1
Phasor Domain Transmission Line Fault Locating
with Three Phase Distributed Parameter Modeling
Yu Liu, Member, IEEE, Zhenyu Tan, Student Member, IEEE, and Jiahao Xie, Student Member, IEEE
Abstract: Accurate fault locating reduces power outage time and
operational costs. In this paper, a novel phasor domain
transmission line fault locating algorithm is proposed. Best
implementation requires GPS synchronized measurements at both
terminals of the line. First, the model of the transmission line with
fault is constructed using three phase distributed parameter
modeling with neutral conductors and grounding representation.
The model is without the assumption that the transmission line is
geometrically balanced. Next, the state estimation algorithm is
utilized to solve the model of the transmission line with fault, where
the fault location is included as a state. Numerical experiments
demonstrate that the method has higher accuracy than traditional
fault locating methods, independent of fault types, locations and
impedances.
Key words: Three Phase Distributed Parameter Modeling, State
Estimation (SE), Fault Locating
I INTRODUCTION
A CCURATE transmission line fault locating minimizes the
time spent searching for the fault and therefore reduces
power outage time and operational costs. Legacy transmission
line fault locating techniques are mainly classified into model
based methods and measurement based methods.
Model based methods locate the fault based on accurate
models of the transmission line of interest. They can be further
categorized into phasor-domain model based methods and
time-domain model based methods. Phasor-domain model
based methods typically use phasors at the fundamental
frequency [1-5]. Single-ended phasor-domain methods utilize
phasor measurements from one terminal of the transmission
line to locate the fault. However, the main disadvantage of
these methods is that they are not accurate with high fault
impedances [4]. To solve the above issue, researchers
proposed dual-ended phasor-domain methods based on phasor
measurements from both terminals of the transmission line, to
eliminate the influence of fault impedances [5]. These
methods can be further classified into methods that require or
do not require GPS synchronized measurements. The main
disadvantages of phasor-domain methods are as follows. (a)
Most of these methods fail to fully consider the distributed
shunt capacitance through the line. Instead, they typically
neglect the capacitance or consider them as lumped
parameters only at terminals of the line. (b) Most of these
methods use sequence line models (sequence lumped
parameter models or sequence distributed parameter models),
which are based on the assumption that the transmission line is
geometrically balanced. (c) The neutral conductors and the
grounding of the transmission lines are typically neglected.
978-1-5386-7703-2/18/$31.00 ©2018 IEEE
Authorized licensed use limited to: ExxonMobil. Downloaded on April 05,2022 at 03:54:33 UTC from IEEE Xplore. Restrictions apply.
Besides the phasor domain model based methods, researchers
also proposed time-domain model based methods [6-8]. The
main advantage of the time-domain methods over phasor-
domain methods is that they do not filter out high frequency
components and therefore they are more accurate especially
during system transients. Most time-domain model based
methods are derived from the Bergeron’s method [9], which
represent the present voltages and currents at one terminal of
the line as functions of the voltages and currents at the other
terminal of the line an extremely short period of time (the
traveling time of the electromagnetic wave between two
terminals of the line) ago. The disadvantages include: (a) the
methods need very high sampling rates to ensure accuracy (a
proper time resolution is required to accurately express the
relationship within this extremely short period of time); (b)
most of these methods depend on modal (eg. sequence
component) decomposition methods, which also assumes that
the transmission line is geometrically balanced; and (c) the
neutral conductors and the grounding of the transmission lines
are typically neglected. Furthermore, researchers also
proposed a time-domain dynamic state estimation based fault
locating method [8]. The method utilizes three phase and
neutral line models and does not need high sampling rates.
However, the method is based on the multi-section cascading
π-equivalent model, which is an approximation of the fully
distributed line model.
For measurement based methods, their accuracy does not
depend on accurate modeling of transmission lines. The most
widely adopted methods are traveling wave based methods
[10-13]. These methods monitors the arrival time of traveling
waves at terminals of the line of interest when a fault occurs
inside the line. Single-ended traveling wave based methods
[10-11] utilize subsequent wavefront arrival time at the local
end and locate the fault by the time differences, including
Type A, C, E and F fault locators. Dual-ended methods [12-13]
utilize arrival time of the first wavefront at both terminals of
the line and locate the fault by the time difference, including
Type B and D fault locators. Further, some researchers utilize
signal processing techniques such as wavelet transform [14] to
accurately detect the wavefront arrival time. The main
disadvantages include: (a) the intensity of the traveling wave
is related to the inception angle of the fault; (b) the accuracy
of the traveling wave based methods highly depends on the
sampling rate (eg. 0.93 miles systematic error with Type D
fault locator and 100 kilo-samples/second sampling rate).
 2

