Received: 21 February 2020 Revised: 11 September 2020 Accepted: 14 September 2020

DOI: 10.1002/2050-7038.12658

RESEARCH ARTICLE

Wide area impedance-based fault location using

synchronized/unsynchronized current measurements

Zahra Shafiei Chafi | Hossein Afrakhte

Department of Electrical Engineering,

Faculty of Engineering, University of
Summary
Guilan, Rasht, Iran Global positioning system (GPS) as an advanced technology provides synchronized
measurements through the network. In spite of enhancing measurement accuracy
Correspondence
Hossein Afrakhte, Department of provided by GPS, losing its signal, using slow communication link for measurement
Electrical Engineering, Faculty of or lack of essential infrastructure for synchronized sampling make the synchronized
Engineering, University of Guilan, Rasht
measurement unavailable across the grid. In these conditions some fault location
41996-13776 Iran.
Email: [email protected] algorithms are not applicable and some others eliminate the corresponding measure-
ments in their fault location process. However valuable information lies in the elimi-
Handling Editor: Reza Jalilzadeh-
Hamidi
nated measurements magnitude which can improve the fault location accuracy. In
this paper a system of equations is developed based on magnitudes of all current
measurements and phase angle of synchronized current measurements. Then it is
solved using an iterative approach with the aim of finding the location of the fault.
To obtain more accurate results the distributed parameter model of line is used. The
proposed method is validated by WSCC 9 bus and IEEE 39 bus test systems. The
results reveal accurate estimation even in conditions that the synchronized measure-
ments are not available in the grid entirely.

KEYWORDS
current measurements, distributed parameter model of line, fault location, synchronized
measurements, unsynchronized measurements, wide area

1 | INTRODUCTION

Occurring fault is unavoidable in power systems. Subsequent to a fault partial outage or in extreme case total failure of
power system occurs.1 To evade and reduce outages and failure effects caused by fault for consumers and power system
equipment, protection system isolates the faulted section in a confined area immediately.2 In this condition accurate

List of symbols and abbreviations: x, fault distance measured from bus i; Zc, faulted line characteristic impedance; Zc_ab, healthy line characteristic
impedance; γ ab, propagation constant of healthy line a-b; lab, healthy line length; γ ij, propagation constant of faulted line; lij, faulted line length; If,
fault current at fault point; Zjj, j-j element of reduced bus impedance matrix; Zij, i-j element of reduced bus impedance matrix; Zii, i-i element of
reduced bus impedance matrix; Z 0ii , i-i element of original bus impedance matrix; Z 0ij , i-j element of original bus impedance matrix; Z 0ji , j-i element of
original bus impedance matrix; Z 0jj , j-j element of original bus impedance matrix; ΔIk, difference between pre- and post-fault current of healthy line in
branch number k; ΔIij, difference between pre- and post-fault current of faulted line in its sending bus; ΔIji, difference between pre- and post-fault
currents in faulted line in its receiving bus; Zia, i-a element of original bus impedance matrix; Zja, j-a element of original bus impedance matrix; Zib, i-
b element of original bus impedance matrix; Zjb, j-b element of original bus impedance matrix; m, number of current measurements; n, number of
synchronized current measurements; i,j, indices of faulted line; a,b, indices of healthy line; k, branch between buses a and b.

Int Trans Electr Energ Syst. 2020;e12658. wileyonlinelibrary.com/journal/etep © 2020 John Wiley & Sons Ltd 1 of 18
https://doi.org/10.1002/2050-7038.12658
2 of 18 SHAFIEI CHAFI AND AFRAKHTE

and swift fault location expedites the repairing and restoration processes. Therefore, by reducing the repair and outage
times the system reliability will be improved.3
A lot of efforts have been devoted to fault location up to now. There are two main groups for fault location:
traditional and wide area. Traditional fault location methods are classified into three groups. The first category
is impedance-based method which uses the fundamental frequency component of measured voltage and current
signals. 4-12 The second group named traveling wave based methods uses traveling wave generated by fault to
pinpoint the fault location 13-18 and the last one is artificial intelligence methods uses neural networks and fuzzy
logic systems for determining the location of the fault. 19-23 Traditional methods demand measurements from
one end or both ends of the faulted line (or from all terminals of multiterminal lines). From practical aspects
due to economical and/or technical reasons the number of measurements is limited in the network. On the
other hand, the faulted line may not be equipped with any measuring devices. In these conditions the tradi-
tional fault location methods may not be applicable. Wide area fault location methods propose solution that
finds the location of the fault with restricted number of measurements that are not obtained from the faulted
line directly. These methods are classified into two groups: Traveling wave based 24-28 and impedance based. 29-39
Traveling wave based methods require high sampling rates for measurements whereas impedance-based
methods need much lower rates. 38 A comparison between traditional and wide area fault location methods can
be found in Table 1.
With the advent of smart grids gathering and metering technology enhanced by wide area measurement sys-
tems. Phasor measurement units (PMUs) as the most important part of these systems play a key role in record-
ing and transmitting data across the grid for different purposes including fault location which impedance-based
wide area fault location methods rely on GPS-synchronized measurements obtained by these units. 29,39 An
important issue about synchronized measurement devices is the GPS signal reliability. There are disparate fac-
tors like atmospheric disturbances, GPS antenna failure, electromagnetic interference, weather change, GPS
signal attack, or solar activity which are uncontrollable and unpredictable and may cause GPS receivers lose
signal occasionally. 40 In confronting with GPS signal loss, the proposed procedure of Reference 41 will be vain.
In References 29-32, 34, 37, 42, and 43, when the GPS-synchronizing signal is not available in some network
buses the acquired measurements do not participate in calculations and the fault is located using the remained
synchronized measurements. Utilizing the magnitude of measurement instead of synchronizing angle which is
absent 33,35,44,45 presented methods in order to overcome the above mentioned issue. Pursuant to worthwhile
information exists in measurements phase angle, 38 put forward a method uses all voltage measurements magni-
tude and synchronized voltage measurements phase angle. However, current measurements give further valu-
able information about the grid. When voltage transformers of some buses are not in service, current
transformers of related buses contribute significantly to fault location issue. Moreover, in a substation there is
only one voltage measurement but multiple current measurements are obtainable due to multiplicity of lines
connected to the substation. 34 Hence, using this sort of measurement provide a lot of information that can con-
tribute to raise the fault location accuracy.

