Electric Power Systems Research 93 (2012) 3945

Contents lists available at SciVerse ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

On-line estimation of transmission line parameters, temperature and sag using

PMU measurements
Yan Du , Yuan Liao
453FPAT, Department of Electrical and Computer Engineering, University of Kentucky, Lexington, KY 40506, USA

a r t i c l e

i n f o

Article history:
Received 27 May 2012
Received in revised form 9 July 2012
Accepted 9 July 2012
Available online 28 July 2012
Keywords:
Dynamic thermal rating
Phasor Measurement Unit
Sag
Transmission line parameter estimation

a b s t r a c t
This paper presents a framework for estimating transmission line positive sequence parameters, temperature, and sag utilizing on-line voltage and current phasors measured by Phasor Measurement Units.
Transmission line parameters, temperature, and sag are vital inputs to various power system analyses
and applications, such as the power ow analysis, the protective relaying application, and the dynamic
thermal rating application. In this study, different types of transmission line topology are classied and
modeled to facilitate the development of corresponding algorithms. All models are based on distributed
parameters, so it automatically considers the shunt capacitance effect of long lines. It is assumed that the
line is transposed. The proposed least squares based algorithms harness the non-linear optimal estimation theory and are capable of detecting and identifying bad data, minimizing the impact of measurement
errors and thus signicantly improving the estimation accuracy. The line temperature and sag are calculated according to the obtained parameter estimates. Numerical case studies based on simulated data
are presented.
2012 Elsevier B.V. All rights reserved.

1. Introduction
The research on power transmission lines requires transmission
line parameters, including series resistance, series reactance and
shunt susceptance as vital inputs to various power system analyses
and applications, such as the power ow analysis and the protective
relaying application [1]. Accurate estimate of line parameters may
also be employed for transmission line thermal condition monitoring [2]. So it will be desirable if line parameters can be accurately
estimated using on-line voltage and current measurements.
Numerous algorithms for estimating transmission line parameters have been proposed in the past. Ref. [3] introduces classical
approaches for estimating line parameters. Deviations from actual
values of employed factors such as conductor type and geometric
parameters may cause errors in estimation. Another procedure presented in [4] harnesses voltage and current phasors from one end
of a line while assuming the other end is open or short-circuited.
Difculties in obtaining the required measurements hinder the
applicability.
The on-line parameter estimation approaches which employ
voltage and current phasors should be benecial to various applications. In [5], the ABCD parameters are obtained by using two sets
of synchronous phasors from two terminals, yet without showing

Corresponding author. Tel.: +1 859 257 6588; fax: +1 859 257 3092.
E-mail address: yan.du@uky.edu (Y. Du).
0378-7796/$ see front matter 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.epsr.2012.07.007

how distributed parameters are obtained. The Global Positioning System provides a way to improve synchronization while the
estimation accuracy relies essentially on the precision of the synchronization angle [6]. As introduced in [7], the characteristic
impedance and propagation constant of a line can be derived by
making use of on-line phasors captured at both ends. According to
[8], by utilizing three sets of synchronous phasors based on Laplace
transform, the proposed algorithm can solve for line parameters. It
is assumed by all these approaches that high precision in synchronization is available; otherwise, it may yield signicant estimation
errors.
To sum up, existing line parameter estimation algorithms work
well with accurate phasors and synchronization. Sometimes in
practice, bad measurements may occur due to various reasons,
such as current transformer saturation [9], data conversion errors
or communications abnormalities [10]. The occurrence of such bad
data could result in substantial estimation inaccuracy, which brings
the challenge for the performance of developed algorithms.
With the recent development of Smart Grid technology, especially in the key technology area of sensing and measurement,
Phasor Measurement Unit (PMU) is considered as a promising tool
for advanced wide-area monitoring systems [11]. It is capable of
capturing time synchronized phasors from a large transmission
network, which are employed for monitoring, control and protection.
A method for monitoring transmission line sag is proposed in
[12], using a tension based model and state estimation method.

