Study of different mathematical approaches in

determining the dynamic rating of overhead power lines

and a comparison with real time monitoring data

P. Castro∗, A. Arroyo, R. Martinez, M. Manana, R. Domingo, A. Laso, R.

Lecuna
University of Cantabria. Department of Electrical and Energy Engineering
Av. Los Castros s/n. 39005. Santander, Spain

Abstract

Electricity generation is changing as new, renewable and smaller genera-

tion facilities are created, and classic topologies have to accommodate this dis-
tributed generation. These changes lead to the creation of smart grids in which
advanced generation, information and communication technologies are needed.
Information metering is important, and one of the most important grid pa-
rameters to be measured and controlled is the temperature of overhead conduc-
tors due to their relation to the maximum allowable sag of the line.
The temperature and current of an overhead conductor and the weather
conditions surrounding the cable are measured every 8 minutes for more than
a year. With these data, the accuracies of the different algorithms presented
in the standards (CIGRE TB601 and IEEE 738) are studied by implementing
them in MATLAB .
The use of precise measurements of solar radiation and low wind speeds with
ultrasonic anemometers, improves the accuracy of the estimated temperature
compared with the real measured conductor temperature. Additionally, using
dynamic algorithms instead of assuming a steady state analysis increases the
accuracy. However, an equilibrium between the accuracy and mathematical

∗ Corresponding author. Tel.: +34-942202065; fax: +34-942201385.

E-mail address: [email protected]

 complexity should be obtained depending on the specific needs.
Keywords: Thermal rating, ampacity, overhead line temperature, weather
parameters, real-time monitoring

1. Introduction

The current needs of global energy management are increasingly challenging

due to the fast changes of our society. The electric sector has to address the
addition of new and renewable sources of energy to the energy mix and to be able
5 to include them into the grid, while maintaining the principles of robustness,
security and reliability. Electricity generation is also changing, and as new and
smaller generation facilities are created, classic topologies have to accommodate
the distributed generation [1]. On the other hand, energy consumers are moving
from being passive to active by increasing their interactions with energy systems.
10 All of these changes point to the creation of smart grids, in which advanced
generation, information and communication technologies are needed [2].
Information metering is one of the critical points of these smart grids, and
accurate knowledge of the electric grid state and environmental conditions of
the surroundings are crucial for operating the line as efficiently as possible [3].
15 One of the most important grid parameters to be measured and controlled is the
temperature of the overhead conductors due to their relation with the maximum
allowable sag of the line.
This paper presents the results of real-time monitoring of the environmental
conditions, current and temperature of an overhead power line for more than a
20 year and compares these results with the different algorithms presented in the
standards CIGRE TB601 and IEEE 738 [4, 5] to estimate the thermal rating
and temperature of the conductor. This study is focused on the influence of the
accuracy of the parameters involved in the thermal rating equations and the way
in which they are implemented in the algorithms. The MATLAB software is
25 used to solve the thermal rating and calculate the temperature of the conductor.

2
 2. Thermal models and mathematical approaches

The thermal behaviour of an overhead conductor is obtained as the balance

of gained and lost heat due to the weather conditions around the conductor
and its electrical load [6]. The main sources of gained heat come from Joule
30 heating, including magnetic effects, and solar radiation. On the other hand,
the principal sources of lost heat are convection and cooling radiation to the
surroundings. The detailed expressions that are used to calculate each contri-
bution are obtained from the standard CIGRE Dec. 2014 [4] and explained as
follows:

Joule Heating (Pj ). The Joule heating gain per unit length for conductors is
obtained from:
Pj = ksk I 2 Rdc (1)

35 where ksk is the skin effect factor, I is the RMS conductor current and Rdc is
the direct current resistance per unit length.

Magnetic Heating (Pm ). A steel-cored conductor causes heating in the steel

core (Pcore ) and heating due to the redistribution of the current densities in the
layers of non-ferrous wires (Predis ).

Pm = Pcore + Predis (2)

The magnetic effects are only relevant for steel-cored conductors with one or
three aluminium layers and high current densities.

Solar Heating (Ps ). The solar heating per unit length is estimated by the
standard as:
  π   π 
Ps = αs DIt = αs D Ib sin(η) + F sin(Hs ) + Id 1 + F (3)
2 2

where αs is the absorptivity of the conductor surface, It is the global radiation

40 intensity, D is the outside diameter of the conductor, η is the angle of the solar
beam with respect to the axis of the conductor, F is the albedo, Hs is the solar

3
 altitude, Id is the diffuse sky radiation to a horizontal surface and Ib is the direct
solar radiation on a surface normal to the sun’s beam. It can be estimated by
considering the worst case situation using its maximum expected value or can
45 be directly measured with a pyranometer.