TABLE 1 comparison between traditional and wide area FL methods

Method Advantage Disadvantage

Traditional • Usually does not use complex devices and has easy • Requires that one end or both ends of the faulted line
implementation. be equipped with measuring device
• Has cheaper implementation • Modeling the faulted line from its terminals and
updating that, are necessary in some methods
• Failure or unavailability of measuring devices make
some methods futile
• Uses the measurements that are close to the FL and
so they are affected by the transformers transients
Wide area • Uses the restricted number of measurements • Usually complex infrastructure is required.
• The measurements are not close to FL and are not affected • Its implementation is expensive
by errors of instrument transformers like CT saturation
SHAFIEI CHAFI AND AFRAKHTE 3 of 18

This article holds forth an impedance-based wide area fault location method incorporating synchronized/
unsynchronized current measurements. In contrary to Reference 33 that used either synchronized or unsynchronized
current measurements the proposed method integrates all available synchronized and unsynchronized current mea-
surements which can also improve the fault location accuracy. Synchronized measurements are incorporated in formu-
lation by their phase angle and magnitude and unsynchronized measurements take part with their magnitudes because
of GPS signal unavailability. Using these measurements, a system of equations is built. Since this system involves
nonlinear equations it solves using Gauss-Newton method. The rest of this paper is organized that in section 2 the pro-
posed fault location formulation is demonstrated, in section 3 evaluating the proposed method is done and eventually,
section 4 is devoted to conclusion.

2 | P R O P O S ED M E T H O D F O R M U L A T I O N

In this subsection the intention is to locate the fault considering current measurements and their circuit equations.
Figure 1 shows the positive sequence network depiction in which the lines are considered with their distributed param-
eter model. A faulted network is shown in (a) where a fault has been occurred at distance x from bus i along the line i-j.
The fault current has been considered a current source which is placed at distance x from bus i. In (b) a positive
sequence network representation with a healthy line a-b is revealed.
According to Figure 1B the bus voltage changes can be shown as (1)46

ΔU a = U pre
a −U a
post
af ðx ÞI f
= ZM ð1Þ

44
where Z M
af is the a-f element of the modified positive sequence bus impedance matrix (see (2)).

Z ia Z
sinhðσ ij lij x Þ
+ sinh σ ljað1 − xÞ
ð ij ij Þ
df ðx Þ =
ZM

 ð2Þ
1
sinhðσ ij lij x Þ
+ sinh 1
+ tanh 2 γ ab lab x
+ tanh γ ab lab2ð1 − x Þ
ðσij lij ð1− x ÞÞ

After simplification, (2) can be rewritten as (3) where f1(x) and f2(x) are the coefficients of Zia and Zja respectively.

af ðx Þ = Z ia f 1 ðx Þ + Z ja f 2 ðx Þ
ZM ð3Þ

FIGURE 1 Positive sequence network model (A) Faulted line model in a network (B) Healthy line model in a network
4 of 18 SHAFIEI CHAFI AND AFRAKHTE

 
sinh σ ij lij ð1 −x Þ
f 1 ðx Þ =   ð4Þ
sinh σ ij lij

 
sinh σ ij lij x
f 2 ðx Þ =   ð5Þ
sinh σ ij lij

substituting (3) in (1) results (6).

 
ΔU a = Z ia f 1 ðx Þ + Z ja f 2 ðx Þ I f ð6Þ

With writing KCL (Kirchhoff's Current Law) for bus a the Equations (7) and (8) are yield for pre- and post-fault con-
ditions where I pre post
ab andI ab are pre- and post-fault current of the line a-b respectively.

  pre
U pre γ ab lab a −U b
U pre
I pre = a
tanh + ð7Þ
ab
Z c_ab 2 Z c_ab sinhðγ ab lab Þ

  post
U post γ ab lab a − Ub
U post
I post = a
tanh + ð8Þ
ab
Z c_ab 2 Z c_ab sinhðγ ab lab Þ

Calculating the difference between (7) and (8) leads to (9).

 
ΔU a γ ab lab ΔU a −ΔU b
ΔI k = I pre post
ab − I ab = tanh + ð9Þ
Z c_ab 2 Z c_ab sinhðγ ab lab Þ

where, ΔUaand ΔUb are the differences between pre- and post-fault voltages of buses a and b respectively. Now by
replacing (6) in (9) for buses a and b, (10) is yield.
With regard to (11) and (12), (10) can be written in the concise form of (13).
    
1 γ ab lab 1 1
ΔIk = tanh + Z ia − Z ib f 1 ðx ÞI f
Z c_ab 2 sinhðγ ab lab Þ Z c_ab sinhðγ ab lab Þ
    
1 γ lab 1 1
+ tanh ab + Z ja − Z jb f 2 ðx ÞI f ð10Þ
Z c_ab 2 sinhðγ ab lab Þ Z c_ab sinhðγ ab lab Þ

   
1 γ ab lab 1 1
Ak = tanh + Z ia − Z ib ð11Þ
Z c_ab 2 sinhðγ ab lab Þ Z c_ab sinhðγ ab lab Þ

   
1 γ ab lab 1 1
Bk = tanh + Z ja − Z jb ð12Þ
Z c_ab 2 sinhðγ ab lab Þ Z c_ab sinhðγ ab lab Þ

ΔI k = Ak f 1 ðx ÞI f + Bk f 2 ðx ÞI f ð13Þ

In Figure 1A if bus i is equipped with measuring device, (14) is used instead of (13).

ΔI ij = C 1 f 1 ðx ÞI f + D1 f 2 ðx ÞI f ð14Þ

where33
SHAFIEI CHAFI AND AFRAKHTE 5 of 18

! !
Z jj Z 0ii Z ij Z 0ij
C1 = − ð15Þ
Z 2ij −Z jj Z ii Z 2ij −Z jj Z ii

! !
Z jj Z 0ji Z ij Z 0jj
D1 = − ð16Þ
Z 2ij −Z jj Z ii Z 2ij −Z jj Z ii

and if there is a measuring device on bus j, (17) is used instead of (13).

