You are on page 1of 14

Electric Power Systems Research 152 (2017) 257–270

Contents lists available at ScienceDirect

Electric Power Systems Research


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

Review

A review of the enabling methodologies for PMUs-based dynamic


thermal rating of power transmission lines
Guido Coletta, Alfredo Vaccaro ∗ , Domenico Villacci
Department of Engineering (DING), University of Sannio, Piazza Roma 21, 82100 Benevento, Italy

a r t i c l e i n f o a b s t r a c t

Article history: In the last years, dynamic thermal rating assessment of overhead lines has gained a critical importance in
Received 13 March 2017 power system operation, since it allows transmission system operators to reliably increase the exploita-
Received in revised form 15 May 2017 tion of existing infrastructures, avoiding the construction of new transmission assets, and increasing the
Accepted 11 July 2017
hosting capacity of renewable power generators. Amongst the possible approaches that can be adopted
Available online 27 July 2017
to solve the thermal estimation problem, the one based on synchrophasor data processing is considered
as one of the most promising enabling technologies, since it does not require the need for deploying
Keywords:
dedicated sensing technologies distributed along the line route, but only the availability of synchronized
Dynamic thermal rating
Wide area monitoring
measurements already available in the control centers for supporting wide area power system applica-
Synchrophasors tions. Anyway, the deployment of this technology in real operation conditions is still at its infancy, and
Phasor measurement unit several open problems need to be addressed, such as the accuracy drop in low loading conditions, and the
need for properly representing and managing the data uncertainties in the thermal estimation process.
In trying to address these issues, this paper presents a comprehensive analysis of the most promising
solution methods proposed in the literature, evaluating their performances on a real case-study based
on a thermally constrained power transmission line located in the north of Italy.
© 2017 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
2. Mathematical preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
3. PMU-based methods for DTR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
3.1. Single and double measurement methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
3.2. Non-linear least square optimal estimator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
3.3. Calibration method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
3.4. Optimization-based methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
4. Case study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
4.1. DTR accuracy for different levels of data uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
4.2. DTR accuracy in a real operation scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.2.1. Single measurement method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.2.2. Non-linear least square optimal estimator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.2.3. Calibration method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.2.4. Optimization-based method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.3. Load capability estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
5. Result and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269

∗ Corresponding author.
E-mail address: vaccaro@unisannio.it (A. Vaccaro).

http://dx.doi.org/10.1016/j.epsr.2017.07.016
0378-7796/© 2017 Elsevier B.V. All rights reserved.
258 G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270