40

Y. Du, Y. Liao / Electric Power Systems Research 93 (2012) 3945

of sub-network are delineated in Section 4. Section 5 introduces

the method to estimate the line temperature and conductor sag.
Numerical case studies and estimation results are presented in
Section 6, followed by the conclusion.
2. The overall framework

Fig. 1. The overall framework.

The required parameters such as wind, ambient temperature, air

and solar data, etc., are obtained through local weather stations or
additional sensors are needed. Hence an alternative way to calculate the conductor temperature and sag using fewer parameters is
desired.
Based on the relationship between conductor temperature and
resistance [1315], a possible online method for transmission
line thermal condition monitoring based on synchronized phasors
installed at terminals of transmission lines is put forth in [2]. The
average line temperature can be estimated based on the conductor characteristics and the average line resistance, from which the
conductor sag can be derived. This type of method will only require
PMUs to be installed to collect phasors, and appropriate algorithms
for estimating the line resistance, eliminating the need for extra
tension or weather sensors.
Building upon [2], this paper further develops and describes the
method for estimating positive sequence line parameters, temperature, and sag using on-line voltage and current synchrophasors. A
bad data detection and identication method capable of minimizing the impact of measurement and synchronization errors is also
presented. The equivalent circuit based on distributed parameter
line models is employed for the development of algorithms. Hence,
the proposed solution automatically considers the shunt capacitance effect of long lines [3]. In this paper, the transmission line is
assumed to be transposed.
In the rest of the paper, Section 2 presents the overall framework
for estimating the line parameters, temperature and sag. Section
3 classies typical network topologies of transmission lines. The
proposed algorithms to estimate line parameters for each type

The overall framework for estimating transmission line parameters, temperature, and conductor sag using on-line voltage and
current phasors measured by PMUs is described in Fig. 1. Real-time
phasors from PMUs are rstly transferred through communication links to on-line parameter estimator, where line parameters
including series resistance, reactance and shunt susceptance are
derived. The average line temperature is then computed based on
the estimated series resistance. Next, the conductor sag of selected
span can be calculated accordingly. The estimated line parameters
can also be exploited for the state estimation, relaying application, and other such power system applications, e.g. power ow
studies and the real and reactive power control for improved grid
operations. The obtained line temperature and conductor sag data
can be employed for the dynamic thermal rating application and
other power system reliability and safety utilization. It is observed
that the line temperature and sag are calculated indirectly from
real-time phasors from PMUs, which means this method does not
require additional tension sensors or weather data, and thus greatly
simplies the process and reduces both equipment and maintenance costs.
3. Typical transmission network
Large power transmission networks can be decomposed based
on the topology of connections to small basic networks. In this
study, totally four types of sub-networks are classied to facilitate the development of line parameter estimation algorithms, as
shown in Fig. 2. The circle labeled Sub-system in Fig. 2 could
connect to a generation source or to the remaining part of a large
system. A bar labeled in numerical order represents a substation or
a bus, where measurement devices such as PMU can be installed.
The Sub-network Type 1 is a basic two-bus single-circuit transmission network with one Sub-system connected at each terminal.
The Sub-network Type 2 is a two-bus single-circuit transmission
network equipped with a series capacitive compensation device,
which can be a capacitor bank or thyristor-based power ow
controller. The Sub-network Type 3 is a three-bus transmission network that shares a common bus at one substation. The Sub-network
Type 4 is a two-bus double-circuit network that shares a common
bus at both terminals.

Fig. 2. Typical transmission network.

Y. Du, Y. Liao / Electric Power Systems Research 93 (2012) 3945

Fig. 3. Equivalent circuit of Sub-network Type 1.