ΔI ji = C 2 f 1 ðx ÞI f + D2 f 2 ðx ÞI f ð17Þ

where33
! !
Z jj Z 0ji Z ij Z 0jj
C2 = − ð18Þ
Z 2ij −Z jj Z ii Z 2ij −Z jj Z ii

! !
Z ii Z 0ji Z ij Z 0jj
D2 = − ð19Þ
Z 2ij −Z jj Z ii Z 2ij −Z jj Z ii

Now, getting the magnitude on both sides of (13) and (14) results in (20) and (21).

jΔI k j = jAk f 1 ðx Þ + Bk f 2 ðx Þj I f ð20Þ

ΔI ij = jC1 f 1 ðx Þ + D1 f 2 ðx Þj I f ð21Þ

Equations (20) and (21) can be employed for all measurement equipped buses regardless of their synchronization.
Calculating the phase angle for both hands of (13) and (14) gives (22) and (23). In contrast to (20) and (21) that
employed for all measurements, (22) and (23) are used in situations that a synchronized measurement is done (synchro-
nized measurements are available).

∡ΔIk = ∡ðAk f 1 ðxÞ + Bk f 2 ðxÞÞ + ∡If ð22Þ

∡ΔIij = ∡ðCf 1 ðxÞ + Df 2 ðxÞÞ + ∡If ð23Þ

Considering (20)-(23) for all accessible measurements yields the equation set of (24). According to (24), the magni-
tude and phase angle of measurements for synchronized measurements and the only magnitude of measurements for
unsynchronized measurements are included. Besides, if a bus does not have any measurement devices the
corresponding measurements of this bus are excluded from (24). Similarly, if there are no measuring devices on one
end or both ends of the faulted line the related measurement will be omitted from (24).
The equations set of (24) can be rewritten in form of (25). Where P denotes the measurement vector, Q is the func-
tion of system unknowns and X = [x, |If|, ∡If]T is the vector of unknown variables. Because of nonlinearity involved in
(24) it can be solved utilizing weighted least squares (WLS) method47 and by methods like Gauss-Newton. The Gauss-
Newton corresponding terms are as (26) and (27), where J is calculated according to J = ∂Q∂XðX t Þ
(see (28)) is the Jacobian
matrix, R is the measurement covariance matrix and t is the counter of iterations. The nonzero components of J can be
obtained as (29)-(40). The derivatives corresponding to measurements magnitudes are calculated as (29)-(34).
6 of 18 SHAFIEI CHAFI AND AFRAKHTE

2 3
jΔI 1 j 2 3
6 jΔI j 7 A1 f 1 ðx ÞI f + B1 f 2 ðx ÞI f
6 2 7
6 7 6 7
6 .. 7 6 A2 f 1 ðxÞI f + B2 f 2 ðx ÞI f 7
6 . 7 6 7
6 7 6 .. 7
6 7 6 7
6 jΔI m j 7 6 . 7
6 7 6 7
6 7 6 7
6 ΔI 7 6 Am f 1 ðx ÞI f + Bm f 2 ðx ÞI f 7
6 ij 7 6 7
6 7 6 7
6 7 6 C1 f 1 ðx ÞI f + D1 f 2 ðx ÞI f 7
6 7 6 7
6 ΔI ji 7 6 7
6 7 6 7
6 7 6 C2 f 1 ðx ÞI f + D2 f 2 ðx ÞI f 7
6 7 6 7
6 ∡ΔI1 7 6 7
6 7 6 ∡ðA1 f 1 ðxÞIf + B1 f 2 ðxÞIf Þ 7
6 7 6 7
6 7=6 7 ð24Þ
6 ∡ΔI 7 6 7
6 2 7 6 ∡ðA2 f 1 ðxÞIf + B2 f 2 ðxÞIf Þ 7
6 7 6 7
6 7 6 7
6 7 6 7
6 . 7 6 . 7
6 .. 7 6 .. 7
6 7 6 7
6 7 6 7
6 7 6 7
6 7 6 7
6 ∡ΔIn 7 6 ∡ðAn f 1 ðxÞIf + Bn f 2 ðxÞIf Þ 7
6 7 6 7
6 7 6 7
6 7 6 7
6 7 6 ∡ðC1 f 1 ðxÞIf + D1 f 2 ðxÞIf Þ 7
6 ∡ΔIij 7 6 7
6 7 6 7
6 7 4 5
6 7
4 5 ∡ðC2 f 1 ðxÞIf + D2 f 2 ðxÞIf Þ
∡ΔIji |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
|fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} QðX Þ
P

P −QðX Þ = 0 ð25Þ

 −1 T −1
ΔX = J T R − 1 J J R ðP −QðX t ÞÞ ð26Þ

X t + 1 = X t + ΔX ð27Þ

" #
H k,2 ½0ðm + 1Þ × 1
J = H k,1 ð28Þ
½0n × 1 H k,3

∂ jΔI k j If ∂ 
H k,1 = = × ½ðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ Ak f 1 ðx Þ + Bk f 2 ðx ÞÞ * ð29Þ
∂x 2jAk f 1 ðx Þ + Bk f 2 ðx Þj ∂x

where k = 1, 2,…, m.

∂ ΔI ij If ∂h i
H k,1 = = × ð C 1 f 1 ð x Þ + D1 f 2 ð x Þ Þ ð C 1 f 1 ð x Þ + D1 f 2 ð x Þ Þ * ð30Þ
∂x 2jC1 f 1 ðx Þ + D1 f 2 ðx Þj ∂x

where k = m + 1.

∂ ΔI ji If ∂h i
H k,1 = = × ð C 2 f 1 ð x Þ + D2 f 2 ð x Þ Þ ð C 2 f 1 ð x Þ + D2 f 2 ð x Þ Þ * ð31Þ
∂x 2jC2 f 1 ðx Þ + D2 f 2 ðx Þj ∂x

where k = m + 2.
SHAFIEI CHAFI AND AFRAKHTE 7 of 18

 
∂ jΔI k j ∂ jAk f 1 ðx Þ + Bk f 2 ðx Þj I f
H k,2 = = = j A k f 1 ðx Þ + B k f 2 ðx Þj ð32Þ
∂ If ∂ If

where k = 1, 2,…, m.
 