In this paper, a novel phasor domain transmission line fault dW  y  dy  BW  y  (2)
locating method is proposed. Best implementation requires
 V  y    0 I n n 
GPS synchronized measurements at both terminals of the line. where W  y     , B  , I n n is the
 d V  y  dy 
 A 0 
The method utilizes three phase distributed parameter
modeling with neutral conductors and grounding identity matrix with the dimension of n, and n is the number of
representation. It accurately models transmission lines without conductors of the line, and A   R1  j L1  G1  jC1  .
the geometrically balanced assumption. Moreover, the state The general solution of (2) is,
estimation algorithm [15-17] is adopted to take full advantage W  y   e yB C (3)
of the redundancy inside the fault location problem, to cross
where C is a constant complex vector (to be determined by
check the fault locating results and to improve the fault
the boundary conditions), and the matrix exponential function
locating accuracy. The rest of the paper is organized as 

  yB 
m
follows. Section II derives a phasor domain three phase eyB is defined as m! .
m 0
distributed parameter model of the transmission line. Section
Next, consider the boundary conditions (last row) in (1),
III provides detail modeling procedure of a transmission line
with fault. Section IV demonstrates the state estimation  BelB C   ZIS ,  BC 1:n  ZIR , elB C   VS , C 1:n  VR (4)
1:n 1:n
procedure to solve the location of the fault. Section V shows where the notation  a:b means row a to row b of vector   ,
simulation results. Section VI draws a conclusion.
and Z  R1  j L1 .
II PHASOR DOMAIN TRANSMISSION LINE THREE PHASE Define elB   M11 , M12 ; M 21 , M 22  , where Mij  i, j  1, 2
DISTRIBUTED PARAMETER MODEL lB
is a submatrix of e with the dimension of n  n .
The three phase distributed parameter model of a From the first two sub-equations in (4), we can express the
transmission line can be generated by combining infinite
number of infinitesimal sections. In time domain, the model constant vector C as functions of VR and VS ,

 
T
can be expressed via a set of partial differential equations.
C  VR  M12  VS  M11VR 
1
(5)
Afterwards, the following equation can be obtained by  
substituting time domain waveforms with phasors [18]: Substitute (5) to the last two sub-equations in (4),
 d V 2 y    R1  jL1 G1  jC1 V  y   M12 1 VS  M11VR  =ZIR
2 


 dy  (6)
   M 21VR  M 22  M12  VS  M11VR   ZIS
1

 dV  y    R  jL  I  y  (1)
 dy 1 1
From (6), represent VS and IS as functions of VR and IR ,
 Boundary Conditions : I  l   I , I  0  I , V  l   V , V  0  V 1
S R S R VS   I n n 0  lB  I n n 0  VR 
  e    (7)
where I  y  and V  y  are current and voltage phasor vectors  IS   0 Z   0 Z   IR 
at location y of the transmission line; IS , VS , IR and VR are The above equation (7) is the three phase distributed
current and voltage phasor vectors at the sending end and the parameter model of a transmission line. In fact, (7) is a
generalization of the modal decomposition based distributed
receiving end of the line; R1 , L1 , G1 and C1 are series
parameter model. (7) is equivalent to the modal
resistance, series inductance, shunt conductance and shunt decomposition based model with the following two
capacitance matrices per unit length. assumptions: (a) the neutral conductors and the grounding
To solve (1), popular methods adopt modal decomposition of the line are neglected and (b) for the parameter
methods (such as sequence component transformation, clarke matrices, the diagonal elements are the same and the off-
transformation, etc) to transform (1) into a group of scalar diagonal elements are the same. Next, the proof of the above
equations (corresponding to several independent single phase statement is provided.
transmission lines). However, these modal transformations are Proof: Take a three phase transmission line as an example.
based on the assumption that (a) the neutral conductors and
the grounding of the circuit are negligible, and (b) the From the assumption (a) IS , VS , IR and VR represent three
transmission lines are geometrically balanced, i.e. the diagonal phase values without neutrals and in this case n = 3.
elements of the parameter matrices R1 , L1 , G1 and C1 are Here the sequence component transformation is adopted
the same and the off-diagonal elements are the same. These (the proofs are similar for other types of modal decomposition
assumptions could potentially generate considerable errors. methods). Define sequence component transformation matrix
To accurately solve (1), the proposed direct method is T  1,1,1; e  j120 , e j120 ,1; e j120 , e  j120 ,1 , we have,
introduced next. Note that this method does not have the  I  T  I T T
 S  S IS  IS 0  , VS  T VS  VS  VS 0 
above two assumptions. (8)
 T T
The first two rows of (1) can be equivalently expressed as  I R  T  IR  IR  IR 0  , VR  T VR  VR  VR 0 