Convective Cooling (Pc ). The convective heat loss can be expressed as a

function of the dimensionless Nusselt number (N u) as follows:

Pc = πλf (Ts − Ta ) N u (4)

where λf is the thermal conductivity of air, Ts is the conductor surface tem-

perature and Ta is the ambient temperature. Depending on the type of air flow
and speed and direction of wind, different Nusselt correlations are used by the
standard.

Radiative Cooling (Pr ). Applying the Stefan-Boltzmann law, the heat loss
from the conductor due to radiation can be expressed as:
 
4 4
Pr = πDσB ǫs (Ts + 273) − (Ta + 273) (5)

50 where σB is the Boltzmann constant and ǫs is the emissivity of the conductor.

To calculate the estimated conductor temperature using the heating and
cooling contributions Eq. (1, 2, 3 and 4), two assumptions can be made, a
Steady State Analysis (SSA) or a Time Dependent Analysis (TDA) (Fig. 1).

4
 PARAMETERS

REAL TIME
MONITORING VARIABLES
(ts=8')

STEADY STATE TIME DEPENDENT

ANALYSIS (SSA) ANALYSIS (TDA)

TRANSIENT ANALYTICAL EQUATION

HEAT BALANCE FOR TRANSIENT STATE
(TRACKING) (DYNAMIC)

STEP STEP OR LINEAR

FUNCTION FOR FUNCTION FOR
ALL VARIABLES THE VARIABLES

Tc DEVIATION
CONTROL
TC CALCULATION TCTMS

Figure 1: Flow chart of conductor temperature calculations of electrical overhead lines.

The simplest approach to determine the thermal state of the conductor is

by assuming that all of the influence variables (wind speed and direction, solar
radiation, ambient temperature and current) are constant in an interval of time
and when the steady state thermal equilibrium is reached [7]. This leads to a
steady state balance in which, the conductor temperature or thermal rating can
be obtained, as shown in Eq. (6).

Pc + Pr = Pj + Pm + Ps (6)

The steady state assumption does not take into account the thermal inertia
of the conductor materials, so it is necessary to be sure that all of the variables
of interest are constant, at least during the thermal time constant of the con-
ductor. Some of these variables can fit into this condition (ambient temperature
and solar radiation), but the most critical ones, which are the wind speed and

5
 its direction, have large variabilities and are difficult to assume steady state
conditions for convection cooling [8]. This is why a time dependent analysis
gives a more accurate idea of the thermal behaviour of the conductor (Fig. 1).
If a non-equilibrium thermal balance is assumed, Eq. (6) is transformed into:

dT
mc = Pj + Pm + Ps − Pc − Pr (7)
dt

where m is the mass per unit length, c is the specific heat capacity, dT is the
55 temperature increment and dt is the time increment. The thermal inertia of
the conductor materials is included. With this approach, both the conductor
temperature and thermal rating can be evaluated assuming a time step that is
small enough to track the variable of interest.
Finally, an analytical approach obtained by V.T. Morgan from the general
heat equation for a homogeneous and isotropic solid can be utilized [4, 9]. This
leads to a time response function for each variable which depends on the thermal
time constant of the conductor (Fig. 1). Combining the different functions
for different variables by assuming the superposition principle, a global time
response function can be obtained that characterizes the thermal behaviour of
the conductor in dynamic states:

θ = θm − (θm − θ1 ) e−t/τj e−t/τs e−t/τr e−t/τc (8)

where θ is the rise of the average temperature of the conductor above ambient
60 temperature at time t, θ1 is the initial rise of the average temperature of the
conductor above ambient temperature at t=0, θm is the asymptotic rise of the
average temperature of the conductor above ambient temperature and τj , τs , τr
and τc are the heating thermal time constants for changes in the current, solar
radiation, ambient temperature and wind speed, respectively. The calculation
65 of τ depends on the type of the variable change.
Now, focusing on the characteristics of the available data (weather variables
and line parameters), some distinctions should also be made. For the steady
state balance, data should be stable enough to assume this condition, i.e., the
variables should not change significantly at least for a period of time equal

6
70 to the thermal time constant of the conductor. For time dependent analysis,
data should be taken with an adequate resolution to track the transient state
and the shape of the transition should also be decided [10]. When using the
tracking method, a step change for all variables is assumed every time the data
are refreshed and Eq. (7) is solved for a specific time step. Otherwise, when
75 using the dynamic approach with analytical Eq. (8), different transition shapes
can be assumed to modify the thermal constant expression. In this paper, a
step change and a linear transition are studied and compared (Fig. 1).