∂ ΔI ij ∂ j C 1 f 1 ð x Þ + D1 f 2 ð x Þj I f
H k,2 = = = j C 1 f 1 ð x Þ + D1 f 2 ð x Þ j ð33Þ
∂ If ∂ If

where k = m + 1.
 
∂ ΔI ji ∂ j C 2 f 1 ðx Þ + D 2 f 2 ðx Þj I f
H K,2 = = = j C 2 f 1 ð x Þ + D2 f 2 ð x Þj ð34Þ
∂ If ∂ If

where k = m + 2.
And the derivatives related to measurements phase angle are as (35)-(40).


ImðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ
∂∡ΔIk ∂∡ðAk f 1 ðxÞ + Bk f 2 ðxÞÞ 1 ∂ ReðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ
H k,1 = = =
2 × ð35Þ
∂x ∂x ImðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ ∂x
1+ ReðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ

where k = m + 3, m + 4,…, m + n + 3.


ImðC1 f 1 ðx Þ + D1 f 2 ðx ÞÞ
∂∡ΔIij ∂∡ðC1 f 1 ðxÞ + D1 f 2 ðxÞÞ 1 ∂ ReðC 1 f 1 ðx Þ + D1 f 2 ðx ÞÞ
H k,1 = = =
2 × ð36Þ
∂x ∂x ImðC 1 f 1 ðx Þ + D1 f 2 ðx ÞÞ ∂x
1+ ReðC1 f 1 ðx Þ + D1 f 2 ðx ÞÞ

where k = m + n + 4.


ImðC2 f 1 ðx Þ + D2 f 2 ðx ÞÞ
∂∡ΔIji ∂∡ðC2 f 1 ðxÞ + D2 f 2 ðxÞÞ 1 ∂ ReðC 2 f 1 ðx Þ + D2 f 2 ðx ÞÞ
H k,1 = = =
2 × ð37Þ
∂x ∂x ImðC 2 f 1 ðx Þ + D2 f 2 ðx ÞÞ ∂x
1+ ReðC2 f 1 ðx Þ + D2 f 2 ðx ÞÞ

where k = m + n + 5.

∂∡ΔIk ∂ ð∡ðAk f 1 ðxÞ + Bk f 2 ðxÞÞ + ∡If Þ

H k,3 = = =1 ð38Þ
∂∡If ∂∡If

where, k = m + 3, m + 4,…, 2m + 2.

∂∡ΔIij ∂ ð∡ðC1 f 1 ðxÞ + D1 f 2 ðxÞÞ + ∡If Þ

H k,3 = = =1 ð39Þ
∂∡If ∂∡If

where k = 2m + 3.

∂∡ΔIji ∂ ð∡ðC2 f 1 ðxÞ + D2 f 2 ðxÞÞ + ∡If Þ

H k,3 = = =1 ð40Þ
∂∡If ∂∡If
where k = 2m + 4.
8 of 18 SHAFIEI CHAFI AND AFRAKHTE

In calculating the J matrix elements, (29)-(31) and (35)-(37) need a little more simplification. The final terms are
mentioned here and the simplification procedure can be found in Appendix A.

3 | EVALUATION STUDIES

To appraise the performance of the presented method WSCC 9 bus and IEEE 39 bus test systems are chosen which their
parameters can be found in Reference 47 and their single-line diagrams are shown in Figures 2 and 3 respectively. In
order to model and simulate the test systems DigSILENT and MATLAB software are used in this paper. The sampling
is done using a high sampling rate(100 KHz). To reject high frequency harmonics involved in the obtained waveforms,
an antialiasing filter which is considered a second order Butterworth filter with cut-off frequency of 350 Hz is used.
Next the output of this filter is down sampled to 4800 Hz to be similar to sampling rate of the PMUs. Then the signals
are passed through a least squares DC removal filter. Finally, in order to extract the fundamental frequency component
involved in the signals discrete Fourier transform (DFT) is applied to them. The magnitude and phase angle of the
phasors utilized in this study are attained by averaging them over an interval before protection system operation. Fig-
ures 4 and 5 illustrate current phasor extraction steps that are used after occurring a fault in WSCC 9 bus test system.

FIGURE 2 Single-line representation of WSCC 9 bus test

system

F I G U R E 3 Single-line
representation of IEEE 39-bus test
system
SHAFIEI CHAFI AND AFRAKHTE 9 of 18

Relative difference between estimated and actual fault location (FL) (see (41)48) is employed as an index to measure
the accuracy of the proposed method.

Estimated FL −Actual FL
FL estimation errorð%Þ = × 100 ð41Þ
Faulted line length

The flowchart of the proposed algorithm can be found in Figure 6. At the first step the current signals are gathered
using the measuring devices that exist in the network beforehand. In the next part the signal processing is done to
extract the phasor representation of the input signals for pre- and post-fault conditions(here a cycle interval which
begins two cycle before and after fault inception are considered for pre- and post-fault phasor calculation respectively)
which the phasor extraction process has been described in the first paragraph of this section (section 3) in details (ie,
applying Butterworth filter, down sampling, DC removal filter, and DFT respectively).After that the output phasors
enter to the Gauss-Newton part. At the beginning the equation set of (24) is formed using (20)-(23), (11)-(12), (15)-(16)
and (18)-(19) which the magnitude and phase angle of the differences between pre- and post-fault extracted phasors in
previous step are considered on their left hand side of(20)-(23).If the synchronized measurement is not possible in a
bus, the angle related rows in (24) are omitted. Then the Jacobian matrix (see (28)) elements are calculated using (29)-
(40) which (11)-(12), (15)-(16), and (18)-(19) are again used in them. In this way the Gauss-Newton can be applied
according to (26)-(27) and its iterative process of calculating the unknown vector (X) is repeated until the error of its
first element (x) which is defined the FL between two consecutive iteration exceeds a predefined value which usually is
considered in range of 1e-3 to 1e-6.