Authorized licensed use limited to: ExxonMobil. Downloaded on April 05,2022 at 03:54:33 UTC from IEEE Xplore. Restrictions apply.
 3

where subscript +, - and 0 correspond to positive, negative and Substitute equation (8) and (13) into (7),
zero sequence values.   cosh    l  Z  sinh    l   
From the assumption (b),  VS     VR  

  I   1 sinh   l    I 
 
 1 1    
 R  j L  Tdiag  r  jl r  jl r  jl T T 1
 0 0
(9)
  S 


 Z 
 cosh   l  R 

  cosh    l  Z  sinh    l   
 
G1  jC1  Tdiag  g  jc g  jc g0  jc0  T 1
T
  
 VS   
    1   VR  
 
(14)
  I S    Z sinh    l  cosh    l    IR  
where diag  is the diagonal matrix with column vector     

as the diagonal elements.   cosh   0 l  Z 0 sinh   0 l   

 VS 0    VR 0 
From (9), the matrix B defined in (2) is,       1 sinh  l  
  IS0   Z  0 cosh   0 l   IR 0 


0 I33 
 0 I33     T6 DT61 (10)
  0
B  
 A 0  Tdiag      0  T
 2 2
 2 T
 1
0 
  From the proof above, the proposed model is a
generalization of the modal decomposition based distributed
T 0  parameter model but without the aforementioned two
where  i   ri  jli  gi  jci   i , , 0 , T6    and assumptions. Consequently, the proposed model is more
0 T  accurate.
 0 I 33 
D .

 diag   2   2  0 2 

T
0 
  III MODELING OF TRANSMISSION LINES WITH FAULT
The transmission line with fault can be accurately modeled
Apply eigenvalue decomposition to matrix D , with the aforementioned three phase distributed parameter
D  PΛP 1 (11) modeling method shown in (7). The mathematical model of
where the matrix Λ is the diagonal eigenvalue matrix of D , the transmission line with fault is a set of complex algebraic
and the columns of matrix P are the corresponding equations describing all the physical laws that the transmission
eigenvectors, line with fault should obey. It has the following format:
     1 1 0 0 0 0   zvi  f  x 
   0 0 1 1 0 0   (15)
      0  g  x 
     0 0 0 0 1 1 
Λ  diag     P    where z vi is the terminal voltage and current phasor
        0 0 0 0  measurement vector of the line with fault, x is the state vector
     0 0   0 0 
  0  of the system (the location of the fault is one element of x ).
     
  0   0 0 0 0 0 0  Note that the terminal neutral voltages and currents are
introduced as pseudo measurements (zero values but with
Therefore, from equation (10) and (11), large error standard deviations) and are included in z vi . Also,
 lT PΛP T61 
m
 lB  T6 P  lΛ P 1T61
m 1 m
  
the second row of (15) corresponds to internal constraints that
e   
lB 6

m0 m! m0 m! m 0 m! (12) describe the relationship among elements of the state vector.