3. Conductor Temperature Calculation

To calculate the thermal heat balance, real-time data of weather variables

80 (wind speed uw [m/s], wind direction φw [ ], solar radiation Qs [W/m2 ] and
ambient temperature Ta [ C]) were measured with a weather station placed
in the electricity tower. Additionally, the current and conductor temperature
were measured by a Temperature Measurement Sensor (TMS) attached to the
conductor (IcT MS [A] and TcT MS [ C]). All of these data were measured every 8
85 minutes (ts =8’, sample time) for more than one year (from December 1, 2014,
to January 30, 2016) in a 132-kV overhead line with a LA 280 Hawk type
conductor [11] located in northern Spain. The main parameters of the studied
line are summarized in Table 1.

7
 Table 1: List of parameters.

Parameters Description
D = 0.0218 m Outside diameter of conductor
D1 = 0.008 m Core diameter
y = 622 m Altitude
φ = 43 Latitude
δl = 31 Line angle
F = 0.1 Albedo
Ns = 1 Clearness Ratio
αs = 0.5 Absorptivity
ǫs = 0.5 Emissivity
ms = 0.319 kg/m Steel mass per unit length
ma = 0.722 kg/m Aluminium mass per unit length
cs,20 = 460 J/kgK Specific heat capacity of steel at 20 C
ca,20 = 880 J/kgK Specific heat capacity of aluminium at 20 C
−4
βs = 1 · 10 1/K Temp. coefficient of steel specific heat capacity
−4
βa = 3.8 · 10 1/K Temp. coefficient of aluminium specific heat capacity
λa = 240 W/mK Aluminium thermal conductivity
Ksk = 1.025 Skin factor
R20 = 0.1194 Ω/km Conductor resistivity per unit length at 20 C
α20 = 4.1 · 10−3 1/K Linear resistivity coefficient at 20 C

The conductor temperature Tc is calculated using three different analyses in

90 MATLAB : Steady State Analysis (SSA), Eq. (6), Time Dependent Analysis
with a Tracking method (TDA-T), Eq. (7) and Time Dependent Analysis using
a Dynamic approach (TDA-D), Eq. (8).

Steady State Analysis (SSA). This analysis uses the line parameters (Ta-
ble 1) and a sample time ts =8’. Initially, the estimated conductor temperate Tcj
95 is assumed to be equal to the ambient temperature Ta . Then, the steady state
equation (Eq. (6)) is iterated, modifying Tcj until the steady state temperature

8
 is reached. Finally, Tcj is compared with the conductor temperature measured
by the TMS (TcT MS ). This process is repeated for every sample (ts =8’), taking
a new initial Tcj+1 =Tcj (Fig. 2).

PARAMETERS

j=0 VARIABLES
Tcj=Ta REAL TIME
FOR j=[1,n]
MONITORING
(n:number
(tS=8')
Tcj+1=Tcj of samples)
j≠0 j=[1,n]
STEADY STATE
ANALYSIS (SSA)
Tci+1 STEADY STATE Tci+1
LOOP

Pji Pmi Psi Pci Pri

H.B.>0 HEAT BALANCE H.B.<0

Tci+1=Tci+ΔT Tci+1=Tci- ΔT
Pji+Pmi+Psi-Pci-Pri
H.B.≈0
Tcj+1
Tcj
Tc DEVIATION
TMS CONTROL TC 1,TC 2,...,TC n
TC
(n: Number of samples)

Figure 2: SSA flow chart.

100 Time Dependent Analysis - Tracking (TDA-T). As in the previous case,

variables are recorded every 8 minutes (ts =8’). Each interval is at the same
time divided into one second steps (∆t=1 s). Initially, Tcj=0 is calculated by
means of the SSA. Then, the heat balance equation Eq. (7) is tracked until
the 8-minute interval is finished. Finally, Tcj is compared with the conductor
105 temperature measured by the TMS (TcT MS ) and the next set of data is loaded,
taking a new initial Tcj+1 =Tcj (Fig. 3).

9
 TRACKING
LOOP (Δt=1 s)

i≠ 480
Δ ΔTi=(Psi+Pji+Pmi-Pci-Pri)·Δt/mici

Figure 3: TDA-T flow chart.

Time Dependent Analysis - Dynamic (TDA-D). Initially, Tcj=0 is cal-

culated by means of the SSA. Then, variables are recorded every 8 minutes
(ts =8’). TDA-D uses the analytical equation of Eq. (8) and makes three calls
110 to calculate the different terms of the equation (Fig. 4).