We assume that synchrophasors of voltages and currents are

obtained by PMUs in real time during normal operations and transmitted to the control center. For each type of sub-network, an
algorithm that is tailored for estimating positive sequence line
parameters will be derived. The line temperature and sag will be
calculated subsequently based on conductor characteristics and
obtained series resistance. The effectiveness and accuracy of all the
algorithms will be evaluated in later sections as well.
4. Proposed method for line parameter estimation
This section delineates the proposed method to estimate positive sequence line parameters. Each sub-network, as shown in
Fig. 2, will be modeled mathematically to form the associated line
equation(s). Then both the classical least squares method and the
optimal estimation theory will be applied to estimate line parameters based on obtained phasors. Due to the similarity of estimation
procedure of the four types of sub-networks, detailed algorithms
will be presented for Sub-network Type 1 only.

where z1 and y1 are the positive sequence series impedance and

shunt admittance of the line, respectively. The real part and imaginary part of z1 , along with the imaginary part of y1 are the three
positive sequence line parameters to be determined.
For each set of measurement, we can obtain two complex equations: (1) and (2), which can be further rearranged into four real
equations, by which the three real unknown line parameters can
be theoretically solved. Alternatively, by using two or more sets of
measurements we can solve for the three unknown line parameters with line equation (1) or (2) only. Apparently, by making use
of multiple sets of measurements, the least squares based method
can generate more robust results.
In order to detect potential synchronization errors, the synchronization angle can also be modeled in the line equations. Eqs. (1)
and (2) can be modied as
VP1i Zc sinh(l)IP1i + sinh(l) tanh
IP1i

It is assumed that the transmission network shown in Fig. 2(a) is

a transposed line between bus P1 and P2, where each Sub-system
is considered as a Thevenin equivalent source. The proposed algorithm employs the steady state voltage and current phasors at bus
P1 and P2, which can be measured instantly at different moments
during normal operations. The equivalent circuit based on distributed parameter model of Sub-network Type 1, as shown in
Fig. 3, is employed for deriving the algorithms.
The following notations are adopted:

 l 
2

VP1i VP2i = 0

tanh(l/2)VP1i
tanh(l/2)VP2i
+ IP2i
=0
Zc
Zc

VP1i VP2i ej = 0

(1)

(5)

(6)

where denotes the synchronization angle between bus P1 and P2.

Similar to synchronous case mentioned above, a total of four
unknowns including three line parameters and the synchronization angle can be theoretically solved utilizing one or more sets of
measurements.
4.1.1. Least squares method
This subsection details the least squares method adopted for
solving line parameters using line equation (5) with synchronization angle in measurements.
Dene the unknown variable vector as
(7)

where x1 , x2 , x3 are the positive sequence series resistance, series

reactance and shunt susceptance per unit length, respectively; x4 is
the synchronization angle. Set x4 = 0 when synchronous measurements are utilized.
By employing the dened variables, (5) can be rewritten as
fi (X) = VP1i Zc sinh(l)IP1i + sinh(l) tanh

 l 
2

VP1i VP2i ejx4 (8)

where i = 1, 2, . . ., N, representing the index of the total number of

measurement set N;
Zc =
=

x1 + jx2
jx3

(9)

(x1 + jx2 )(jx3 )

(10)

Dene the function vector F(X) as:

F2i1 (X) = real(fi (X)),
F2i (X) = imag(fi (X)),

Based on Fig. 3, assuming synchronous phasors are adopted, the

following equations can be obtained:

z1 /y1
Zc =

 = z1 y1

tanh(l/2)VP1i
tanh(l/2)VP2i ej
+ IP2i ej
=0
Zc
Zc

VP1i , VP2i : ith positive sequence voltage phasors at bus P1 and P2,
respectively;
IP1i , IP2i : ith positive sequence current phasors at bus P1 and P2,
respectively;
i = 1, 2, . . ., N: where N is the total number of measurement sets
Zc : characteristic impedance of the line;
: propagation constant of the line;
l: the total length of the line (known);

IP1i

 l 

X = [x1 , x2 , x3 , x4 ]T

4.1. Sub-network Type 1

VP1i Zc sinh(l)IP1i + sinh(l) tanh

41

i = 1, 2, . . . , N
i = 1, 2, . . . , N

(11)
(12)

where real(fi (X)) and imag(fi (X)) yield the real and imaginary part
of fi (X).
Then the unknown variable vector X can be derived iteratively
as follows:
F(Xk )
Xk