 P T
l  l  l   l  l 0 l 0 T In this paper, the modeling of two-terminal lines is
 T6 Pdiag e e e e e e  1
6
1
presented as an example. Note that similar modeling
From equation (9) and (12), procedure can be applied to two-terminal, multi-terminal and
1 inhomogeneous transmission lines. The construction process
 Inn 0  lB  Inn 0  of the model is shown in Figure 1 (a 4-conductor transmission
 0 Z e  0 Z
 1  1 line as an example). It consists of three parts: the line at the
1 left side of the fault, the line at the right side of the fault, and
 I33 0   I33 0 
    the fault itself. Note that both transmission lines utilize the
 Z     lB   Z    
   e  
modeling method in (7). The variables are defined as follows:
 0 Tdiag Z   T 1   0 Tdiag Z   T 1  l f is the distance between the fault and the left side of the
       
   
   Z0 0       Z0 0    line; l is the total length of the line; I1 , V1 , I2 and V2 are
  cosh   l     Z sinh   l    three phase and neutral current and voltage phasor vectors at
     
 diag  cosh   l    diag  Z sinh   l    both terminals of the line; V f , I f 1 and I f 2 are three phase
  cosh   l      Z sinh   l   
 0   0 0  and neutral voltage phasor vector, left side current phasor
 T6   T 1
vector, and right side current phasor vector measured at the
  sinh   l  Z    cosh   l    
6

      location of the fault.

diag  sinh   l  Z   diag  cosh   l     The detail matrices of the model are,
  sinh   l  Z    cosh   l     T T
  0 0   0    (13) z  V1 I1 V2 I2  , x  V f I f 1 I f 2 l f  ,

Authorized licensed use limited to: ExxonMobil. Downloaded on April 05,2022 at 03:54:33 UTC from IEEE Xplore. Restrictions apply.
 4

  I 4 4 04 4  1 l B  I 4 4 04 4   I 4 4 04 4 04 4 041   as shown in Figure 2. The rest of the network is not shown.
   e 0  
f
 The tower structure is also shown in Figure 2. Three phase
  04  4 Z 1   4 4 Z1   04 4 I 4 4 04 4 041   ;
f  x   1 x voltage and current measurements are installed at both
  I 4 4 04 4   l  l f  B  I 4 4 04 4   I 4 4 04 4 04 4 041   terminals of the line. Extensive number of events with
 0  e 0  
  4  4 Z1   4 4 Z1   04 4 04 4 I 4 4 041   different fault time, types, locations and impedances has been
g  x   0n fault  4 T fault T fault0n fault 1  x ;

simulated. The instantaneous measurement data are stored in
COMTRADE files for experimentation with the sampling rate
where T fault and n fault are defined in Table 1. of 80 samples per cycle (4.8 kilo-samples per second)
lf l  lf according to the IEC61850-9-2LE standard. The phasors of
corresponding measurements are afterwards calculated
I1 I f 1 I f 2 I2
a according to the IEEE C37.118 standard.
b
c
V1 n V f V2