SSA(j) to calculate the asymptotic rise of the average temperature of the

conductor above ambient temperature (θmj ).

TDA-T(j) to calculate the gain and loss heat associated with sample j
after one second (i=1).

115 TDA-T(j-1) to calculate the gain and loss heat associated with sample j-1
after one second (i=1).

10
 PARAMETERS

CALL TO S.S.A. VARIABLES

FOR j=0

VARIABLES
REAL TIME
FOR j=[1,n]
Tcj+1=Tcj MONITORING
(n:number
(tS=8')
of samples)
j=(1,n)
TIME DEPENDE NT
VARIABLES FOR j CALL TO SSA ANALYSIS-DYNAMIC
(TDA-D)
VARIABLES FOR j CALL TO TDA-T (i=1)

VARIABLES FOR j-1 CALL TO TDA-T (i=1)

Psj Pjj Pmj Pcj Prj mjcj τj θj

ANALYTICAL EQUATION
Tcj=θmj -(θmj-θ1j)·e-(t/τ sj)·e-(t/τ jj)·e-(t/τcj)·e-(t/τrj)
Tcj+1
Tcj
Tc DEVIATION
CONTROL Tc1,Tc2,...,Tcn
TCTMS
(n: Number of samples)

Figure 4: TDA-D flow chart.

4. Results

As previously stated, data were recorded for more than one year (from De-
cember 1, 2014, to January 30, 2016) in a 132-kV overhead line with a LA 280
120 Hawk type conductor [11] located in northern Spain. During that time, due to
a change in the type of anemometer, two different intervals can be distinguished
and compared. In the first period (from December 1, 2014, to September 24,
2015), a cup anemometer was used. In the second period (from September 24,
2015, to January 30, 2016), the cup anemometer was replaced by a 2D ultrasonic
125 anemometer. This made it possible to study the influence of the accuracy and
resolution of the wind sensor. Later in this section the results of the different
algorithms used are also presented.

11
 To obtain an idea of the influence of the different variables and the accuracy
of the algorithms and to compare this influence with the temperature measured
130 by the TMS, the calculated cases with a deviation of less than +/- 2.5 C from
the conductor temperature measured by the TMS are analysed. The results
for the CIGRE standard algorithms are presented, and in the last part of this
section, they are compared with the results obtained for the IEEE standard
algorithms.

135 4.1. Influence of the wind sensor resolution

It is commonly known that low speed winds are difficult to measure due to
their great variability in direction and the limits of the sensor resolution [12].
This can strongly affect the temperature output of the algorithms, which also
loses accuracy at low wind speeds due to the difficulty of having good convection
140 correlations. When using cup anemometers, the region from 0 to 1 m/s cannot
be well measured because the cups need a minimum speed to start moving. This
leads to many values of 0 m/s being recorded (more than 600 of the 26,657 values
recorded) and can be reasonably thought of as very small winds blowing at a
particular speed. In fact, when changed to an ultrasonic anemometer, values
145 lower than 0.1 m/s were not recorded in the measured interval.
In practice, this difference means that algorithms using more accurate data
for low wind speeds are able to better fit the real temperature measured by the
conductor sensor. Fig. 5 shows the temperature deviation from the conductor
temperature using cup and ultrasonic anemometers for the tracking algorithm
150 (TDA-T).
In Fig. 5, large deviations for the uw =0 m/s measurements from the cup
anemometer are clearly shown, with values overestimating the real conductor
temperature by over 20 C. It is also interesting to note that the cup anemometer
measurements tend to overestimate the temperature of the conductor, so in the
155 case of using cup anemometers, these inaccuracies remain on the conservative
side.

12
 (°C)

20
Cup Anemometer
15
TMS

10
-T c

5
Estimated

0
-5
Tc

0 2 4 6 8 10 12 14 16 18
Wind speed u (m/s)
w
(°C)

20
Ultrasonic Anemometer
15
TMS

10
-T c

5
Estimated

0
-5
Tc

0 2 4 6 8 10 12 14 16 18
Wind speed u (m/s)
w

Figure 5: Deviations from the measured conductor temperature vs. the wind speed.

The differences of the percentage of accurate values (percentage of values

with a deviation less than +/- 2.5 C) using the cup or ultrasonic anemometer
are summarized in Table 2.

Table 2: Percentage of values with a deviation less than +/- 2.5 C using the cup and ultrasonic
anemometers.