(2)

H=

(3)

X = (H T H)

(4)

Xk+1 = Xk + X

(13)
1

[H T F(Xk )]

(14)
(15)

42

Y. Du, Y. Liao / Electric Power Systems Research 93 (2012) 3945

where H is composed of derivatives of the function vector with

respect to X; Xk , Xk+1 are variable vectors for kth and (k + 1)th iteration; k is the iteration number starting from 1; X denotes the
variable update.
4.1.2. Optimal estimation method
Since measurements may contain errors, the nonlinear estimation theory can be employed to derive an optimization algorithm
for solving unknown parameters, capable of minimizing the negative impacts of bad data and thus improving the estimation
accuracy and robustness. Likewise, we only use line equation (5)
for demonstration.
Dene the voltage and current phasors along with the synchronization angle as known measurements
M = [VP11 , IP11 , VP21 , IP21 , . . . , VP1N , IP1N , VP2N , IP2N , ]

(16)

Set = 0 when synchronous measurements are utilized. Modeling in the line equation(s) makes it possible to detect potential
synchronization errors.
Dene the measurement function as
Yi (X) = x2i1 ejx2i

(17)

Y4N+1 (X) = x8N+4

(18)

where i = 1, 2, . . ., N, representing the index of the total number of

measurement set N; and the unknown variable vector X is dened
as
X = [x1 , x2 , . . . , x8N , x8N+1 , x8N+2 , x8N+3 , x8N+4 ]T

(19)

where x1 , x2 , . . ., x8N are unknown variables, representing the 4N

complex measurements; x8N+1 , x8N+2 , x8N+3 are positive sequence
series resistance, series reactance and shunt susceptance per unit
length, respectively; x8N+4 denotes the synchronization angle.
Following the dened variables, (8) can be rewritten as
fi (X) = x8i7 ejx8i6 Zc sinh(l)x8i5 ejx8i4
+ sinh(l) tanh

 l 
2

x8i7 ejx8i6 x8i3 ejx8i2 ejx8N+4

(20)

where i = 1, 2, . . ., N, representing the index of the total number of

measurement set N;

Zc =
=

x8N+1 + jx8N+2
jx8N+3

(21)

(x8N+1 + jx8N+2 )(jx8N+3 )

(22)

Introduce S as the measurement vector with elements shown

below:
Si = 0,

i = 1, 2, . . . , 2N

(23)

S2i1+2N = abs(Mi ),

i = 1, 2, . . . , 4N

(24)

S2i+2N = angle(Mi ),

i = 1, 2, . . . , 4N

(25)

S10N+1 = M4N+1

(26)

where abs(Mi ) and angle(Mi ) yield the magnitude and angle of Mi .

Introduce F(X) as the function vector with elements shown
below:
F2i1 (X) = real(fi (X)),
F2i (X) = imag(fi (X)),

i = 1, 2, . . . , N
i = 1, 2, . . . , N

F2i1+2N (X) = abs(Yi (X)) = x2i1 ,

F2i+2N (X) = angle(Yi (X)) = x2i ,
F10N+1 = Y4N+1 (X) = X8N+4

i = 1, 2, . . . , 4N
i = 1, 2, . . . , 4N

(27)

The measurement vector and function vector are related by

S = F(X) + 

(32)

where  represents measurement errors.