extract actionable intelligence from the large quantity of data made


List of acronyms available by the large pervasion of wide area monitoring systems
(WAMS) into transmission grids, which allow to reliable estimate
OHL Overhead transmission line the power system state by directly acquiring both current and volt-
TSO Transmission system operator age phasors in some strategical network buses [13,14]. Starting
DTR Dynamic thermal rating from these time synchronized measurements, these WAMS-based
PMU Phasor measurement unit DTR techniques try to estimate the actual conductor thermal state
PE Parameter estimation by periodically identifying the electrical parameters of the power
NLLS Non-linear least square lines [15]. According to this computing paradigm, the average line
TVE Total vector error temperature can be inferred from the knowledge of the actual
WAMS Wide area monitoring System values of the electrical line parameters, by using a proper regres-
sive model, which correlates the variations of these parameters,
especially the positive sequence resistance, to the conductor tem-
1. Introduction perature. This thermal estimation process does not require the need
for deploying dedicated sensing technologies distributed along the
The increased loading of the power components, especially the line route, since the only requested inputs for solving thermal
overhead transmission line (OHL), is one of the most challenging estimation problem are the synchronized measurements of the
issues to address in modern electrical transmission systems, which voltage and current phasors at the both ends of the monitored
are being operated closest to their loadability limits, with severe line, which are already available in the control centers for imple-
impacts on the security and reliability of the entire power network. menting wide-area power systems applications [16]. Although this
The main phenomena driving this trend include the increasing positive feature is considered a strategic benefit for the TSOs, the
power transactions due to the difference in locational marginal deployment of this technology in real operation conditions is still
prices, the massive penetration of renewable power generators, at its infancy, and several open problems need to be addressed,
and the difficulties in upgrading the existing transmission assets such as the accuracy drop in low loading conditions, and the need
due to social and environmental issues [1,2]. for properly representing and managing the data uncertainties. The
It is well known that reliable OHLs operation requires, amongst latter represents one of the most challenging research objectives to
other things, the rigorous satisfaction of proper thermal con- address in the context of phasor measurement unit (PMU)-based
straints, which limit the conductor temperature in order to keep applications [17,18], which has recently attracted the attention of
the line sag in fixed allowable ranges, avoiding the risk of ground wide-area monitoring, protection and control systems designers.
faults. Traditionally, transmission system operators (TSOs) aimed In particular, paper [19] describes four methods for the solu-
at converting these thermal constraints into a maximum allowable tion of the OHL thermal estimation problem in the presence of
value of the power transfer capability, which is obtained by solv- measurement uncertainty, based on both single and double pha-
ing the steady-state conductor thermal model using the worst-case sor measurements, and using both linear and non-linear least
values of the boundary conditions (i.e. wind speed 0.6 m/s, sun irra- square-based solution methods. The results discussed in this paper
diation 1000 W/m2 , etc.) [3,4]. This conservative approach allows demonstrate that the non-linear least square approach based on
to reliably operate the OHLs, reducing the risk of faulty conditions, multiple measurements represents the most viable solution to mit-
but at the cost of a sensible under-utilization of the transmission igate the effects of measurement errors. This approach has been
assets, since the real conductor temperature is extremely overesti- improved in [20], which allows improving the overall thermal
mated, especially when the weather conditions sensibly differ from estimation accuracy by defining proper bounds on the electrical
the worst-case profiles assumed in solving the conductor thermal parameters, in [21], where the non-linear regression technique is
model. Hence, the definition of more effective OHLs loading poli- applied for estimating the electrical line parameters of both power
cies become appealing in order to reliably improve the components lines and transformers, and in [22,23] where a new method for esti-
explotation by a more accurate assessment of their real loadability mating the line conductor temperature and the corresponding line
margins [5–8]. sag is proposed.
In this domain, the application of advanced methodologies for An alternative solution technique for uncertainty management
dynamic thermal rating (DTR) represents a very promising research has been described in [24], which proposes a two-stage solution
direction, since it could allow the on-line estimation of the actual method to reduce the effects of systematic errors by identifying
conductor temperature, and the assessment of the corresponding proper calibration coefficients, and applying a least square tech-
load capability curves, supporting the TSOs in defining reliable and nique on the calibrated phasors to estimate the uncertain line
effective OHLs loading policies. parameters.
Modern DTR architectures are based on pervasive and ubiq- Finally, Refs. [25,26] propose two reliable thermal identifica-
uitous sensing technologies, which can directly measure the tion techniques based on the solution of uncertain optimization
conductor temperature, or estimating it by periodically solving a problems, whose objective is to minimize the error between the
calibrated thermal model with the actual values of the environmen- measured and the estimated phasors. Although these optimization-
tal variables measured along the line route [6,9,10]. Although the based methods exhibited good performances in several operation
performance of these methods have been experimentally validated scenario, they suffer from severe limitations, mainly related to the
in different operation scenarios [11,12], several open problems difficulties in identifying the most suitable error function to mini-
need to be addressed in order to support their large-scale pervasion mize, and the need for large computational resources.
in existing power transmission networks. The need for deploy- On the basis of the results discussed in these papers, much of
ing pervasive and reliable peer-to-peer networks, the difficulties them based on simulation studies, it can be argued that the reli-
arising in installing the sensors on the conductor surface, and the able solution of the DTR problem by PMUs data processing can be
integrity assessment of the sensed data are some of the most critical addressed according to different techniques, which sensibly dif-
issues to address in this domain. fer in terms of accuracy, reliability, and promptness. Hence, the
To face some of these limitations, alternative solution tech- selection of the most effective technique to adopt in real opera-
niques based on synchronized data processing have been recently tion scenarios is not straightforward, and it is further complicated
proposed in the literature. The insight of these methods is to try to by the lack of a comprehensive analysis aimed at comparing the
G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270 259

Fig. 2. -Circuit model of a short OHL.

Fig. 1. -Equivalent circuit model of the OHL.


the average conductor temperature by the following approximated
equation [6]:
performance of these techniques on a common experimental case
study. r(T ) = r(Tref ) + ˛(T − Tref ) (3)
In the light of these needs, in this paper the performance of the where ˛ is the temperature-resistance regression constant, r(T) is
most advanced methods for PMUs-based DTR have been exper- the series resistance at the temperature T and, Tref is a reference
imentally assessed in the task of solving the thermal estimation conductor temperature.
problem for a complex transmission asset, which is based on a Various techniques can be adopted to solve this parameter
400 kV overhead line located in the north of Italy. A comparison estimation (PE) problem in modern PMU-based DTR assessment
between the average line temperature estimated by the analyzed procedures, and the most significant of them will be analyzed in
methods for a 6h time horizon, and the real conductor temperature the next section.
directly sensed by a measurement system installed in the critical
line span, will be presented and discussed in order to outline the
3. PMU-based methods for DTR
benefits and the limitations of the considered techniques.
3.1. Single and double measurement methods
2. Mathematical preliminaries
The simpler approach for PMU-based DTR assessment [19] is
A fully-transposed electric power line can be modeled at the based on the solution of the PE problem by manipulating the tele-
direct sequence through a concentrated parameters quadrupole, graphers’ equations. In particular, the main idea is to identify the
represented by the equivalent  circuit shown in Fig. 1. quadrupole constants by inverting (1), using the voltage and cur-
The electrical state of this circuit is described by the voltage and rent phasors measured at the both line ends. This solution paradigm
current phasors at the first line end, V̄s and Īs , which are related to is typically referred as double measurement method, since it requires
the corresponding phasors at the second line end, V̄r and Īr , by the the acquisition of two independent sets of phasor measurements,
so called telegraphers equations: namely the voltage and current phasors at both line ends.
An alternative solution approach, requiring only the knowledge
V̄s = Ȧ V̄r + Ḃ Īr
(1) of a single phasor measurement set, can be adopted for solving
Īs = Ċ V̄r + Ḋ Īr the PE problem for the so called short transmission lines, which are
overhead lines characterized by a limited length, typically less than
where the transmission parameters for the power line can be 100–150 km. In this case, it is possible to approximate the hyper-
defined as follows: bolic functions in the transmission parameter equations by their
 
first order Taylor’s series terms, obtaining the simplified model
Ȧ = cosh (r + jωl) (g + jωc)
shown in Fig. 2, which is described by the following equations:
    
