Sei sulla pagina 1di 6

The 20th International Symposium on High Voltage Engineering, Buenos Aires, Argentina, August 27 – September 01, 2017

MONTE CARLO LIGHTNING PERFORMANCE ANALYSIS OF


OVERHEAD TRANSMISSION LINES PROTECTED BY ZNO
ARRESTERS USING MATLAB/ATP

R. S. Oliveira1,2 and I. J. S. Lopes3


1Graduate Program in Electrical Engineering - Federal University of Minas Gerais
UFMG - Av. Antônio Carlos, 6627, 31.270-901
2Minas Gerais State Power Utility - CEMIG
3Electrical Engineering Department - Federal University of Minas Gerais

Belo Horizonte, MG, Brazil


*Email: rafasdeoliveira@gmail.com

Abstract: Non-scheduled outages, especially in overhead transmission lines (OHLs) rated


below 230kV, are mainly caused by lightning. In many cases, the intense lightning activity
combined with high soil resistivity, leads the OHL to perform under the acceptable limits.
Outages can cause revenue reductions for power distribution and transmission companies,
as well as personal and material losses for costumers. In Brazil, ANEEL (National Electricity
Regulatory Agency) and ONS (National Power System Operator), are increasing the
requirements, forcing the utilities to pay more attention to the time and frequency outage
limits. In order to mitigate this problem, Zinc-Oxide (ZnO) surge arresters have been used
as an interesting solution for OHLs with poor lightning performance. Because of their non-
linear VxI curve, the arresters installation in all phases of all towers reduces the number of
outages to zero. However, installing a large amount of arresters requires high investment.
Alternatives, such as the installation of arresters in the lower phases and/or in some towers
with high footing resistance values, may be a solution to reduce the number of outages to
an acceptable level. Based on this scenario, this paper presents a methodology to estimate
the lightning performance of OHLs using Monte Carlo Method and considering the use of
surge arresters along the line. The proposed method was implemented, a program was
built in MATLAB environment and the ATP version of the Electromagnetic Transient
Program was used for the circuit simulations. The program estimates the number of
outages due to back flashovers and shielding-failures. Results are compared with well-
known programs “IEEE Flash” and “Sigma SLP”. Additionally, a case study based on real
138 kV OHL data from CEMIG (Minas Gerais State Power Utility) is presented. A procedure
to reduce the number of line outages, based on individual tower performance, is proposed.
The method can be used as a design tool to search for a cost effective solution on the
number of arresters to be installed along an OHL to reach an acceptable lightning
performance.

1 INTRODUCTION residual voltage of the ZnO and the supportability of


the OHL insulation [2].
Overhead Transmission Lines (OHL) having
nominal voltages under 230kV are susceptible to
Line arresters installation in all phases of all towers
lightning caused outages. Some factors, such as
drops the number of outages of an OHL to zero.
high soil resistivity and intense lightning activity,
However, in a limited financial resources scenario,
may lead the OHL to perform under the acceptable
this may not be feasible. Based on that, and due to
limits. Considering that the Brazil National Electricity
the random nature of lightning [3], this work
Regulatory Agency (ANEEL) and the National
presents methodology for lightning performance
Power System Operator (ONS) have been
analysis of OHL protected by ZnO arresters based
increasing the regulatory requirements (time and
on Monte Carlo Method.
frequency outage limits), power utility companies
must be aware of the possible revenue reduction
related with such outages. The proposed methodology is implemented in
MATLAB environment, together with the Alternative
Transient Program (ATP) to perform the
Different approaches can be used to reduce the
electromagnetic transient calculations. It allows
number of lightning outages, such as improving the
estimating the number of outages/100km.year of an
shielding angle and reducing the tower footing
OHL considering any number/configuration of
resistance [1]. However, in some cases, this may
arresters installed tower by tower.
not be possible. For these cases, the installation of
metal-oxide surge arresters (ZnO) is an interesting
The OHL spans are simulated using “JMarti”
solution, due to the great difference between the
frequency-dependent model; the insulation
breakdown is evaluated with the Voltage x Time in each side of the tower. At the line terminations,
curve of the insulator strings; the ZnO arresters and matching impedances are connected to avoid
the tower footing are modeled as nonlinear reflections.
resistances (TACS-Models ATP representation)
and lightning parameters are assumed to present a The towers are simulated by means of lossless,
log-normal distribution [3][7]. The validation process single-phase transmission lines [8][9][10]. The
is performed comparing the results with two well- length of each line is equal to the height difference
known programs: IEEE Flash [4], and Sigma SLP between cross arms and/or earth. The tower
[5]. Additionally, the lightning performance impedance is calculated as recommended by [4].
calculated by the routine and real data observed on
a 138 kV OHL located in Minas Gerais State, are
compared. Finally, a case study is carried out and a
procedure is presented showing how different
arrester configurations can be used to achieve a
desired performance.