Wind speed Cup Ultrasonic

sensor anemometer anemometer
% 92.96 96.78

160 Due to these large differences between the cup and ultrasonic anemometer
values, all of the subsequent results are only presented for the time interval
recorded with the ultrasonic anemometer (from September 24, 2015 to January
30, 2016).

13
 4.2. Influence of the solar radiation estimation
165 CIGRE and IEEE standards give information on how to estimate the solar
radiation reaching the conductor as a function of the position and date. This
value gives an upper bound of the real radiation present in one specific site and
date. If the location where the line is placed usually has cloudy weather (as it is
in the case of this study) or has significant shadows, the estimated values may
170 give a high overestimation of the conductor temperature.
If solar radiation is measured with a pyranometer, this information can be
included in the algorithm to obtain a better solution. CIGRE (Eq. (3)) and
IEEE (Eq. (9)) present similar equations to calculate the heat gained by the
conductor due to solar radiation.

qs = αs A′ Qse sin(θ) (9)

where A′ is the outside diameter of the conductor and θ is the angle of the solar
beam with respect to the axis of the conductor.
The value of the pyranometer may be interpreted as the total radiation (Qse
for IEEE or It for CIGRE) or the direct beam radiation (Ib for CIGRE). In
175 the second case, the albedo value F should also be indicated and the diffuse
radiation Id should be estimated. For this study F = 0.1, was chosen because
the line mainly crosses a forest. The three different options, that is: 1) es-
timated radiation by CIGRE or IEEE, 2) measured as It radiation (CIGRE)
or Qse radiation (IEEE) and 3) measured as Ib radiation (CIGRE), were used
180 to calculate the conductor temperature and were then compared with the real
conductor temperature measured by the TMS. Table 3 summarizes the results
indicating the percentage of the values with a deviation less than +/- 2.5 C
from the TMS measurement for the ultrasonic anemometer values and TDA-T
algorithm.
185 The loss of accuracy using the CIGRE estimated values instead of the mea-
sured ones is shown as a 10 % decrease of the deviation temperature values
between +/- 2.5 %. The use of the pyranometer value as the total radiation It
instead of direct radiation Ib has little influence on the percentage of accuracy.

14
 If the values of the median, maximum overestimation and underestimation and
190 the standard deviation are calculated (Table 3), additional conclusions can be
reached.
The lower underestimation using standard estimated radiation is clear due
to its conservative approach, which is also shown in its higher overestimation
value. Using the pyranometer value as Ib gives lower underestimations and
195 similar overestimations than using the It value for all cases with the cup and
ultrasonic anemometer data. This may depend on the value of F , and more
studies on the algorithm sensitivity to F may be of interest.
Comparing CIGRE and IEEE standards it can be concluded that the latter
presents slightly worse estimations when using the pyranometer value as total
200 radiation (It vs. Qse ).

Table 3: Percentage of values with a deviation less than +/- 2.5 C, median, maximum
overestimation and underestimation and the standard deviation using estimated radiation,
measured as It or Qse radiation and as Ib radiation.

Radiation Standard Pyranometer value used as

Source Estimation It Qse Ib
CIGRE IEEE CIGRE IEEE CIGRE
% 86,79 86,73 97,04 94,99 96,78
Median [ C] 0,77 0,76 0,42 0,5 0,53
Max. overestimation [ C] 9,01 9,09 7,38 8,54 7,38
Max. underestimation [ C] -2,53 -2,9 -5,48 -5,46 -4,29
Standard deviation [ C] 1,36 1,33 0,92 1,02 0,93

Fig. 6 shows the median of the deviation of the CIGRE estimated tempera-
ture from that measured by the TMS (red large line), 50 % of the values (blue
box) and the whiskers (black lines) and outliers (small red lines) calculated by
1.5 times the interquartile range.

15
 8

6
T cEstimated-T cTMS (°C)

-2

-4

-6
Standard Estimation Total Radiation Direct Radiation
It Ib

Figure 6: Box plot of the deviation from the measured conductor temperature using CIGRE
estimated radiation, measured as It radiation and measured as Ib radiation.