F I G U R E 4 Current waveform of line 7-8 for phase a- A SLG

fault at 50% of the line 7-5 length in 0 sec has been occurred and
has been cleared in 0.06 second

F I G U R E 5 Current waveforms of
line 9-6- A SLG fault at 50% of the line
7-5 length in 0 sec has been occurred
and has been cleared in 0.06 second
(in this Figure the phase current
waveforms are the output of the least
squares filter)
10 of 18 SHAFIEI CHAFI AND AFRAKHTE

F I G U R E 6 Flow Chart of the

proposed method

3.1 | Small scaled network: WSCC 9 bus test system

The single-line diagram of this system is depicted in Figure 2. This system comprises three generators, three trans-
formers, and six transmission lines. It is assumed that PMUs are on buses number 4,7, and 9.39
Table 2 shows the fault locating results for different fault types at 0.5 p.u. of the lines length. It is evident that the
fault position is estimated correctly irrespective of fault type. It can be seen that the maximum error does not exceed
to 0.5%.
Table 3 presents the effect of the fault distance on the proposed algorithm. Different fault types at the middle (50%)
and the end (90%) of the line 7-5 are considered. It can be seen that the position of the fault can be pinpointed accu-
rately. The estimation error is below 0.1%.
SHAFIEI CHAFI AND AFRAKHTE 11 of 18

TABLE 2 FL results for different

FL estimation error (%) Fault type
fault types
Faulted line 3PH LL SLG LLG
7-5 −0.0203 −0.0820 −0.1078 −0.0279
9-8 0.0950 0.1617 −0.0317 0.0920

TABLE 3 The effect of fault

FL estimation error (%) fault type
distance
Faulted line Fault distance (p.u) 3PH LL SLG LLG
7-5 0.5 −0.0203 −0.0820 −0.1078 −0.0279
0.9 0.0712 −0.0606 −0.0933 −0.0571

T A B L E 4 FL results for GPS signal

FL estimation error (%) fault type
loss in bus number 4.
Faulted line 3PH LL SLG LLG
7-5 −0.0932 −0.3979 −0.0683 −0.0755
9-8 0.0949 0.0331 0.0818 0.0920

TABLE 5 The effect of fault

Estimation error (%) fault type
resistance.
Faulted line 3PH LL SLG LLG
7-5 −0.0891 0.0596 −0.0383 −0.0462
7-8 0.1214 −0.3357 x0.4297 0.0703
9-8 −0.0890 0.0079 −0.0050 0.0341

Table 4 tabulates the impact of GPS signal loss on the proposed method. It is assumed that the losing GPS synchro-
nization signal is occurred in bus number 4 and all types of the fault at 0.5 p.u. of each line length is studied. In this
condition the phase angle of current measurements that measured at this bus (bus number 4) do not take part in fault
locating procedure and only their magnitudes are used. Also, the magnitude and phase angle of measurements in buses
number 7 and 9 take part in FL formulation. It can be seen that in spite of GPS synchronization signal loss, the FL is
determined successfully.
The fault resistance is another substantial issue affects the FL estimation precision and is known as one of the error
origins in pinpointing the location of the fault. Table 5 shows the FL outcomes for different types of the fault at 0.5 p.u.
of the lines length when the fault resistance is high which here 50 Ω resistance is selected. It can be deduced that in
presence of high resistance the presented approach performs well.
For more investigation a SLG fault at 50% of line 7-5 length with different fault resistance values is considered. The
results are shown in Figure 7. It can be seen that the FL estimation error does not follow a monotonic pattern when the
resistance increases but the values are very close.
In order to show the advantage of the proposed approach over previous wide area FL methods an LLG fault with
50 Ω resistance is simulated at 90% of the length of the line 7-5 and it is assumed that the GPS signal loss is occurred on
the bus number 4. The result of FL is shown in Table 6. In this situation the proposed procedure in Reference 41 is
inapplicable and the algorithm in Reference 33 uses only the magnitude of the measurements when the GPS signal is
missed. FL result show that utilizing the magnitude of the measurements beside the phase angle of the synchronized
measurements can enhance the FL precision markedly.
More comparisons are reflected in the remained rows of the Table 6.The FL is done for an LLG fault at 10% of the
length of the line 7-5 considering the occurrence of GPS signal loss on the bus number 4. In this condition the proposed
method in Reference 41 which uses only the synchronized measurements is inapplicable and the method in Reference
12 of 18 SHAFIEI CHAFI AND AFRAKHTE

F I G U R E 7 The effect of fault

resistance on FL estimation error

TABLE 6 FL methods comparison

Method Actual fault distance (p.u.) Fault resistance (Ω) FL estimation error (%)
Proposed 0.9 50 −0.0021
Reference 33 0.08
Reference 41 —
Proposed 0.1 0 0.0125
Reference 38 −0.028
Reference 30 −0.08
Reference 41 —

TABLE 7 The effect of PMUs

Single phase to ground fault
configuration
Faulted line FL estimation error (%)
7-5 0.0769
7-8 −0.9661
5-4 −0.0879
4-6 −0.0543

38which considers the magnitude and phase angle of synchronized voltage measurements and the magnitude of
unsynchronized voltage measurements is slightly more inaccurate compared to the proposed method. As it can be
observed in Table 6, the multiplicity of the current measurements has increased the FL precision. Other FL approach
chosen to compare is Reference 30 which considers synchronized measurements and neglects the unsynchronized mea-
surements in its formulation when the GPS signal loss occurs. It is evident that taking into account the current phase
angle measurements rather than dismissing them the FL accuracy can be improved.
Table 7 tabulates the fault locating results with different measurements configuration proposed in Reference 30. In
Reference 30, it is considered that the measurements are situated on buses number 4 and 8 instead of buses number
4, 7, and 9. The SLG fault as the most common fault type in power system is considered at 50% of each line length. It
can be seen that the approximation error is under 1%.
In a number of references including Reference 29, it has been stated that with increasing in the number of PMUs
the FL error will reduce. Three PMUs are employed in the first configuration but in the second one two PMUs are
installed. This case can be investigated by comparing the results of Tables 2 and 7 for SLG fault on lines 7-5 and 7-8. It
clarifies that the results for three PMUs in Table 2 are more accurate than the two measurements in Table 7.
SHAFIEI CHAFI AND AFRAKHTE 13 of 18