2 ALGORITHM FOR OHL LIGHTNING


PERFORMANCE ESTIMATION
Monte Carlo Procedure to estimate the number of
outages/100km.year is summarized in Figure 1. The
algorithm is described below:

 Data related to the overhead transmission line,


such as tower geometry and electrical
parameters of the conductor, span length, tower
footing resistance, insulation strings length, ZnO
configuration and lightning statistical parameters
are informed;
 Using data related to the geometry of the
conductors in each span, the algorithm runs the
ATP to perform the construction of the simulation
cards of each span;
 According to lightning statistical parameters,
discharge currents are generated based on a
log-normal distribution. Each discharge
incidence point is obtained using he
Electrogeometric Model [1]. Lightning events are
generated until an specified minimum number of Figure 1: Monte Carlo Procedure Flow-chart.
discharges (hitting the shield wires or phase
conductors) is reached; The tower footing impedance is simulated by a non-
 From all the input parameters, simulation cards linear resistance as represented by Equation 1
are generated; [9][10][12]. This is included in the simulation by a
 The algorithm runs ATP once again to calculate combined TACS-Models representation.
the resulting voltages along the OHL insulator
strings; (1)
R i = R 0 /√(1 + IR /Ig )
 The algorithm reads the overvoltages of each
simulation and evaluates if a flashover occurred
or not. At the end, the lightning performance is where, R0 represents the tower footing resistance,
calculated. IR the current injected and Ig the current from which
it is observed the ionization effect.
2.1 System Modeling The supportability of the insulation strings is
simulated by comparing the overvoltages calculated
The OHL spans are simulated by means of ATP in each simulation with the voltage x time curve. The
“JMarti” frequency-dependent model [6][8]. Due to curve is calculated using Equation 2 [4].
the tower positions and configuration of the shield
wires and phase conductors, each line span is V(t) = 0.4W + 0.71/t 0.75 (2)
identified by a specific card. However, not all the where W represents the length of the insulation
spans are considered in each simulation, to save strings in meters, t time in microseconds and V
memory and time. For each lightning incidence voltage in kV.
point, a number N of nearby spans are considered
The metal-oxide surge arresters are modelled by 3.1 Comparison with IEEE Flash
their Voltage x Current (V x I) curve [4]. The
To perform a comparison with IEEE Flash program,
algorithm allows the user to consider the arresters
a typical 138 kV was simulated. The OHL is
installation tower by tower, phase by phase.
composed by twenty 300-m spans. The typical
tower is 26.75m high has a surge impedance of
Lightning discharges are modeled by current
152Ω. The insulation strings are composed by nine
sources whose polarity can be positive or negative.
254 x 146mm insulator units. Phase conductors and
The wave-shape is considered to be a slop-ramp
shield wire data are provided in Table 2, and the
[6][11].
ground flash density is 5.0 discharges per km2.
Monte Carlo Method is used to generate the
Table 2: Conductor and shield cables geometry
discharge parameters: point of incidence
coordinates; front time (tf); tail time (th) and peak Conductor X(m) Y(m) r(mm)
value (IP). The point of incidence is chosen 1 -2.9 23.72 9.15
randomly in an area of 1km of width and length “L”
2 2.9 21.86 9.15
equal to the length of the OHL. The parameters tf,
th, and Ip are assumed to follow a log-normal 3 -2.9 20.00 9.15
distribution, with probability density function p(x) [7]: 4 0 26.75 4.63