Ḃ = (r + jωl)/(g + jωc) sinh (r + jωl) (g + jωc) Ż Ẏ
V̄s = 1+ V̄r + Ż Īr
(2) 2
1
    (4)
Ċ =  sinh (r + jωl) (g + jωc)
Īs = Ẏ V̄r + 1+
Ż Ẏ
Īr
(r + jωl)/ (g + jωc) 2
Ḋ = Ȧ
where Z = (r + jωl) * L, Y = (g + jωc) * L and L is the line length. Since the
These parameters depend on the angular frequency ω = 2f, and obtained model is a two variables-two equations system, its inver-
the primary line constants per unit length, namely, the series resis- sion requires only the knowledge of 4 complex quantities, namely
tance r, the series inductance l, the shunt conductance g, and the a single phasor measurement set. Hence, this approach is typically
shunt susceptance c. The latter depend on both the overhead line referred as single measurement method.
characteristics, e.g. the pole geometry, and the conductor thermal Both the single and the double measurement methods allow
state, especially the conductor temperature. obtaining reasonable estimation accuracy when input data are
Consequently, once the voltage and current line phasors are accurate enough, but their performances tend to degenerate in real
known, the electrical line parameters can be estimated by manipu- operation scenario, due to the lack of intrinsic tools for data uncer-
lating the model formalized in (1), and the corresponding conductor tainty modeling. Although the use of redundant measurements can
thermal state can be inferred by using proper regression models, be adopted in trying to increase the performances of these estima-
describing the correlations of the electrical line parameters with the tion algorithms, by reducing the effect of random measurement
average conductor temperature. To this aim, the parameter char- errors, the tolerance of the estimated line parameters could not
acterized by the highest sensitivity to the conductor temperature be enough to reliably estimate the average conductor temperature
is known to be the series resistance, which can be correlated to with an acceptable level of confidence.
260 G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270

3.2. Non-linear least square optimal estimator 3.3. Calibration method

A challenging idea aimed at solving the parameter estimation Errors in PMU measurements are not only induced by ran-
(PE) problem in the presence of data uncertainty by processing dom errors, which can be effectively managed by the previously
redundant measurements is to design estimation techniques based described techniques, but also by systematic errors, mainly affect-
on the non-linear least square (NLLS) theory [19]. In particular, ing the measurement transformers. To face this issue, in [24] a
given a set of N measurements, Eq. (1) can be written as follows: solution method based on a calibration process is proposed. The
⎧ insight principle is to model systematic errors in phasor mea-
⎪ R(V̄s ) = R(Ā)R(V̄r ) − I(Ā)I(V̄r ) + R(B̄)R(Īr ) − I(B̄)I(Īr )

⎪ surements as an error in magnitude and an error in phase, hence

⎨ I(V̄s ) = R(Ā)I(V̄r ) + I(Ā)R(V̄r ) + R(B̄)I(Īr ) + R(B̄)I(Īr ) expressing each measure phasor as follows:
(5)

⎪ R(Īs ) = R(C̄)R(V̄r ) − I(C̄)I(V̄r ) + R(D̄)R(Īr ) − I(D̄)I(Īr ) P̄ = P̄ ∗ (1 + a)ej (10)



I(Īs ) = R(C̄)I(V̄r ) + I(C̄)R(V̄r ) + R(D̄)I(Īr ) + I(D̄)R(Īr )
where P̄ ∗ ∈ C is the phasor measured by the PMU and a, ∈ R are
where R( · ) and I(·) denote the real and imaginary part of each the corresponding phase and magnitude errors, respectively.
complex quantity. Starting from these equations, and knowing the Since the phase errors are typically less than 0.530◦ , in order to
measured line phasors, it is possible to compute the corresponding satisfy the accuracy requirements of the IEEE Std. C37.118.1.2011
residuals, namely the difference between the measured and the for synchrophasor measurements [27,28], small angle approxima-
theoretical phasors for each set of line parameters, as follows: tion for trigonometric functions can be used, obtaining:

⎪ f1 (x, ) = R(V̄s )−R(Ā)R(V̄r ) + I(Ā)I(V̄r )−R(B̄)R(Īr ) + I(B̄)I(Īr ) ej ≈ 1 + j (11)



⎨ f2 (x, ) = I(V̄s )−R(Ā)I(V̄r )−I(Ā)R(V̄r )−R(B̄)I(Īr )−R(B̄)I(Īr )
Hence, Eq. (10) becomes:

⎪ f3 (x, ) = R(Īs )−R(C̄)R(V̄r ) + I(C̄)I(V̄r )−R(D̄)R(Īr ) + I(D̄)I(Īr )

⎪ P̄ = P̄ ∗ (1 + a)(1 + j) = P̄ ∗ (1 + a + j + ja)
⎩ (12)
f4 (x, ) = I(Īs )−R(C̄)I(V̄r )−I(C̄)R(V̄r )−R(D̄)I(Īr )−I(D̄)R(Īr )
(6) and, by neglecting the second order error terms, it follows that:

P̄ = P̄ ∗ (1 + a)(1 + j) = P̄ ∗ (1 + a + j) (13)