The optimal estimate of X is obtained by minimizing the cost
function dened as
J = [S F(X)]T Q 1 [S F(X)]

(33)

where Q is the covariance matrix, determined according to meter

accuracy.
The solution to (33) can be obtained iteratively as follows:
H=

F(Xk )
Xk

X = (H T WH)

(34)
1

[H T W (S F(Xk ))]

Xk+1 = Xk + X

(35)
(36)

where H is composed of derivatives of the function vector with

respect to X; Xk , Xk+1 are variable vectors for kth and (k + 1)th iteration, respectively; k denotes the iteration number starting from 1;
and X is variable update; W is the weighted matrix, denoted as
W = Q1 .
After X is obtained, (17) can be applied to compute the estimated
values of measured phasors.
4.1.3. Bad data detection and identication
Since bad data can lead to inaccuracy of line parameter estimation, it is highly desirable to develop an effective method capable
of detecting and identifying bad data. The general procedure of
the bad data detection and identication technique based on chisquare test, referring to [3,16], is summarized as follows. First, the
expected value of cost function, equal to the number of degrees of
freedom denoted as k, is calculated as the number of elements contained in the measurement vector S minus the number of unknown
variables [3]. Then the estimated value of cost function CJ is computed. If the estimated value of the cost function equals or exceeds
the expected value of cost function, i.e. CJ 2k, , the presence
of bad data is suspected with a probability of (1 ), and the
associated data corresponding to the largest standardized error
is identied as the bad data. The value of 2k, can be calculated
with a specic k and based on chi-square distribution. In this
study, we choose = 0.01, indicating a 99% condence interval of
the detection.
The general procedures of the line parameter estimation process can be applied to all the remaining sub-networks discussed
thereinafter.
4.2. Sub-network Type 2
Referring to Fig. 2(b), with synchronous phasors adopted, for the
line segment P1C, we obtain
IP1i cosh(l1 )

sinh(l1 )VP1i
= IC3i
Zc

(37)

Likewise, for the line segment P2C, we attain

IP2i cosh(l2 ) +

sinh(l2 )VP2i
= IC3i
Zc

(38)

(29)

where l1 and l2 represent the length of line segment P1C and P2C,
respectively; IC3i is the ith positive sequence phasor current owing
through the compensation device.
It follows from (37) and (38) that

(30)

IP1i cosh(l1 )

(28)

(31)

sinh(l1 )VP1i
sinh(l2 )VP2i
IP2i cosh(l2 )
=0
Zc
Zc
(39)

Y. Du, Y. Liao / Electric Power Systems Research 93 (2012) 3945

For each set of measurements, (39) can be separated into two

real equations with measurements at bus P1 and P2 only. There are
three unknown line parameters. Thus, two or more sets of measurements will enable solutions of the unknown parameters.
4.3. Sub-network Type 3
Referring to Fig. 2(c), utilizing synchrophasors and eliminating
the intermediate variables at bus P3, we can acquire
VP1i cosh(1 l1 ) IP1i Zc1 sinh(1 l1 )
VP2i cosh(2 l2 ) IP2i Zc2 sinh(2 l2 ) = 0

(40)

Eq. (40) only contains measurements at bus P1 and P2. The subscript 1 and 2 signify parameters of line segment P1P3 and P2P3,
respectively. For each set of measurements, (40) can be separated
into two real equations. There are six unknown line parameters.
Thus, three or more sets of measurements will yield solutions for
unknown parameters.
4.4. Sub-network Type 4
For double-circuit line, the afore-mentioned algorithms are
also applicable since generally there is negligible mutual coupling
between the two circuits in the positive sequence network [3].
In the conguration referring to Fig. 2(d), where the two circuits share a common bus at both terminals, we need to have
measurements at both ends of the sub-network if the two circuits
have identical line parameters; otherwise, it is possible to derive
line parameters using measurements from only one bus, as shown
below.
Based on phasors at bus P1, we can acquire the voltage phasors at bus P2, utilizing parameters of either of the parallel circuits.
Equalizing both of them results in
VP1i cosh(1 l) I1P1i Zc1 sinh(1 l)
VP1i cosh(2 l) + I2P1i Zc2 sinh(2 l) = 0