205 As a representative example, a specific day (September 27, 2015) is presented

in more detail for the tracking algorithm (TDA-T) to examine these differences.
In Fig. 7, the green, red and yellow lines correspond to the temperature obtained
by the CIGRE estimated radiation, measured radiation as total radiation (It )
and measured radiation as direct radiation (Ib ). The black line represents the
210 temperature of the conductor measured by the TMS, and the blue line represents
the ambient temperature. Without radiation (during the night), it is obvious
that the three approaches obtain the same estimated temperature. However,
when the influence of the solar radiation appears the three approaches obtain
different estimated temperatures. The solid green line of Fig. 8 is the measured
215 radiation during that day, and the dotted green and purple lines are the values
of the estimated radiation by the IEEE and CIGRE standards, respectively.
Differences in the estimations of IEEE and CIGRE are observed because IEEE
only considers direct radiation in its solar intensity calculations.
For the estimated radiation case, greater differences are shown from appro-
220 ximately 8:00 to 14:00 due to the intervals of cloudy weather in the morning. A

16
 better fit is obtained in the evening as the estimated radiation becomes closer
to the measured radiation. In regards to the use of total or direct radiation,
the temperature calculated using the value of the pyranometer as total radia-
tion is slightly lower than using the value as the direct radiation because diffuse
225 radiation and albedo effects are not taken into account.

30
TTMS
28 c
TStandard
c
Estimation

26 TTotal Radiation
c
24 TDirect
c
Radiation
Temperature [ºC]

Ta
22

20

18

16

14

12

10
00 2:00 4:00 6:00 8:00 0:00 2:00 4:00 6:00 8:00 0:00 2:00 0:00
00: 0 0 0 0 1 1 1 1 1 2 2 0

Figure 7: Ambient, conductor and CIGRE estimated temperature for September 27, 2015.

1100 27.5
Wind Direction [º] Ambient Temperature [ºC]
1000 Conductor current [A] Wind speed [m/s] 25
Measured Radiation [W/m2 ]
900 CIGRE Estimated Radiation [W/m 2 ] 22.5
IEEE Estimated Radiation [W/m2 ]
800 20
700 17.5
φw - IcTMS - Q s

T a - uw

600 15
500 12.5
400 10
300 7.5
200 5
100 2.5
0 0
00 2:00 4:00 6:00 8:00 0:00 2:00 4:00 6:00 8:00 0:00 2:00 0:00
00: 0 0 0 0 1 1 1 1 1 2 2 0

Figure 8: Weather variables for September 27, 2015.

17
 4.3. Influence of the type of algorithm

As previously explained, the conductor temperature was calculated using

three different analysis: steady state analyses (SSA), time dependent analy-
sis with a tracking method (TDA-T) and time dependent analysis using the
230 dynamic approach with an analytical equation (TDA-D).
The most accurate option, i.e., the temperature calculated with data from
the ultrasonic anemometer and using the measured radiation as the direct ra-
diation (Ib ), is studied in this section. To obtain an idea of the accuracy of
the algorithms in comparison with the temperature measured by the TMS, the
235 percentage of cases with a deviation lower than +/- 2.5 C is calculated and
presented in Table 4.

Table 4: Percentage of values with a deviation less than +/- 2.5 C using SSA, TDA-T and
TDA-D algorithms.

SSA TDA-T TDA-D

% 95.72 96.78 96.57

The main difference appears between the steady state and two types of time
dependent analyses, with an increment of 1 % from steady to transient analyses.
Focusing on the time dependent analysis, the tracking algorithm (TDA-T)
240 seems to be the best fit for the TMS temperature values. When using a more
complex algorithm to include the analytical equation (Eq. (8)), (TDA-D) does
not increase the accuracy. In fact, the temperatures obtained by TDA-T better
match the temperatures of the conductor in the transient periods.
Combining step or ramp responses in the TDA-D does not significantly vary
245 the results. Table 5 shows the type of time responses for each variable and the
differences in the percentage.

18
Table 5: Percentage of values with a deviation less than +/- 2.5 C for different types of time
responses.

Current Solar Ambient Wind %

radiation temperature speed
C. 1 Step Step Step Step 96.57
C. 2 Step Ramp Ramp Ramp 94.61
C. 3 Step Ramp Ramp Step 95.42
C. 4 Step Ramp Step Ramp 95.41

Table 6 summarizes the main statistical values for the algorithms used: me-
dian, maximum overestimation and underestimation and the standard deviation.

Table 6: Median, maximum overestimation and underestimation and the standard deviation
for all of the algorithms used.

Type of SSA TDA-T TDA-D TDA-D TDA-D TDA-D

algorithm (C. 1) (C. 2) (C. 3) (C. 4)
Median [ C] 0.52 0.53 0.54 0.57 0.59 0.59
Max. Overestimation [ C] 9.50 7.38 6.69 5.89 5.89 5.89
Max. Underestimation [ C] -4.32 -4.29 -7.22 -8.80 -6.84 -6.84
Standard Deviation [ C] 1.00 0.93 0.97 1.19 1.05 1.05

Fig. 9 represents the median, 50 % of the values and the outliers.