Then, once indicating with x and  the vectors of the syn-
chrophasor measurements, and the line parameters, respectively, Starting form this result, all the phasors at the both line ends
as follows: can be expressed as:

x = R(V̄s ) I(V̄s ) R(Īs ) I(Īs ) R(V̄r ) I(V̄r ) R(Īr ) I(Īr )
V̄s = V̄s∗ (1 + as + js )
V̄r = V̄r∗ (1 + ar + jr )
= r l c g (14)
Īs = Īs∗ (1 + bs + js )
Eq. (6) can be generalized for a set of N phasor measurements
as follows: Īr = Īr∗ (1 + br + jr )
⎡ ⎤ ⎡ ⎤
f11 (x, ) 11 (x, ) and the total errors in phasor measurements can be expressed as:
⎡0⎤ ⎢ f 1 (x, ) ⎥ ⎢ 1 (x, ) ⎥
⎢2 ⎥ ⎢ 2 ⎥
⎢0⎥ ⎢ ⎥ ⎢ ⎥ ıV̄s = V̄s∗ (as + js )
⎢ ⎥ ⎢ f3 (x, ) ⎥ ⎢ 3 (x, ) ⎥
⎢ 1 ⎥ ⎢ 1

⎢0⎥ ⎢
⎢ ⎥ ⎢ f 1 (x, ) ⎥ ⎥
⎢ 1



ıV̄r = V̄r∗ (ar + jr )
⎢0⎥ ⎢ 4 ⎥ ⎢  (x, )
⎥ (15)
⎢ ⎥ ⎢ ⎥ ⎢
4
⎥ ıĪs = Īs∗ (bs + js )
⎢ .. ⎥ ⎢ . ⎥ ⎢ ⎥ = F(x, ) + 
⎢ . ⎥ = ⎢ .. ⎥ +⎢..
⎥ (7)
⎢ ⎥ ⎢ ⎥ ⎢ .
⎥ ıĪr = Īr∗ (br + jr )
⎢0⎥ ⎢ N ⎥ ⎢ N (x, ) ⎥
⎢ ⎥ ⎢1 f (x, ) ⎥ ⎢  ⎥
⎢0⎥ ⎢ 1
⎢ ⎥ ⎢ f N (x, ) ⎥ ⎥
⎢ N



If a single set of phasor measurements are available, the sin-
⎣0⎦ ⎢ 2 ⎥ ⎢ 2 (x, ) ⎥ gle measurement method is applied, in order to provide a first
⎢ N ⎥ ⎢ N ⎥ approximation of the line parameters, trough the inversion of the
0 ⎣ f3 (x, ) ⎦ ⎣ 3 (x, ) ⎦
telegraphers’ equations:
f4N (x, ) N4 (x, )
V̄s2 − V̄r2
Thanks to this mathematical formulation, it is possible to esti- Ż =
V̄s Īr2
2 + V̄r2 Īs2
mate the line parameters by identifying  that minimizes the (16)
residual of F(x, ), . This scalar minimization problem can be solved 2(Īs − Īr )
Ẏ =
using a NLLS regression technique, by deploying the following iter- V̄s + V̄r
ative procedure:
The first order approximation of the corresponding estimation
 k+1 =  k +  k (8) errors can be expressed as:
−1
 k = (HT H) HT (−F(x,  k )) (9) ∂Ż ∂Ż
ıŻ = ıV̄s + ıV̄r
where H is the Jacobian matrix of F(x,  k ). This iteration process
∂V̄s ∂V̄r
(17)
is typically interrupted when the variable update, ␪k , results ∂Ẏ ∂Ẏ
ıẎ = ıĪs + ıĪr
smaller than a specified tolerance. ∂Īs ∂Īr
G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270 261

Hence, by assuming that the overall measurement errors can be Table 1


Main line characteristics.
defined as:
a = ar − as Phase conductors 2
Conductor type ACSR
 = r − s Length 112.2 km
(18) Nominal voltage 400 kV
b = br − bs Conductor diameter 31.5 mm
Positive sequence resistance 3.15 
 = r − s Positive sequence reactance 38.2 
Positive sequence susceptance 2 × 191.5 × 10−6 s
where, according to the argumentation presented in paper, the
maximum allowable ranges for these error functions can be esti-
mated as: of the squared errors between the measured and the estimated
|a| < 0.02 phasors, as detailed in the following equations:
|| < 0.02
(19) min [E (|V̄ |) + E (∠V̄ ) + E (|Ī|) + E (∠Ī)] (24)
|b| < 0.02 [r,xl ,g,b]

|| < 0.02 where


Starting from these results, it is possible to compute the corre- N  2
 |V̄ |s,i − |V̄ |s,i,estimated
sponding relative errors in Z̄ and Ȳ estimation as follows: E (|V̄ |) = (25)
  |V̄ |s,i
ıZ̄ −2(ar − as + jr − js )V̄r∗2 i=1
=
Z̄ app
V̄s∗2 − V̄r∗2 N 
 2
  (20)
E (∠V̄ ) =
∠V̄s,i − ∠V̄s,i,estimated
(26)
ıȲ −2(br − bs + jr − js )Īr ∠V̄s,i
=
Ȳ app
V̄s∗ + V̄r∗ i=1