where VP1i denotes the ith positive sequence voltage phasors at

bus P1, I1P1i and I2P2i denote the ith positive sequence current phasors at bus P1 for the two circuits, respectively. l is the total line
length between bus P1 and P2. The subscript in Zc and  signies
parameters of the two circuits.
Comparing with (39) and (40), (41) does not contain any measurements at bus P2; therefore, this formulation is less complex and
more efcient. For each set of measurements, (41) can be separated
into two real equations. There are six unknown line parameters.
Thus, three or more sets of measurements will generate solutions
for unknown parameters. In addition, this sub-network only contains synchronous measurements since all measurements come
from single bus.
5. Proposed method for line temperature and sag
estimation
5.1. Line temperature estimation
Practically within the conductor thermal limit, the line resistance variation with its temperature is approximately linear [14].
Based on the estimated line resistance, as discussed in the previous section, the average temperature of the line conductor can be
determined as
T=

(R/Rref ) 1

+ Tref

Table 1
Conductor characteristics.
System frequency
Conductor type
Conductor weight per unit length
Thermal elongation coefcient
Thermo-resistivity coefcient
Horizontal tension component
Rated strength

(42)

60 Hz
ACSR, icker
= 0.614 lbs/ft
AS = 9.0718 106 /F
= 0.0033/ C
H = 25% of RBS (rated breaking strength)
17,200 lbs

where T is the estimated line temperature and R is the estimated

resistance; Rref is the line resistance at reference temperature Tref ;
is the thermo-resistivity coefcient of the conductor.
The conductor resistance at reference temperature can be
referred by conductor characteristics and the thermo-resistivity
coefcient can be calculated accordingly. The accuracy of line
temperature estimation is directly related to the accuracy of line
resistance estimation.
5.2. Line sag estimation
The method from [15] is employed to determine the line sag for
a selected span length. Firstly the reference line sag Dref for span
length  is calculated at reference temperature Tref , then the actual
conductor length LTref due to sag at Tref is derived. Meanwhile, the
estimated temperature Test is computed from the estimated line
resistance Rest based on the line temperature estimation method.
Following that, the conductor length LTest due to sag at Test is
derived. Finally the line sag Dest can be determined according to
obtained LTest . Only level span is investigated in this study and the
span length can be assigned at discretion or according to real data.
The characteristics by conductor type are available.
The following equations are employed for sag estimation:
Dref =

2
8H

LTref =  +
(41)

43

(43)
2
8Dref

3

(44)

LTest = LTref [1 + AS (Test Tref )]

(45)

3(LTest )
8

(46)

Dest =

where is the weight of conductor per unit length,  is the level

span length, H is the horizontal tension component, proportional
to the rated strength of the conductor, and AS is the coefcient
of linear thermal elongation which can be calculated referred to
conductor characteristics.
6. Case studies and results analysis
In this section, numerical case studies and representative results
are presented to demonstrate the performance of the proposed line
parameter, temperature and sag estimation methods.
The Matlab simpowersystem tool is utilized to create simulation models to construct typical transmission networks in Fig. 3
and generate voltage and current phasors during normal operations. Multiple measurement sets have been obtained by varying
the phase angle difference between the two Thevenin equivalent
sources. The per unit system is used in this study with a base voltage of 500 kV and a base power of 1000 MVA. Universal conductor
characteristics are listed in Table 1.
The characteristic for level span sag of conductor type icker
versus temperature variations is in Fig. 4, where it shows that
not only the level span sag will increase when the temperature
increases, but also longer span length will lead to larger sag. Such

44

Y. Du, Y. Liao / Electric Power Systems Research 93 (2012) 3945

200,400,600,800ft Level Span Sag Vs. Temperature for conductor "Flicker"

25
200ft
400ft
600ft
800ft

20

Parameters

R1 (p.u./km)
X1 (p.u./km)
B1 (p.u./km)
R2 (p.u./km)
X2 (p.u./km)
B2 (p.u./km)
( )
T1 ( C)
Sag1 (ft)
T2 ( C)
Sag2 (ft)

Sag(ft)

15

10

0
20

30

40

50

60

70

80

90

100

110

120

Conductor Temperature (C)

Fig. 4. Sag versus temperature of conductor icker.