19
 10

4
T Estimated-T TMS

-2

-4

-6

-8

SS
A A-
T .1) .2) .3) .4)
TD D (C D (C D (C D (C
A- A- A- A-
TD TD TD TD

Figure 9: Box plot of the deviation from the measured conductor temperature for all of the
algorithms used.

250 Due to the good results for the tracking algorithm, a sensitivity study of the
time step was performed from 0.1 to 120 s but no differences in the algorithm
accuracy were found.
Finally, it is also important to note that the IEEE and CIGRE standards
give very similar results in this study, which has also been noticed in previous
255 studies of steady state analysis [13].
Table 7 and Fig. 10 compare both standards. The main differences between
both standards are that the percentage of values with a deviation less than +/-
2.5 C is around 1.5 % higher for CIGRE and that IEEE standard obtains higher
overestimation and underestimation.

20
 Table 7: Percentage of values with a deviation less than +/- 2.5 C for IEEE and CIGRE
standards.

Standard IEEE CIGRE

algorithm SSA TDA-T SSA TDA-T
% 94.25 94.94 95.72 96.78
Median [ C] 0.50 0.50 0.52 0.53
Max. Overestimation [ C] 11.39 8.54 9.50 7.38
Max. Underestimation [ C] -5.44 -5.46 -4.32 -4.29
Standard Deviation [ C] 1.08 1.02 1.00 0.93

12

10

6
T Estimated-T TMS

-2

-4

-6
IEEE SSA IEEE TDA-T CIGRE SSA CIGRE TDA-T

Figure 10: Box plot of the deviation from the measured conductor temperature for IEEE and
CIGRE standards.

260 4.4. Sensitivity study of sensor measurements

Due to the variations of the sensor measurements, a Monte Carlo test was
made to evaluate their sensitivity. The sensor accuracies provided by the ma-
nufacturers are: ambient temperature thermometer Ta +/- 0.2 C, ultrasonic
anemometer uw +/- 0.3 m/s and φw +/- 2 , pyranometer Qs +/- 5 % and
265 power quality analyzer IcT MS +/- 0.1 %.
This test was made for the steady state analysis (SSA), using the ultrasonic

21
 anemometer and including the pyranometer values as the direct radiation Ib .
The first step was to decide the number of Monte Carlo combinations for each
set of data. In order to simulate the worst possible scenario, all the sensor mea-
270 surements were considered to follow a rectangular distribution in their ranges of
accuracy. Monte Carlo tests with n=10, 100, 1000, 10000 and 100000 combina-
tions of the sensor measurements (uw , φw , Qs , Ta and IcT MS ) were obtained for
one set of data, and the corresponding n conductor temperatures (T1 ...Tn ) were
calculated. The mean temperature value for each test T 1:n and the estimated
275 temperature Testimated using the original values directly obtained by the sensors
were calculated.
Table 8 shows the average and the maximum positive and negative values of
the differences Testimated -T 1:n . It can be seen that the average of the difference
tends to zero as the number of Monte Carlo combinations increases and that
280 the range [minimum-maximum] of this difference does not increase significantly
from 1000 to 100000 combinations. This is the reason why a 1000 Monte Carlo
combinations test was chosen to evaluate the sensitivity of the sensors.

Table 8: Sensitivity study of sensors using Monte Carlo method.

Number of Testimated -T 1:n [ C]

combinations n average minimum maximum
10 -0.0706 -0.3552 0.2137
100 -0.0330 -0.4284 0.3418
1000 -0.0077 -0.4355 0.4521
10000 -0.0070 -0.4680 0.4338
100000 -0.0070 -0.5077 0.4670

Then, the 1000 combinations Monte Carlo test was run for all the set of
data. The temperature standard deviations (σT1:1000 ) for each set of data were
285 calculated and added as a correction to the estimated conductor temperatures
Testimated . In this case, the percentage of values with a difference less than +/-
2.5 C from the measured temperature TT MS was 94.17%. If these standard

22
 deviations were subtracted instead, the percentage of these values was 96.85%.
Both values are similar to the result 95.72% presented in Table 4 where the
290 sensor variations were not taken into account.