N 
 2
This important result allows computing the values of the uncer- |I|s,i − |I|s,i,estimated
tain variables a, b,  and , which minimize, on a fixed time window, E (|Ī|) = (27)
|I|s,i
the time integral of the square error between the measured and the i=1

estimated phasors at both line ends. N 


 2
In solving this problem, a linear drift of both line resistance ∠Īs,i − ∠Īs,i,estimated
E (∠Ī) = (28)
and reactance is assumed, while conductance and susceptance ∠Īs,i
i=1
are assumed constant during the time window. Thanks to these
hypothesis, which are consistent with the assumptions defined in and N is the number of measured phasors used for estimating each
IEEE Std 738-2012 [7], it is possible to represent the primary con- line parameter set.
stants R = r * L, X = x * L, G = g * L and B = b * L as follows: Similarly to the previously described NLLS methods, this solu-
fR (t) = qR t + rR tion approach tries to minimize the time integral of the squared
error between the measured and the estimated phasors at both line
fX (t) = qX t + rX ends, but rather than solving the PE problem by using a two-step
(21) optimization procedure, which firstly identifies the phasor calibra-
fG (t) = rG
tion constants and then estimates the line parameters, it employs
fB (t) = rB a single optimization procedure, without doing any assumption on
the time evolution of the line parameters.
where the quantities qR , qX , rR , rX , rG and rB are the coefficients of the
line parameter models estimated through a least square procedure.
Consequently, the unknown phasor calibration parameters can 4. Case study
be estimated by solving the following non-linear, constrained opti-
mization problems: In order to characterize the performances of the described meth-
ods in the task of solving complex thermal estimation problems
min gZ (a, ) = SR + SX in a real operation scenario, an experimental case study has been
[a,]
{ |a| < 0.02 (22) considered. To this aim, the data acquired from the Italian wide
area monitoring system [13] for a 400 kV overhead line over a 6 h
|| < 0.02
time window have been used. This is a strategic asset for the Ital-
min gY (b, ) = SG + SB ian power transmission network, since it interconnects two critical
[b,]
system sections, characterized by “large” power transactions. The
{ |b| < 0.02 (23)
main characteristics of the line under study are summarized in
|| < 0.02 Table 1.
Once a, b, ,  have been identified, the corrected phasors can be The measured data include voltage and current phasors at both
computed and a conventional least square procedure can be applied ends of the considered line sampled each 50 ms (Figs. 3 and 4), and
to solve the PE problem as described in the previous section. the conductor temperature of the most critical line span sampled
each 1 min by a high accurate temperature sensor.
3.4. Optimization-based methods The available data are organized as:

Alternative solution approaches recently proposed in the liter- - four vectors of complex number containing Vr , Vs , Ir and Is mea-
ature [25,26] to solve the PE problem are based on the deployment surements, with a cardinality of [1,080,000 × 1];
of optimization frameworks aimed at identifying the set of line - a vector of time tags associated with the PMU measurement, with
parameters, which minimize, on a fixed time window, the sum the same cardinality.
262 G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270

Fig. 3. PMU measurements-starting section.

Fig. 4. PMU measurements-receiving section.

The first issue to address in solving the PE problem is to fix Table 2


Computational time.
the time window, namely the number of data samples that should
be adopted in estimating the line parameter. To this aim, various Single measurement method 0.15 Seconds
options have been considered: NLLS method 6 Minutes
Calibration method 1.6 Hours
Optimization method 13 Minutes
- estimation of one parameter-set every N measurement;
- estimation of one parameter-set every N measurement through
which is based on the Intel(R) Core(IM) i7-4510 dual core processor,
a moving time window; ®
equipped by the Matlab environment.
- estimation of one parameter-set every Nk measurement through
Starting from these data, two experimental sessions have been
a moving time window, where Nk is a set of data containing the
implemented: the first one was aimed at assessing the performance
average phasors computed every k measurements.
of the analyzed methods in the task of solving the PE estimation
- hybrid approach – estimation of one parameter every N measure-
problem for different level of data uncertainty, which was gener-
ment, through a moving window, every k measurement.
ated by properly merging the real measurements with simulation
data, while the second session was aimed at testing the DTR meth-
In this context, the firsts experimental results demonstrated ods on a real case study, by using only the experimental data.
that the application of the last approach is the most effective option,
representing a good trade-off between estimation accuracy and 4.1. DTR accuracy for different levels of data uncertainty
computational burden. Hence, it has been selected for developing
the experiments discussed in this paper. To generate realistic data-sets corrupted by a fixed level of
This conclusion can be confirmed by analyzing the data summa- data uncertainty, a reverse-engineering based procedure has been
rized in Table 2, which reports the computational times requested adopted. The main idea is to start from the conductor temperature
to solve the PE problem by using the analyzed solution techniques. profile measured by the sensor to obtain a realistic evolution of the
These data have been observed on the same processing system, line parameters for a fixed time window. Then, by using these pro-
G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270 263

Fig. 5. TVE histogram (with 0.005 uncertainty level).