Table 2
Estimation results of Sub-network Type 1.
Parameters

Real values

R (p.u./km)
X (p.u./km)
B (p.u./km)
( )
T ( C)
Sag (ft)

0.000619303
0.00146432
0.00114016
0
97.88
11.47

Estimation results
Least squares

Optimal

0.000696184
0.00146389
0.00114017
0.0774845
136.8
13.92

0.00062797
0.00141887
0.00114006
9.9123e5
102.5
11.77

Table 3
Estimation results of Sub-network Type 2 with bad synchronization ( = 5 ).
Parameters

Real values

R (p.u./km)
X (p.u./km)
B (p.u./km)
( )
T ( C)
Sag (ft)

0.000619303
0.00146432
0.00114016
0
97.88
11.47

Estimation results
Least squares

Optimal

0.00427354
0.00819419
0.00112863
0.952987
2034
55.64

0.00319199
0.00758921
0.00114527
0.016863
1461
47.10

data can be stored in a database from which we can determine the

line sag given the estimated line temperature for any selected span
length.
Estimation results for Sub-network Types 14, including positive sequence line parameters, temperature and sag by comparing
the least square method and optimal estimator for a level span of
500 ft are summarized in Tables 26. Table 2 shows the estimation
results for a case with synchronization error, which results in huge
errors in temperature and sag estimates. The synchronization angle
is identied as bad data based on the method outlined in Section

Table 4
Estimation results of Sub-network Type 2 after removal of bad synchronization.
Parameters

R (p.u./km)
X (p.u./km)
B (p.u./km)
( )
T ( C)
Sag (ft)

Table 5
Estimation results of Sub-network Type 3.

Real values

0.000619303
0.00146432
0.00114016
0
97.88
11.47

Estimation results
Least squares

Optimal

0.000698338
0.00146342
0.00114018
0.952987
139.7
13.99

0.000698354
0.00146345
0.00114018
0.047025
139.8
13.99

Real values

0.000619303
0.00146432
0.00114016
0.000629205
0.00165173
0.00125729
0
97.88
11.47
103.1
11.81

Estimation results
Least squares

Optimal

0.000616220
0.00146082
0.00121149
0.000621388
0.00164777
0.00134601
1.34634e10
96.24
11.36
98.98
11.54

0.000612071
0.00146293
0.00117007
0.000622295
0.00164979
0.00129407
1.7116e12
94.05
11.21
99.46
11.57

Table 6
Estimation results of Sub-network Type 4.
Parameters

Real values

R1 (p.u./km)
X1 (p.u./km)
B1 (p.u./km)
R2 (p.u./km)
X2 (p.u./km)
B2 (p.u./km)
( )
T1 ( C)
Sag1 (ft)
T2 ( C)
Sag2 (ft)

0.000619303
0.00146432
0.00114016
0.000629205
0.00165173
0.00125729
n/a
97.88
11.47
103.1
11.81

Estimation results
Least squares

Optimal

0.000621371
0.00146395
0.00123194
0.000621371
0.00165160
0.00122185
n/a
98.97
11.54
98.97
11.54

0.000620359
0.00146404
0.00121636
0.000629959
0.00165172
0.00124120
n/a
98.44
11.50
103.5
11.84