Figure 1. Phasor device model of two-terminal line with fault

Table 1. Values of T fault and nfault for different fault types

130 feet
100 feet
Fault Type Value of matrix Tfault nfault
A-N [1 0 0 1; 0 1 0 0; 0 0 1 0] 3
B-N [1 0 0 0; 0 1 0 1; 0 0 1 0] 3
C-N [1 0 0 0; 0 1 0 0; 0 0 1 1] 3
A-B [1 1 0 0; 0 0 1 0; 0 0 0 1] 3 Figure 2. Example test system: two-terminal transmission line
B-C [1 0 0 0; 0 1 1 0; 0 0 0 1] 3 Two legacy fundamental frequency phasor based methods
C-A [1 0 1 0; 0 1 0 0; 0 0 0 1] 3
AB-N [1 1 0 1; 0 0 1 0] 2
are used for comparison. The two methods are both based on
BC-N [0 1 1 1; 1 0 0 0] 2 dual-ended three phase voltage and current phasors. The first
CA-N [1 0 1 1; 0 1 0 0] 2 legacy method (also known as the impedance based method)
3 phase [1 1 1 0; 0 0 0 1] 2 refers to the IEEE standard [10], which neglects capacitive
charging currents through the transmission line and uses
IV STATE ESTIMATION PROCEDURE sequence line models. The second legacy method uses
To solve the state vector (including the location of the fault) sequence distributed parameter model of transmission line
based on modal (sequence component) decomposition. We
in equation (15), the state estimation procedure is adopted.
compare the proposed method to the above two legacy fault
The method we use is the unconstraint weighted least square
locating schemes via the following test cases. Here we do not
method, where the constraints are treated as virtual
consider time-domain model based legacy methods or
measurements (with zero values but with much smaller traveling wave based legacy methods for comparison since
measurement error standard deviations compared to the actual their fault locating accuracy cannot be guaranteed with low
measurements). Therefore, equation (15) becomes: sampling rates.
z   zactual 0  h  x 
T
(16) Test Case 1: Single Phase to Neutral Faults
A 0.01 ohm phase A to neutral fault occurs at 80 miles from
Note that for two-terminal transmission lines, we have 16 + side A1 and at time 1.0 to 1.1 seconds. The fault locating
nfault measurements and 13 states (i.e. 38.5% to 46.2% results, including the legacy impedance based method, the
redundancy) to cross check the fault locating results and legacy modal decomposition based method and the proposed
improve the fault locating accuracy. method, are shown in Figure 3. We can observe that the
The best estimation of the state vector can be obtained by proposed method (79.9660 miles) has higher accuracy
solving the following optimization problem: compared to legacy impedance based method (81.1237 miles)
min J   h  x   z  W  h  x   z 
T
(17) and legacy modal decomposition based method (80.8011
miles).
where W  diag   ,1  i2 ,   , and  i  i  1, 2,  is the To further validate the effectiveness of the proposed method,
error standard deviation of measurement i. three groups of phase A to neutral faults at different fault
The solution is given with the following Newton’s iterative locations, with 0.01 ohm, 1 ohm and 10 ohm fault impedances,
method until convergence: are tested. The results are depicted in Figure 4. We can
observe that the proposed method is more accurate than both
x 1  x   H T WH  H T W (h  x   z )
1
(18) legacy methods.
Test Case 2: High Impedance Faults
where H  h  x  x x  x .
Three groups of high impedance phase A to neutral faults at
different fault locations, with 100 ohm, 300 ohm and 500 ohm
V SIMULATION RESULTS fault impedances, are tested. The results are depicted in Figure
The proposed algorithm is validated via a 500 kV, 135.22- 5. We can observe that the proposed method is more accurate
mile, two-terminal transmission line of interest (line A1-A2), than both legacy methods. In this case the maximum absolute

Authorized licensed use limited to: ExxonMobil. Downloaded on April 05,2022 at 03:54:33 UTC from IEEE Xplore. Restrictions apply.
 5