files and the measured phasors at one line end, which have been ing processing did not match the accuracy requirements of DTR
assumed with no uncertainty, the corresponding phasors at the applications.
other line end have been computed by using (1).
These two phasor sets, which have been assumed as the “true” 4.2.2. Non-linear least square optimal estimator
data-set on both line ends, have been corrupted by three levels of To test this method, a set of 50 samples for each phasor at
data uncertainty, applying the concept of total vector error (TVE). both line ends is used for estimating a single line parameter set,
The latter represents one of the most important figures of merit using a moving window-based approach for solving the estimation
characterizing PMUs accuracy. It is defined in [27] as: problem at each time step. The corresponding results for the ana-
 lyzed time window are organized in a matrix composed by 21,600
(R(At ) − R(A))2 + (I(At ) − I(A))2 rows, each one representing a set of estimated line parameters,
TVE = (29)
R(At )2 + I(At )2 that are time averaged in order to estimate the conductor temper-
ature every 5 min. Figs. 11 and 12 depict the line parameters and
where At is the true synchrophasor, A is the sensed phasor, R( · ) and the corresponding conductor temperature estimated through the
I(·) denotes the real and imaginary part, respectively. This param- application of this methodology.
eter is used in the cited IEEE standard as a quality index of the
synchrophasors measurements in power systems. 4.2.3. Calibration method
Errors in measures have been modeled in both phase and mag- This method allows managing both statistical and systematic
nitude, generating four independent random Gaussian variables, errors affecting synchrophasor measurements. Correction constant
without doing any assumption on the correlation between the mea- identified for the 4 phasorial quantities are shown in Fig. 13 and the
surements errors of the same instrument channel. The considered corresponding temperature profile, which has been averaged on
level of uncertainty corresponds to average TVE of 0, 0.005 and 0.01, 1 min time windows, is shown in Fig. 15. Line parameters, instead,
respectively. are shown in Fig. 14.
The corresponding histograms of the TVE are reported in
Figs. 5 and 6, for TVE levels of 0.005 and 0.01, respectively. Start- 4.2.4. Optimization-based method
ing from these data-sets, the analyzed methods have been applied The optimization-based methodology has been applied in the
obtaining the results summarized in Figs. 7–9, which report the task of solving the following NLP programming problem:
conductor temperature measured at the critical span, and the cor- ⎧  
⎪ min ∗ [E (|V̄ |) + E (∠V̄ ) + E (|Ī|) + E (∠Ī)]
responding estimated profiles, for average TVE of 0, 0.005 and 0.01 ⎪
⎪ l
respectively. In this context, it is worth noting that, as expected, ⎪ [r,x ,b]
⎨ 0 < r ∗ < inf
the calibration method failed to solve the PE problem, identifying (30)

⎪ 0.9 xLnom < xl∗ < 1.1 xLnom
inconsistent temperature profiles, for TVE greater than 0.005. ⎪

⎩ ∗
0.9 bnom < b < 1.1 bnom
4.2. DTR accuracy in a real operation scenario
where * is a proper scaling factor, and E (|V̄ |), E (|Ī|), E (∠V̄ ) and
4.2.1. Single measurement method E (∠Ī) are the error functions defined in Section 3.4.
The result obtained by applying this method to solve the PE The latter optimization problem has been solved through an
problem for the analyzed time window are shown in Fig. 10, interior-point based method, using the built-in routines available
®
which depicts the measured and the estimated temperature pro- in the optimization toolbox of the Matlab environment.
files. These data have been obtained by averaging the conductor Line parameter and conductor temperature estimated through
temperature over a time window of 5 min, in order to compen- this method are depicted in Figs. 16 and 17, respectively.
sate the errors by using redundant measurements. The deployment
of proper data filtering functions is strictly recommended for the 4.3. Load capability estimation
correct application of this estimation method on real measured
data. This was also confirmed by the experimental results obtained The knowledge of the average conductor temperature is a pre-
on the analyzed case study, where it has been observed that the requisite for assessing the real loadability margins of the monitored
temperature profiles estimated without performing the averag- transmission asset on a short-medium time horizon. To this aim,
264 G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270

Fig. 6. TVE histogram (with 0.01 uncertainty level).

Fig. 7. Estimated conductor temperature TVE = 0.

Fig. 8. Estimated conductor temperature TVE = 0.005.

the dynamic assessment of the load capability curve, reporting the To address this issue, an indirect method for weather parame-
maximum allowable time for each hypothetical line current, repre- ters estimation has been applied in this paper. The rationale is to
sents one of the most useful tools. Anyway, this computing process estimate the wind speed, the wind direction, and the ambient tem-
is a very complex issue to address, since it requires the knowledge of perature by solving an optimization problem that minimizes, on
the actual conductor temperature, the worst-case estimation of the a fixed time window, the prediction error between the estimated
weather variables along the line route, and the repetitive solution conductor temperature, and the corresponding one obtained by
of a detailed conductor thermal model. applying a first order thermal model [7]. Once the actual weather
parameters have been estimated, and the maximum allowable con-
G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270 265

Fig. 9. Estimated conductor temperature TVE = 0.01.

Fig. 10. Estimated conductor temperature – single measurement method.

Fig. 11. Estimated parameter – NLLS method.

ductor temperature has been fixed, the load capability curve is dure summarized in Fig. 18. The obtained results for a fixed thermal
computed by solving the aforementioned thermal model for the state, and a maximum conductor temperature of 90◦ , have been
overall set of hypothetical load currents according to the proce- summarized in Fig. 19.
266 G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270

Fig. 12. Estimated conductor temperature – NLLS method.

Fig. 13. Correction coefficients.