4.1.3. Once the bad data is eliminated, the new estimates are
provided in Table 4, and improved estimates have been obtained.
Compared to the least squares method, the optimal estimation
algorithm generally produces more or comparably accurate results,
and particularly is capable of detecting and identifying potential
bad data such as synchronization errors and thus ensuring accurate
and reliable solutions.
7. Conclusion
This paper proposes an overall framework for estimating positive sequence transmission line parameters (series resistance,
series reactance and shunt susceptance), temperature and sag by
employing on-line voltage and current phasors measured by PMUs
installed at terminals of transmission lines. Typical transmission
networks are modeled to facilitate the development of corresponding algorithms. Distributed parameters are employed to accurately
consider the shunt capacitances and long line effect. The proposed algorithms harness the non-linear optimal estimation theory
and are capable of detecting and identifying bad data, minimizing
the impact of measurement and synchronization errors and thus
signicantly improving the estimation accuracy. The line temperature and sag are calculated according to the obtained parameter
estimates and provide a clear picture of the prevailing thermal conditions of transmission lines. It is worth noting that the presented
methods do not require additional tension or weather sensors,
and thus greatly simplify implementation and reduce both equipment and maintenance costs. Numerical case studies indicate that
encouraging results have been achieved.
References
[1] S.H. Horowitz, A.G. Phadke, Power System Relaying, Research Studies Press Ltd.,
Taunton, Somerset, England, 1995.

Y. Du, Y. Liao / Electric Power Systems Research 93 (2012) 3945

[2] Y. Du, Y. Liao, Online estimation of power transmission line parameters, temperature and sag, in: Proc. 2011 North American Power Symposium, Boston,
MA, August 46, 2011, pp. 16.
[3] J. Grainger, W. Stevenson, Power System Analysis, New York, McGraw-Hill,
1994.
[4] S. Kurokawa, J. Pissolato, M.C. Tavares, C.M. Portela, A.J. Prado, A new
procedure to derive transmission-line parameters from synchronized measurements, Electric Machines and Power Systems 27 (January (1)) (2006)
492498.
[5] R.E. Wilson, G.A. Zevenbergen, D.L. Mah, A.J. Murphy, Calculation of transmission line parameters from synchronized measurements, Electric Machines and
Power Systems 27 (December (12)) (1999) 12691278.
[6] J.S. Thorp, A.G. Phadke, S.H. Horowitz, M.M. Begovic, Some applications of phasor measurements to adaptive protection, IEEE Transactions on Power Systems
3 (May (2)) (1988) 791798.
[7] J.-A. Jiang, Y.-H. Lin, J.-Z. Yang, T.-M. Too, C.-W. Liu, An adaptive PMU based
fault detection/location technique for transmission linesPart II: PMU implementation and performance evaluation, IEEE Transactions on Power Delivery
15 (October (4)) (2000) 11361146.
[8] C.S. Chen, C.W. Liu, J.A. Jiang, A new adaptive PMU based protection scheme
for transposed/untransposed parallel transmission lines, IEEE Transactions on
Power Delivery 17 (April (2)) (2002) 395404.

45

[9] S.E. Zocholl, D.W. Smaha, Current transformer concepts, in: Proc. 19th
Annu. Western Protective Relay Conference, Spokane, WA, October 2022,
1992.
[10] J.R. Linders, C.W. Barnett, J.W. Chadwick, et al., Relay performance considerations with low-ratio CTs and high-fault currents, IEEE Transactions on Industry
Applications 31 (MarchApril (2)) (1995) 392404.
[11] National Energy Technology Laboratory, U.S. Dept. of Energy, A system
view of the modern grid, Appx. B2: sensing and measurement, March
2007.
[12] P. Ramachandran, V. Vittal, On-line monitoring of sag in overhead transmission lines with leveled span, in: Proc. 2006 North American Power Symposium,
Carbondale, IL, September 1719, 2006, pp. 405409.
[13] M. Weibel, W. Sattinger, P. Rothermann, U. Steinegger, M. Zima, G. Biedenbach,
Overhead line temperature monitoring pilot project, in: CIGRE, 2006, Paris,
August 27September 1, 2006.
[14] M. Bockarjova, G. Andersson, Transmission line conductor temperature impact
on state estimation accuracy, in: IEEE Power Tech, Lausanne, Switzerland, July
15, 2007.
[15] Electric Power Research Institute (EPRI), Increased power ow guidebook:
increasing power ow in transmission and substation circuits, November 2005.
[16] A. Abur, A.G. Exposito, Power System State EstimationTheory and Implementation, Marcel Dekker, New York, 2004.