p(x)
2 As IEEE Flash uses the 2 point method, some
1 1 ln(x) − ln(xm ) (3)
= exp [− ( ) ] changes and/or simplifications were made for a
√2πxσln(x) 2 σln(x) closer similarity of the input parameters used by
where: 𝜎ln(𝑥) is the standard deviation of ln(x), and both programs: All discharges were considered
negative; the front time was fixed at 2µs and tail time
xm is the median value of x. Table 1 shows the
infinity; peak value was sorted according to a
parameters used to sort the discharges using
cumulative probability distribution given by
Equation 3 [7].
Anderson [4] (Equation 6).
Table 1: Statistical Parameters of Negative and
Positive Strokes [7]. 𝑝(𝐼) = 1/(1 + (𝐼𝑐 /31)^2.6) (6)

Negative Positive
Parameter where, Ic is the crest value of stroke current.
Median σ Median Σ
31.1 0.48 35.0 1.21 Two sets of simulations were performed to calculate
IP, kA
3.83 0.55 22.0 1.23 the lightning performance: (a) varying the tower
tf, kA footing resistance and (b) varying the number of
77.5 0.58 230.0 1.33 insulators. The results are presented in Figure 2,
th, kA
and show that, for both cases, there is a strong
Ninety percent of the discharges are considered of similarity between the algorithm proposed and the
negative polarity. Besides that, 60% of the sorted IEEE Flash.
strokes hit the tower, and 40% hit the mid-span.
3.2 Comparison with Sigma SLP
A single value for the average ground flash density To compare the proposed method with Sigma SLP,
along the line should be used. The Electrogeometric data from a 90kV shielded OHL presented in [5] was
Model [1] is used to determine the last step of each used. The lightning performance was calculated at
sorted return stroke (Equations 4 and 5). 4 different arrester configurations: (a) without any
ZnO; (b) two ZnO in all towers; (c) two ZnO in
alternate towers and (d) one ZnO in alternate
𝑟𝑐 = 10𝐼 0.65 (4)
towers. The tower footing resistance values vary
[3.6 + 1.7 ln(43 − 𝑦𝑐 )]𝐼 0.65 𝑦𝑐 < 40𝑚 from 5 to 60Ω.
𝑟𝑔 = { (5)
5.5𝐼 0.65 𝑦𝑐 ≥ 40𝑚
The OHL is composed by sixteen 150-m spans. The
typical tower is 23.3m high and has an impedance
of 175Ω. The insulator strings are composed by six
where: rc is the attractive radius of the shield and
255 x 127mm insulator units. The phase conductors
phase conductors, rg the attractive radius of earth
and shield wires data is given in Table 3 and the
and yc the height (m).
ground flash density is 2.8 discharges per km2 per
year.
3 VALIDATION
To validate the algorithm, two comparisons are
presented. First, with IEEE Flash [4] and secondly
with Sigma SLP [5].
50.0 30.00
Num Outages/100km.year 45.0

Num outages/100km.year
25.00
40.0
35.0 20.00
30.0
15.00
25.0
20.0 10.00
15.0
10.0 5.00
5.0
0.00
0.0 5.00 15.00 25.00 35.00 45.00 55.00
0.0 10.0 20.0 30.0 40.0 50.0 60.0
Tower footing resistance
Tower footing resistance Algorithm (a) Sigma LP (a)
Algorithm Flash Algorithm (b) Sigma LP (b)
(a)
25.00

Num outages/100km.year
35.0 20.00
Num Outages/100km.year

30.0
15.00
25.0
20.0 10.00

15.0
5.00
10.0
0.00
5.0
5.00 15.00 25.00 35.00 45.00 55.00
0.0 Tower footing resistance
7 8 9 10
Algorithm (c) Sigma LP (c)
Number of Insulators
Algorithm (d) Sigma LP (d)
Algorithm Flash