Fig. 14. Estimated parameters – calibration method.


G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270 267

Fig. 15. Estimated conductor temperature – calibration method.

Fig. 16. Estimated parameters – optimization method.

Fig. 17. Estimated conductor temperature – optimization method.

These results confirmed the sensible upset of the conductor vary from 800 to 1200 A in function of the particular solution tech-
estimation accuracy on the loadability assessment. In particular, nique adopted to solve the PE problem. This effect is also evident
by observing the vertical asymptotes of the load capability curves, for the other operation points in the load capability curve, hence
which represents the static thermal ratings, it can be noted as they
268 G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270

Fig. 18. Capability curve assessment.

Fig. 19. Capability curve.

confirming the strategic importance of implementing a reliable and Moreover, in terms of conductor temperature accuracy, both
accurate method for DTR assessment. Optimization-based and single measurement methods exhibited
better performances compared to the other considered techniques,
both in the simulated and the real-time application scenario, but
5. Result and discussion as outlined in Fig. 20, their performances tend to deteriorate when
line current drops below under certain threshold values. On the
Result obtained by the application of the various described other hand, the numerical stability of both the NLLS and the cal-
methodologies allows us to make some useful considerations. In ibration methods is less sensitive to the line current, but their
particular, by analyzing Fig. 7 it is worth observing the sensible estimation accuracy is lower. In particular, the calibration method
impacts of input data uncertainty on the accuracy of the estimated does not provide feasible temperature values in the simulated
line temperature profiles, which lead the calibration method to
completely fail in identifying feasible solutions to the PE problem.
G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270 269

Fig. 20. Temperature profiles.

scenario, giving acceptable, although overestimated, results in real- perature by processing the real data sensed by the PMUs. The results
case applications. obtained demonstrated that both systematic and random mea-
Another interesting argumentation emerging from the analysis sured data influences the accuracy performances of the analyzed
of the obtained results concerns with the values of the estimated estimation algorithms to a considerable extent, and the selection
parameters. In this context, the calibration method provides esti- of the most effective paradigm for data uncertainty management
mations of the shunt susceptance, which are very close to the in PMU-based DTR is still an open problem, requiring further inves-
nominal value, and estimations of the series line parameters, which tigations. Further experimental activities, based on longer time
sensible differ from the corresponding nameplate values. This horizons, higher loading levels, and extremely variable weather
feature is consistent with previous experimental studies, which conditions, are currently under investigation by the Authors in
demonstrated that the series parameters exhibit larger sensitivity order to confirm these conclusive remarks on different, and more
to the line temperature variations. complex, operation scenario.
The analyzed methodologies have been originally conceptu-
alized for estimating the electrical line parameters. Anyway, as
demonstrated by the experimental results obtained in this paper, References
and confirmed by other field studies reported in the literature,
the deployment of these methodologies in real operation sce- [1] E.M. Carlini, C. Pisani, A. Vaccaro, D. Villacci, Dynamic line rating monitoring
nario is not straightforward, and requires the application of proper in WAMS: challenges and practical solutions, 2015 IEEE 1st International
Forum on Research and Technologies for Society and Industry Leveraging a
data miming techniques aimed at mitigating the effects of data Better Tomorrow (RTSI) (2015) 359–364.
uncertainty, such as missing data, outliers, incoherent samples and [2] E. Carlini, S. Favuzza, S. Giangreco, F. Massaro, C. Quaciari, Uprating an
filtering other kind of defective data. These uncertainty sources overhead line. Italian TSO applications for integration of RES, 2013
International Conference on Clean Electrical Power (ICCEP), IEEE (2013)
can affect the accuracy of the estimated parameters, leading to
470–475.
unrealistic results. To address this issue, the role of data pre- [3] J. Heckenbergerova, P. Musilek, K. Filimonenkov, Assessment of seasonal
processing techniques has been extensively emphasized in the static thermal ratings of overhead transmission conductors, Power and
literature only for specific power system application, such as PMU- Energy Society General Meeting, 2011 IEEE, IEEE (2011) 1–8.
[4] V. Morgan, The thermal rating of overhead-line conductors: Part I. The
based electromechanical oscillation analysis, while only standard steady-state thermal model, Electr. Power Syst. Res. 5 (2) (1982) 119–139.
outlier detection techniques has been deployed for PMU-based line [5] S. Jupe, M. Bartlett, K. Jackson, Dynamic thermal ratings: the state of the art,
parameter estimation, to the best of our knowledge. Anyway the 21st Int. Conf. on Electricity Distribution (2011).
[6] J. Iglesias, G. Watt, D. Douglass, V. Morgan, R. Stephen, M. Bertinat, D. Muftic,
adoption of these techniques did not allow us to obtain a reliable R. Puffer, D. Guery, S. Ueda, et al., Guide for Thermal Rating Calculations of
parameter estimation. Hence, this is an open problem, which asks Overhead Lines, Cigré, 2014.
for further research efforts. [7] IEEE Standard for Calculating the Current-Temperature Relationship of Bare
Overhead Conductors, IEEE Std 738-2012 (Revision of IEEE Std 738-2006 –
Finally, it is possible to note that the line temperature estimated Incorporates IEEE Std 738-2012 Cor 1-2013), 2013, pp. 1–72.
by applying the calibration procedure is highly conservative, which [8] M. Zafran, M.N. Arbab, I. Ahmad, M. Khan, A case study on alleviating electric
can be considered as an advantage, since it reduces the risk of ther- transmission congestion using dynamic thermal rating methodology, 2014
International Conference on Energy Systems and Policies (ICESP), IEEE (2014)
mal line overloading, but at the cost of an asset under-utilization. 1–6.
[9] E. Carlini, F. Massaro, C. Quaciari, Methodologies to uprate an overhead line.
6. Conclusion Italian TSO case study, J. Electr. Syst. 9 (4) (2013) 422–439.
[10] R. Stephen, J.-L. Lilien, D. Douglass, M. Lancaster, G. Biedenbach, G. Watt, R.
Pestana, P. Ferrières, M. Schmale, et al., Guide for Application of Direct
In this paper a review of the most promising approaches for Real-Time Monitoring Systems, Cigré, 2012.
PMU-based conductor temperature estimation of overhead lines [11] G. Kosec, M. Maksić, V. Djurica, Dynamic thermal rating of power lines –
model and measurements in rainy conditions, Int. J. Electr. Power Energy Syst.
have been presented, and their performance have been assessed 91 (2017) 222–229.
and compared by using experimental errors obtained from a 400 kV [12] D.A. Douglass, A.-A. Edris, Real-time monitoring and dynamic thermal rating
power line located in the North of Italy. The experimental ses- of power transmission circuits, IEEE Trans. Power Deliv. 11 (3) (1996)
1407–1418.
sions were aimed at both characterizing the uncertain propagation [13] D. Cirio, D. Lucarella, G. Giannuzzi, F. Tuosto, Wide area monitoring in the
for fixed values of the total vector error, and the accuracy of the Italian power system: architecture, functions and experiences, Eur. Trans.
analyzed techniques in the task of estimating the average line tem- Electr. Power 21 (4) (2011) 1541–1556.
270 G. Coletta et al. / Electric Power Systems Research 152 (2017) 257–270