3.2 | Large scaled network: IEEE 39-bus test system

To illustrate the efficacy of the presented method in a large scaled network IEEE 39-bus test system has been chosen.
As the Figure 3 depicts this system consists of 10 generators and 34 transmission lines. The lines are modeled using
their distributed parameters.
It is considered that PMUs are placed on buses number 2, 6, 10, 14, 17, 19, 22, 23, 25, 29, and 39.36
Table 8 reflects the FL results of IEEE 39 bus test system for different fault cases. Diverse fault types at various
places on each transmission line with different resistance values are regarded. The FL is performed utilizing synchro-
nized measurements obtained by installed PMUs in the network. It is obvious that the FL is approximated correctly at
each position of the lines. The supremacy of wide area algorithms also can be inferred from the Table 8 that the faults
on the faulted lines even with no measuring devices can be located prosperously. The relative error for all cases in
Table 8 has a plausible value.
The proposed method has the flexibility to locate the fault irrespective of synchronization. In other words, if
the synchronized measurement is not possible in some buses of the network or even in any of the buses, the pro-
posed method can effectively locate the fault. In order to evaluate the ability of the presented scheme in these
situations, two cases have been considered. At first it has been supposed that the GPS signal loss has been
occurred on buses number 14 and 29. Therefore, the synchronized measurement cannot be done in these two
buses and they take part in (24) only with their magnitude and the other measurement equipped buses take
part with their magnitude and phase angle. The next case is related to the conditions that none of the buses have
access to the synchronized measurement. So, the buses that can measure the current signals can be considered
in the proposed formulation by their magnitudes. The results of these two cases are shown in Table 9. Different

TABLE 8 FL results under different fault conditions

Faulted line Actual distance (p.u.) Fault resistance (Ω) Fault type FL estimation error (%)
One end PMU equipped lines
6-7 0.5 0 LL 0.0259
0.7 10 3PH −0.0002
10-13 0.9 20 3PH 0.0647
0.5 5 LLG −0.0429
14-15 0.1 50 SLG 0.0685
0.8 12 LL −0.0354
6-11 0.85 40 SLG 0.0074
0.1 2 LL 0.0498
16-17 0.6 3 LLG 0.0414
0.2 70 3PH 0.0352
Two ends PMU equipped line
2-25 0.3 5 LLG −0.7079
0.5 35 SLG −0.0893
16-24 0.4 25 LLG 0.0277
0.7 43 3PH −0.0236
No PMU equipped lines
3-18 0.25 100 3PH −0.0587
0.75 19 SLG 0.0081
8-9 0.9 31 3PH 0.0048
0.2 5 LL 2.0559
26-27 0.3 62 SLG 0.0642
0.6 12 LLG −0.0751
14 of 18 SHAFIEI CHAFI AND AFRAKHTE

TABLE 9 The effect of unsynchronized and partially synchronized measurements

FL estimation error (%)

Faulted Actual Fault Fault Partially synchronized Unsynchronized

line distance (p.u.) resistance (Ω) type measurements measurements
6-7 0.7 10 3PH −0.0021 0.0392
14-15 0.1 50 SLG 0.074 0.0687
2-25 0.3 5 LLG 0.036 −0.0594
3-18 0.75 19 SLG 0.005 0.0042
16-24 0.4 25 LLG −0.0011 0.0897

Fault inception angle (degree) FL estimation error (%)

TABLE 10 The effect of fault
inception angle
0 0.0142
45 −0.0374
90 −0.0321
101.25 0.0398
173.75 −0.0999

fault types with different resistance and positions have been considered. It can be observed that both cases can
pinpoint the FL with acceptable accuracy.
In order to show the effect of the fault inception angle on the proposed algorithm, a LL fault in the location of 20%
of the line 8-9 length with 5 Ω resistance at the various fault inception angles from 0 to about 180 has been applied.
The simulation results are reflected in Table 10. The results show appropriate accuracy for various fault inception
angles.
One of the main benefits of the wide area approaches is that they do not require to use measurements from the
faulted line in order to locate the fault. So, current transformer (CT) saturation has least effect on the presented
approach. In this paper the considered CT has an apparent power of 50 VA which has a piecewise linear saturation
model. With the noisy measurements as long as the sampling rate is in the range of KHz the estimation results will be
valid.49
It should be mentioned that although the uncertain zero sequence quantities are not employed but simulations
show that the proposed method is capable of pinpointing the FL and it is done regardless of the fault type.

4 | C ON C L U S I ON

In this paper a wide area impedance-based FL method based on synchronized and (or) unsynchronized current mea-
surements is proposed. This approach

• Integrates all synchronized and unsynchronized current measurements and therefore is immune of GPS signal loss.
• Make this possibility to locate the fault using only synchronized or unsynchronized current measurements or both
of them.
• Can improve the FL accuracy by valuable information provided by current measurements that exist according to
multiplicity of lines connected to a substation.
• Can improve the FL accuracy considering phase angle of current measurements in conditions that the GPS signal is
not available.
• Give a great contribution to FL problem in conditions that the voltage measurements are not available in the system
buses.
SHAFIEI CHAFI AND AFRAKHTE 15 of 18

• Improves the FL accuracy by integrating both synchronized and unsynchronized current measurements in contrast
to previous methods that considered either synchronized current measurements in all buses or unsynchronized cur-
rent measurements in them.

Several tests have been applied in order to evaluate the efficiency of the proposed method. It has been seen that it
has an acceptable performance in absence of GPS signal. Different fault types and resistance at various line positions
have been studied and it has been seen that the fault is pinpointed successfully in all cases. Using positive sequence net-
work in the proposed method provides this facility to find the location of faults without determining its type in advance.
The fault inception angle as the other test case has been examined. Also, the effect of PMU configuration on the pro-
posed procedure is evaluated and it has been observed that with decreasing in the number of measurements the preci-
sion of the estimation will be reduced.

