] A. S. Dobakhshari, “Noniterative parameter-free fault location on untrans-
[2] F. H. Magnago and A. Abur, “Fault location using wavelets,” IEEE Trans.
posed single-circuit transmission lines,” IEEE Trans. Power Del., vol. 32,
Power Del., vol. 13, no. 4, pp. 1475–1480, Oct. 1998.
no. 3, pp. 1636–1644, Jun. 2017.
[3] M. Kezunovic, “Smart fault location for smart grids,” IEEE Trans. Smart
[26] Communication Networks and Systems in Substations, IEC Standard
Grid, vol. 2, no. 1, pp. 11–22, Mar. 2011.
61850, 2003.
[4] Y. Liu, A. P. Meliopoulos, Z. Tan, L. Sun, and R. Fan, “Dynamic state
[27] IEEE Standard for Synchrophasor Data Transfer for Power Systems, IEEE
estimation-based fault locating on transmission lines,” IET Gener. Trans-
Standard C37.118.2-2011, 2011.
miss. Distrib., vol. 11, no. 17, pp. 4184–4192, Nov. 2017.

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.
LU et al.: IMPROVED NONITERATIVE PARAMETER-FREE FAULT LOCATION METHOD ON UNTRANSPOSED 1369

[28] N. I. Elkalashy, “Simplified parameter-less fault locator using double- Shi Chen received the B.S. degree in electrical engi-
end synchronized data for overhead transmission lines,” Int. Trans. Elect. neering from Southeast University, Nanjing, China,
Energy Syst., vol. 24, no. 6, pp. 808–818, 2014. in 2012 and the Ph.D. degree in electrical engineer-
[29] M. H. Hesse and J. Sabath, “EHV double-circuit untransposed transmis- ing from Shanghai Jiao Tong University, Shanghai,
sion line-analysis and tests,” IEEE Trans. Power App. Syst., vol. PAS-90, China, in 2017. He is currently a Senior Engineer with
no. 3, pp. 984–992, May 1971. State Grid Jiangsu Electric Power Company Research
[30] S. Kang, Y. Ahn, Y. Kang, and S. Nam, “A fault location algorithm based on Institute. His research interests include the protection
circuit analysis for untransposed parallel transmission lines,” IEEE Trans. and control of renewable energy and protection of
Power Del., vol. 24, no. 4, pp. 1850–1856, Oct. 2009. high-voltage power systems.
[31] Q. Ma, B. Zheng, L. Ban, and Z. Xiang, “Secondary arc current anal-
ysis of an untransposed EHV/UHV transmission line with controllable
unbalanced shunt reactor,” IEEE Trans. Power Del., vol. 30, no. 3,
pp. 1458–1466, Jun. 2015.

Binglin Wang (Student Member, IEEE) received the


B.S. degree in electrical engineering and intelligent
control from the Xi’an University of Technology,
Dian Lu (Student Member, IEEE) received the B.S. Xi’an, China, in 2018. He is currently working to-
degree in electrical engineering and automation from ward the Ph.D. degree in electrical engineering with
the Huazhong University of Science and Technology, the School of Information Science and Technology,
Wuhan, China, in 2017. He is currently working to- ShanghaiTech University, Shanghai, China. His re-
ward the Ph.D. degree in electrical engineering with search interests include protection, fault location, and
the School of Information Science and Technology, state estimation of HVAC and HVDC transmission
ShanghaiTech University, Shanghai, China. His re- lines.
search interests include phasor domain protection and
fault location of transmission lines.

Dayou Lu (Student Member, IEEE) received the B.S.


degree in electrical engineering and automation from
Yu Liu (Senior Member, IEEE) received the B.S. and the Huazhong University of Science and Technology,
M.S. degrees in electrical power engineering from Wuhan, China, in 2017. He is currently working to-
Shanghai Jiao Tong University, Shanghai, China, in ward the Ph.D. degree in electrical engineering with
2011 and 2013, respectively, and the Ph.D. degree in the School of Information Science and Technology,
electrical and computer engineering from the Georgia ShanghaiTech University, Shanghai, China. His re-
Institute of Technology, Atlanta, GA, USA, in 2017. search interests include modeling, protection, and
He is currently a Tenure-Track Assistant Professor fault location of transmission lines.
with the School of Information Science and Technol-
ogy, ShanghaiTech University, Shanghai, China. He
has authored or coauthored two book chapters and
more than 60 technical papers. His research interests
include modeling, protection, fault location, and state or parameter estimation
of power systems and power electronic systems.

Authorized licensed use limited to: ABB Power Grids. Downloaded on May 27,2022 at 04:53:09 UTC from IEEE Xplore. Restrictions apply.

You might also like