5. Conclusions

The temperature and current of an overhead conductor and the weather

conditions surrounding the cable were measured every 8 minutes for more than
one year. With these data, the accuracies of different algorithms presented in
295 the standards [4, 5] were studied by implementing them in MATLAB .
One result was the large difference in the accuracy of the estimated tem-
perature using cup or ultrasonic anemometers. Although the cup anemometer
gave an overestimation of the conductor temperature, it introduced a large de-
viation between the estimated and measured temperatures in the presence of
300 low winds. The percentage of values with a deviation less than +/- 2.5 C was
92.96 % for the cup anemometer and 96.78 % for the ultrasonic anemometer.
Thus, ultrasonic anemometers are recommended to be used when possible.
Similar conclusions were found regarding estimated and measured solar ra-
diation. The percentages of values with a deviation less than +/- 2.5 C were
305 86.79 % (CIGRE) and 86.73 % (IEEE) for estimated radiation and 97.04 %
(CIGRE) and 94.99 % (IEEE) for measured radiation. Using estimated radia-
tion gave a very high temperature overestimation for cloudy sites; thus, using a
pyranometer is highly recommended. The differences in including estimations of
albedo F and diffuse radiation Id were negligible in practice. Comparing CIGRE
310 and IEEE standards, the latter presents slightly worse estimations when using
the pyranometer value as total radiation (It vs. Qse ).
Steady state analysis gave good results when the recording interval was small
enough, but if transitions were to be studied, the tracking time dependent ana-
lysis was the best option. More complex algorithms, including time function
315 responses, can increase the amount of mathematical code and the computing
time, while not improving the results significantly. The percentages of values

23
 with a deviation less than +/- 2.5 C were 95.61 % (CIGRE) and 94.25 %
(IEEE) for the steady state analysis, 96.78 % (CIGRE) and 94.94 % (IEEE) for
the tracking algorithm and 96.57 % (CIGRE) for the time dependent algorithm
320 with an analytical equation. Thus, the main differences between both standards
are that the percentage of values with a deviation less than +/- 2.5 C is around
1.5 % higher for CIGRE and that IEEE standard obtains higher overestimation
and underestimation.
The Monte Carlo method was used to analyze the sensor sensitivities for the
325 steady state analysis with 1000 combinations of the sensor measurements. The
results were similar to those obtained when the sensor variations were not taken
into account.

Acknowledgments

This work was supported by the Spanish Government under the R+D ini-
330 tiative INNPACTO with reference IPT-2011-1447-920000 and Spanish R+D
initiative with reference ENE2013-42720-R. The authors would also like to ac-
knowledge Viesgo for its support.

References

[1] J. Zbojnk, M. Dvok, Power grid simulation model for long term operation
335 planning, Applied Thermal Engineering 70 (2014) 1294 –305.

[2] S. Nykamp, A. Molderink, J. L. Hurink, G. J. Smit, Statistics for pv,

wind and biomass generators and their impact on distribution grid plan-
ning, Energy 45 (2012) 924 –32. The 24th International Conference on
Efficiency, Cost, Optimization, Simulation and Environmental Impact of
340 Energy, (ECOS) 2011.

[3] H. Perez, S. Richards, B. Leyde, When intermittent power production

serves transient loads, Applied Thermal Engineering 50 (2013) 1549 –56.

24
 [4] CIGRE technical brochure 601, guide for thermal rating calculation of over-
head lines, December 2014.

345 [5] IEEE Std. 738-2012, standard for calculating the current-
temperature relationship of bare overhead conductors, 2013.
doi:10.1109/IEEESTD.2013.6692858.

[6] M. W. Davis, A new thermal rating approach: The real time thermal rating
system for strategic overhead conductor transmission lines – part I: General
350 description and justification of the real time thermal rating system, IEEE
Transactions on Power Apparatus and Systems 96 (1977) 803–9.

[7] V. Morgan, The thermal rating of overhead-line conductors part i. the

steady-state thermal model, Electric Power Systems Research 5 (1982) 119
–39.

355 [8] CIGRE technical brochure 498, guide for application of direct real-time
monitoring systems, June 2012.

[9] V. Morgan, Rating of bare overhead conductors for intermittent and cyclic
currents, Electrical Engineers, Proceedings of the Institution of 116 (1969)
1361–76.

360 [10] J. Rodriguez, C. Franck, Evaluation of the accuracy of a thermal rat-

ing model to estimate the temperature of operational transmission lines,
CIGRE Science & Engineering. Issue 004. (2016).

[11] EN 50182:2001, conductors for overhead lines. Round wire concentric lay
stranded conductors, 2001.

365 [12] Cigre technical brochure 299, guide for selection of weather parameters for
bare overhead conductor ratings, August 2006.

[13] A. Arroyo, P. Castro, R. Martinez, M. Manana, A. Madrazo, R. Lecuna,

A. Gonzalez, Comparison between ieee and cigre thermal behaviour stan-
dards and measured temperature on a 132-kV overhead power line, Energies
370 8 (2015) 12391.

25