P EE R R EV IE W
The peer review history for this article is available at https://publons.com/publon/10.1002/2050-7038.12658.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are openly available in (Online) http://www.pserc.cornell.edu/
matpower.47

ORCID
Hossein Afrakhte https://orcid.org/0000-0002-7377-5611

R EF E RE N C E S
1. Ali A, Khan AQ, Hussain B, Raza MT, Arif M. Fault modelling and detection in power generation, transmission and distribution sys-
tems. IET Gen Trans Distrib. 2015;9(16):2782-2791.
2. Korkali M, Abur A. Robust fault location using least-absolute-value estimator. IEEE Trans Power Syst. 2013;28(4):4384-4392.
3. Jiang Q, Wang B, Li X. An efficient PMU-based fault-location technique for multiterminal transmission lines. IEEE Trans Power Deliv.
2014;29(4):1675-1682.
4. Liao Y, Kezunovic M. Optimal estimate of transmission line fault location considering measurement errors. IEEE Trans Power Deliv.
2007;22(3):1335-1341.
5. Brahma SM. New fault-location method for a single multiterminal transmission line using synchronized phasor measurements. IEEE
Trans Power Deliv. 2006;21(3):1148-1153.
6. Izykowski J, Molag R, Rosolowski E, Saha MM. Accurate location of faults on power transmission lines with use of two-end
unsynchronized measurements. IEEE Trans Power Deliv. 2006;21(2):627-633.
7. Brahma SM. Fault location scheme for a multi-terminal transmission line using synchronized voltage measurements. IEEE Trans Power
Deliv. 2005;20(2):1325-1331.
8. de Morais Pereira CE, Zanetta LC. Fault location in transmission lines using one-terminal postfault voltage data. IEEE Trans Power
Deliv. 2004;19(2):570-575.
9. Brahma SM, Girgis AA. Fault location on a transmission line using synchronized voltage measurements. IEEE Trans Power Deliv. 2004;
19(4):1619-1622.
10. Popovic LM. A digital fault-location algorithm taking into account the imaginary part of the grounding impedance at the fault place.
IEEE Trans Power Deliv. 2003;18(4):1489-1496.
11. Kawady T, Stenzel J. A practical fault location approach for double circuit transmission lines using single end data. IEEE Trans Power
Deliv. 2003;18(4):1166-1173.
12. Saber A. A new fault location technique for four-circuit series-compensated transmission lines. Int Trans Electr Energy Syst. 2019;29(4):
e2791.
13. Aurangzeb M, Crossley PA, Gale P. Fault location using the high frequency travelling waves measured at a single location on a transmis-
sion line. 7th International Conference on Developments in Power Systems Protection (DPSP); 2001;403-406.
14. Zheng X, Tao Y. Analysis on using transmission lines traveling wave fault location based on fault information system. International Con-
ference on Advanced Power System Automation and Protection. 2011;2:1636-1640.
15. Lei A, Dong X, Shi S, Wang B. A novel current travelling wave based single-ended fault location method for locating single-phase-to-gro-
und fault of transmission line. 50th International Universities Power Engineering Conference (UPEC); 2015;1-6.
16. He Z, Li X, Chen S. A traveling wave natural frequency-based single-ended fault location method with unknown equivalent system
impedance. Int Trans Electr Energy Syst. 2016;26(3):509-524.
17. Eng NE, Ramar K. Single-ended traveling wave based fault location on two terminal transmission lines. TENCON 2009 - 2009 IEEE
Region 10 Conference; 2009;1-4.
16 of 18 SHAFIEI CHAFI AND AFRAKHTE