errror of the prroposed meth hod during a 500 ohm high algorithhm has been prresented to acccurately determ mine the fault
im
mpedance phasee A to neutral fault
f is only aro
ound 0.7 miles.. locationn in transmisssion lines. F First, the moodel of the
Actual fault location from side A1 (miles) 83 transmi ssion line w with fault is built using three phase
Imped
dance based meth hod
82.5 distribuuted parameterr modeling wiith neutral connductors and
Modall decomposition based
b method
Propo
osed method
groundiing representattion. Next, the state estimatioon algorithm
82
is appli ed to computee the best estim mate of the statte (including
81.5 the locaation of the fauult) of the line with fault. Thee advantages
of the proposed algoorithm includee: (a) it uses three phase
81
distribuuted parameter modeling insteead of modal (eg. sequence
80.5 componnent) decompposition baseed distributedd parameter
80
modelinng, and does nnot assume thaat the line is ggeometrically
balanceed; (b) it conssiders neutral conductors annd grounding
79.5 represenntation; (c) it takes full advvantage of the redundancy
79 inside the fault locaating problem m to further iimprove the
1 1.02 1.0
04 1.06 1.08 1.1 accuraccy; (d) it demoonstrates higheer accuracy com mpared with
Time (s)
existingg fault locatingg schemes.
Fiigure 3. Fault lo
ocating results comparison,
c a 0.01
0 ohm phase A
to neutral
n fault, 80 miles from sidee A1
REFERENCEES
[1] SELL-351 Protection SSystem Instruction Manual, Schweitzer Engineering
Labboratories Inc., Pulllman, WA, 2015.
[2] GE D90Plus Line D Distance Protectioon System Instruuction Manual,
Genneral Electric Multtilin, Markham, Onntario, Canada, 20012.
[3] Siemmens SIPROTEC C Line Differential Protection with Distance
Pro tection 7SD5, Veersion 4.70, Siemeens, Munich, Bavvaria, Germany,
201 1.
[4] Ha, H., Zhang, B., annd Lv, Z., "A novvel principle of sinngle-ended fault
locaation technique forr EHV transmissioon lines", IEEE Traans. Power Del.,
vol. 18, no.4, pp. 11477- 1151, Oct. 2003.
[5] Kaw wady, T., and Steenzel, J., “A pracctical fault locatioon approach for
doub
uble circuit transm mission lines usingg single end dataa,” IEEE Trans.
Powwer Del., vol. 18, nno. 4, pp. 1166–1173, Oct. 2003.
[6] Sonng, G., Suonan, J.., Xu, Q., et al: ““Parallel transmisssion lines fault
locaation algorithm baased on differentiial component nett,” IEEE Trans.
Powwer Del., 2005, 20,, (4), pp. 2396-24006.
[7] Goppalakrishnan, A., Kezunovic, M., McKenna, S. M.., et al:, “Fault
locaation using the disstributed parameteer transmission linee model,” IEEE
Tranns. Power Del., 20000, 15, (4), pp. 11169-1174.
[8] Liu,, Y., Meliopouloss, A. P., Tan, Z., eet al: “Dynamic S State Estimation
Bas ed Fault Locatiing on Transmission Lines,” IE ET Generation
Figgure 4. Fault locating results comparison,
c 0.01 ohm, 1 ohm and
a Trannsmission Distribuution, in press.
10 ohm phase A to neutral faults,
f variable fault location [9] Dom mmel, H. W., ““Digital Computeer Solution of E Electromagnetic
Trannsients in Single--and Multiphase N Networks, ” IEEE E Trans. Power
Appp. Syst., vol. PAS-888, no. 4, pp. 388-399, April 1969.
[10] IEEEE Guide for Deteermining Fault Loocation on AC Trransmission and
Disttribution Lines, IEEEE Standard C37.114, 2014.
[11] Azizzi, S., Sanaye-Passand, M., Abedinni, M., et al.: ‘A ttraveling wave-
baseed methodology ffor wide-area fauult location in muultiterminal DC
systtems,’ IEEE Transs. Power Del., 2014, 29, (6), pp. 25552–2560.
[12] Dew we, M. B., Sankarr, S., and Arrillagaa, J.: ‘The applicaation of satellite
timee references to HV VDC fault locationn’, IEEE Trans. Poower Del., 1993,
8, (33), pp. 1295–13022.
[13] Jafaarian, P., and Sanaaye-Pasand, M.: ‘AA traveling wave bbased protection
techhnique using waveelet/PCA analysis’’, IEEE Trans. Poower Del., 2010,
25, (2), pp. 588–599.
[14] Liu,, Y., Meliopouloos, A. P., Tai, N N., et al.: “Protecction and Fault
Loccating Method off Series Compennsated Lines by Wavelet Based
Eneergy Traveling Waave”, in IEEE Pow wer Energy Soc (PE ES). Gen. Meet.,
Julyy. 2017, pp. 1–5.
[15] Zhaao, J., Zhang, G., Das, K., Korres,, et al: "Power syystem real-time
monnitoring by using pmu based robustt state estimation method," IEEE
Trannsactions on Smarrt Grid, vol. 7, no. 1, pp. 300-309, Jaan 2016.
[16] Melliopoulos, A. P., C Cokkinides, G. J., Myrda, P., et al: ‘Dynamic state
estimmation based prootection: status annd promise’, IEEE E Trans. Power
F
Figure 5. Fault locating
l results comparison, 10 00 ohm, 300 ohm m Del.., 2017, 32, (1), ppp. 320-330.
aand 500 ohm ph hase A to neutraal faults, variab
ble fault location
n [17] Liu,, Y., Meliopouloss, A. P., Fan, R., et al: ‘Dynamic state estimation
baseed protection on seeries compensatedd transmission linees’, IEEE Trans.
Powwer Del., 2017, 32,, (5), pp 2199-22009.
VI CON
NCLUSION
[18] Melliopoulos, A. P.,, Power System Grounding and Transients: an
A state estim
mation based phasor domaiin fault locating Introoduction. M. Dekkker, 1988.

Authorized licensed use limited to: ExxonMobil. Downloaded on April 05,2022 at 03:54:33 UTC from IEEE Xplore. Restrictions apply.