[14] V. Terzija, G. Valverde, D. Cai, P. Regulski, V. Madani, J. Fitch, S. Skok, M.M. [21] C. Borda, A. Olarte, H. Diaz, PMUB based line and transformer parameter
Begovic, A. Phadke, Wide-area monitoring, protection, and control of future estimation, Power Systems Conference and Exposition, 2009. PSCE’09.
electric power networks, Proc. IEEE 99 (1) (2011) 80–93. IEEE/PES, IEEE (2009) 1–8.
[15] A. Phadke, J. Thorp, History and applications of phasor measurements, Power [22] C. Indulkar, K. Ramalingam, Estimation of transmission line parameters from
Systems Conference and Exposition, 2006. PSCE’06. 2006 IEEE PES, IEEE measurements, Int. J. Electr. Power Energy Syst. 30 (5) (2008) 337–342.
(2006) 331–335. [23] Y. Du, Y. Liao, On-line estimation of transmission line parameters,
[16] J. De La Ree, V. Centeno, J.S. Thorp, A.G. Phadke, Synchronized phasor temperature and sag using PMU measurements, Electr. Power Syst. Res. 93
measurement applications in power systems, IEEE Trans. Smart Grid 1 (1) (2012) 39–45.
(2010) 20–27. [24] D. Ritzmann, P.S. Wright, W. Holderbaum, B. Potter, A method for accurate
[17] J. Tang, J. Liu, F. Ponci, A. Monti, C. Muscas, S. Sulis, Impact of synchrophasor transmission line impedance parameter estimation, IEEE Trans. Instrum.
measurement uncertainty on detecting voltage stability margin in power Meas. 65 (10) (2016) 2204–2213.
systems, 2012 3rd IEEE PES International Conference and Exhibition on [25] C. Pisani, A. Vaccaro, D. Villacci, Conceptualization and experimental
Innovative Smart Grid Technologies (ISGT Europe), IEEE (2012) 1–6. deployment of an adaptive synchronized sensing system for power line
[18] M. Asprou, E. Kyriakides, The effect of time-delayed measurements on a thermal monitoring, IEEE Trans. Ind. Inf. 12 (6) (2016) 2158–2165.
PMU-based state estimator, PowerTech, 2015 IEEE Eindhoven, IEEE (2015) [26] A. Dán, D. Raisz, Estimation of transmission line parameters using wide-area
1–6. measurement method, 2011 IEEE Trondheim PowerTech (2011) 1–6.
[19] D. Shi, D.J. Tylavsky, N. Logic, K.M. Koellner, Identification of short [27] IEEE Standard for Synchrophasor Measurements for Power Systems, IEEE Std
transmission-line parameters from synchrophasor measurements, Power C37.118. 1-2011 (Revision of IEEE Std C37. 118-2005), 2011, pp. 1–61.
Symposium, 2008. NAPS’08. 40th North American, IEEE (2008) 1–8. [28] K. Narendra, D.R. Gurusinghe, A. Rajapakse, Dynamic performance evaluation
[20] X. Zhao, H. Zhou, D. Shi, H. Zhao, C. Jing, C. Jones, On-line PMU-based and testing of phasor measurement unit (PMU) as per IEEE C37. 118.1
transmission line parameter identification, CSEE J. Power Energy Syst. 1 (2) standard, Doble Client Committee Meetings Int. Protect. Testing Users Group
(2015) 68–74. (2012).

You might also like