18. Naidu OD, Pradhan AK. A traveling wave-based fault location method using unsynchronized current measurements. IEEE Trans Power
Deliv. 2018;34(2):505-513.
19. Lai LL, Rajkumar N, Vaseekar E, Subasinghe H, Carter A, Gwyn BJ. Wavelet transform and neural networks for fault location of a teed-
network. PowerCon 2000. International Conference on Power System Technology. Proceedings (Cat. No. 00EX409); 2000;2:807-811.
20. Cheong WJ, Aggarwal RK. A novel fault location technique based on current signals only for thyristor controlled series compensated
transmission lines using wavelet analysis and self organising map neural networks. Eighth IEE International Conference on Developments
in Power System Protection. 2004; 224-227.
21. Aggarwal RK, Blond SL, Beaumont P, Baber G, Kawano F, Miura S. High frequency fault location method for transmission lines based
on artificial neural network and genetic algorithm using current signals only. 11th IET International Conference on Developments in
Power Systems Protection (DPSP) 2012; 111-111.
22. Oleskovicz M, Coury DV, Aggarwal RK. A complete scheme for fault detection, classification and location in transmission lines using
neural networks. 7th International Conference on Developments in Power Systems Protection (DPSP); 2001;335-338.
23. Amorim HP, Huais L. Faults location in transmission lines through neural networks. IEEE/PES Transmision and Distribution Confer-
ence and Exposition: Latin America (IEEE Cat. No. 04EX956). 2004;691-695.
24. Liang R, Wang F, Fu G, Xue X, Zhou R. Wide-area fault location based on optimal deployment of the traveling wave recorders. Int Trans
Electr Energy Syst. 2016;26(8):1661-1672.
25. Azizi S, Sanaye-Pasand M, Abedini M, Hasani A. A traveling-wave-based methodology for wide-area fault location in multiterminal DC
systems. IEEE Trans Power Deliv. 2014;29(6):2552-2560.
26. Chen Y, Liu D, Xu B. Wide-area traveling wave fault location system based on IEC61850. IEEE Trans Smart Grid. 2012;4(2):1207-1215.
27. Korkali M, Lev-Ari H, Abur A. Traveling-wave-based fault-location technique for transmission grids via wide-area synchronized voltage
measurements. IEEE Trans Power Syst. 2011;27(2):1003-1011.
28. Chen R, Yin X, Li Y. Fault location method for transmission grids based on time difference of arrival of wide area travelling wave. J Eng.
2019;2019(16):3202-3208.
29. Dobakhshari AS, Ranjbar AM. A novel method for fault location of transmission lines by wide-area voltage measurements considering
measurement errors. IEEE Trans Smart Grid. 2014;6(2):874-884.
30. Azizi S, Sanaye-Pasand M. A straightforward method for wide-area fault location on transmission networks. IEEE Trans Power Deliv.
2014;30(1):264-272.
31. Hosseini SA, Sadeh J, Mozafari B. Robust wide-area impedance-based fault location method utilising LAV estimator. IET Gen Trans Dis-
trib. 2016;10(10):2475-2485.
32. Majidi M, Etezadi-Amoli M, Fadali MS. A sparse-data-driven approach for fault location in transmission networks. IEEE Trans Smart
Grid. 2015;8(2):548-556.
33. Salehi-Dobakhshari A, Ranjbar AM. Robust fault location of transmission lines by synchronised and unsynchronised wide-area current
measurements. IET Gen Trans Distrib. 2014;8(9):1561-1571.
34. Dobakhshari AS, Ranjbar AM. A closed-form solution for transmission line fault location using local measurements at a remote substa-
tion. Electr Pow Syst Res. 2014;111:115-122.
35. Dobakhshari AS, Ranjbar AM. A wide-area scheme for power system fault location incorporating bad data detection. IEEE Trans Power
Deliv. 2015;30(2):800-808.
36. Azizi S, Sanaye-Pasand M, Paolone M. Locating faults on untransposed, meshed transmission networks using a limited number of syn-
chrophasor measurements. IEEE Trans Power Syst. 2016;31(6):4462-4472.
37. Azizi S, Sanaye-Pasand M. From available synchrophasor data to short-circuit fault identity: formulation and feasibility analysis. IEEE
Trans Power Syst. 2016;32(3):2062-2071.
38. Dobakhshari AS. Wide-area fault location of transmission lines by hybrid synchronized/unsynchronized voltage measurements. IEEE
Trans Smart Grid. 2016;9(3):1869-1877.
39. Mortazavi SH, Sadeh J. An analytical fault location method based on minimum number of installed PMUs. Int Trans Electr Energy Syst.
2016;26(2):253-273.
40. Yao W, Zhou D, Zhan L, et al. GPS signal loss in the wide area monitoring system: prevalence, impact, and solution. Electr Power Syst
Res. 2017;147:254-262.
41. Neyestanaki MK, Ranjbar AM. An adaptive PMU-based wide area backup protection scheme for power transmission lines. IEEE Trans
Smart Grid. 2015;6(3):1550-1559.
42. Feng G, Abur A. Fault location using wide-area measurements and sparse estimation. IEEE Trans Power Syst. 2015;31(4):2938-2945.
43. Kang N, Liao Y. Double-circuit transmission-line fault location utilizing synchronized current phasors. IEEE Trans Power Deliv. 2013;28
(2):1040-1047.
44. Liao Y. Fault location for single-circuit line based on bus-impedance matrix utilizing voltage measurements. IEEE Trans Power Deliv.
2008;23(2):609-617.
45. Kang N, Liao Y. Double-circuit transmission-line fault location with the availability of limited voltage measurements. IEEE Trans Power
Deliv. 2011;27(1):325-336.
46. Exposito AG, Abur A. Power System State Estimation: Theory and Implementation. CRC press: Taylor & Francis; 2004. https://www.
routledge.com/Power-System-State-Estimation-Theory-and-Implementation/Abur-Exposito/p/book/9780824755706
47. Matpower: A matlab power system simulation package. [Online]. Available from http://www.pserc.cornell.edu/matpower/
SHAFIEI CHAFI AND AFRAKHTE 17 of 18

48. Saha MM, Izykowski J, Rosolowski E. Fault Location on Power Networks. Berlin, Germany: Springer; 2010.
49. Dobakhshari AS. Fast accurate fault location on transmission system utilizing wide-area unsynchronized measurements. Int J Electr
Power Energy Syst. 2018;101:234-242.

How to cite this article: Shafiei Chafi Z, Afrakhte H. Wide area impedance-based fault location using
synchronized/unsynchronized current measurements. Int Trans Electr Energ Syst. 2020;e12658. https://doi.org/
10.1002/2050-7038.12658

A. APPENDIX

A.1. Magnitude derivative with respect to x

In this subsection according to similarity of derivative process in (21)-(23), derivative in (21) is calculated and it can be
calculated for (22)-(23) in a similar way. The magnitude derivative with respect to x is as (A1).
 
∂ jΔI k j ∂ I f jAk f 1 ðx Þ + Bk f 2 ðx Þj ∂ ðjAk f 1 ðx Þ + Bk f 2 ðx ÞjÞ
H k,1 = = = If ðA1Þ
∂x ∂x ∂x

Consider z = |Akf1(x) + Bkf2(x)| then (A1) is rewritten as (A2).

pffiffiffiffiffiffiffi
∂ ðj z j Þ ∂ zz* I f ∂ ðzz* Þ I f ∂ ðzz* Þ
H k,1 = I f = If = pffiffiffiffiffiffiffi = ðA2Þ
∂x ∂x 2 zz ∂x
* 2jzj ∂x

substituting z with its assumed value yields

If ∂h i
H k,1 = × ðA k f 1 ðx Þ + B k f 2 ðx ÞÞðA k f 1 ðx Þ + B k f 2 ðx Þ Þ* ðA3Þ
2jAk f 1 ðx Þ + Bk f 2 ðx Þj ∂x

A.2. Phase angle derivative with respect to x

In this subsection according to similarity of derivative process in (27)-(29), derivative in (27) is calculated and it can be
calculated for (28)-(29) in a similar way. Consider (A4) for current difference.

∡ΔIk = ∡ðAk f 1 ðxÞ + Bk f 2 ðxÞÞIf = ∡ðAk f 1 ðxÞ + Bk f 2 ðxÞÞ + ∡If ðA4Þ

ðz Þ
Let z = Akf1(x) + Bkf2(x) and u = Im
ReðzÞ then (A4) is rewritten as (A5).

∡ΔIk = tan − 1 u + ∡If ðA5Þ

Now applying the derivative with respect to x results in (A6).

∂∡ΔIk ∂∡ðAk f 1 ðxÞ + Bk f 2 ðxÞÞ ∂ ðtan − 1 uÞ 1 ∂u

H k,1 = = = = ðA6Þ
∂x ∂x ∂x 1 + u2 ∂x
18 of 18 SHAFIEI CHAFI AND AFRAKHTE

Replacing u and then z with their hypothetic value yields


ImðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ
1 ∂ ReðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ
H k,1 =
2 × ðA7Þ
ImðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ ∂x
1+ ReðAk f 1 ðx Þ + Bk f 2 ðx ÞÞ