(b) Figure 3: Comparison with Sigma SLP


Figure 2: Comparison with IEEE Flash.
Utility), was chosen. The OHL is 44.6 km long,
Table 3: Conductor and shield cables geometry composed by 103 towers (typical tower shown in
Figure 4). Phase conductors and shield wires are
Conductor X(m) Y(m) r(mm)
Linnet and 5/16” HS respectively. The insulator
1 2.35 20.40 11.03 strings are composed by nine 254 X 146 mm
2 -2.95 18.70 11.03 insulator units. Lightning average ground flash
3 2.95 17.00 11.03 density is 3.49 discharges per km² per year.
4 0 23.30 6.30
Figure 5 presents the tower footing resistance
values and the span length distributions along the
Similarly to the previous case, some modifications line. A total of 140 ZnO line arresters are installed
and simplifications were performed to establish a along the line [13]: one tower has arresters installed
similarity between the input parameters used by in each phase, 45 towers have 2 arresters, and 47
both programs: All discharges were considered towers have one arrester installed. The sets of two
negative; the front time was sorted according to a and one arresters were installed at the lower
log-normal distribution with median 3.83 and σ 0.55; phases.
tail time fixed in 75µs; peak value sorted according
to log-normal distribution with median 26.0 and σ Detailed information about this OHL and its
0.605. observed lightning performance can be found in
[13].
Figure 3 presents the results of the four different
arrester configurations. As seen, for all cases, the Since the arresters were installed along this line,
results obtained by the proposed algorithm and from 2007 to 2015 its lightning performance has
Sigma SLP are in close agreement. been recorded by CEMIG. The average outage rate
for the period is 2.89 outages per year. There is no
4 CASE STUDY information available on the lightning performance
4.1 Line under study before the installation of the arresters.

In order to make a comparison between the


algorithm results and real line data, the 138 kV OHL
Itutinga – Minduri, located in Minas Gerais and
operated by CEMIG (Minas Gerais Power
Doing so, the effectiveness of the arresters
installation in each tower performance was
evaluated. In this case (blue line in Figure 6), the
estimated OHL lightning performance was 17.64
outages/100km.year.
1.20

Number of outages/100km.year
1.00

0.80

0.60

0.40

0.20

0.00
0 10 20 30 40 50 60 70 80 90 100
Figure 4: OHL Tower Geometry. Tower
Without ZnO With ZnO 1 ZnO 2 ZnO 3 ZnO
660
600
540 Figure 6: Number of outages/100km.year per
480 tower for OHL Itutinga – Minduri.
420
Rt (Ω)

360
300
240
Finally, to evaluate the effectiveness of different
180 arrester configurations, the towers were ranked by
120
60 their performance without arresters. After that,
0 several simulations were performed considering an
1,600
increasing number of arresters installed from the
1,400
1,200
highest failure rate towers to the lowest ones. Figure
Span Length (m)

1,000 7 presents the OHL reduction in the number of


800 outages by the amount of worst performance towers
600 protected by ZnO arresters.
400
200 100.0%
0 90.0%
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 80.0%
Reduction of Outages

Tower Number 70.0%


60.0%
Figure 5: OHL Itutinga - Minduri tower footing
50.0%
resistance and span length distribution. 40.0%
30.0%
4.2 Lightning Performance Analysis 20.0%
10.0%
0.0%
Based on the OHL configuration and performance 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
data, initial simulations were performed to compare Towers protected by Surge Arresters
the results obtained with the proposed method and 1 ZnO 2 ZnO 3 ZnO

the actual outage data. A total of 65406 discharges Figure 7: Reduction in the number of outages for 3
were simulated from which approximately 10000 hit different ZnO arrester configurations, OHL Itutinga
the OHL. The obtained number of - Minduri.
outages/100km.year is presented in Table 4. As
seen, the calculated lightning performance is about As seen, if one arrester is installed per tower, the
40% higher than the observed average maximum reduction in the outage number is around
performance. These results are considered to be in 30% when all towers receive one arrester. On the
good agreement. other hand, when two arresters are installed per
tower, the outages can be reduced in 60%. The
Table 4: OHL Failure Rate number of outages can be reduced to zero only
Number of when three arresters are installed per tower and
Item
outages/100km.year every tower receives the arresters. The same
Observed 6.48 reduction in outages (30%) can be reached by
Simulated 9.11 installing sets of two arresters in 30% of the highest
failure rate towers, or installing sets of tree arresters
A further investigation on the contribution of each in 15% of the towers with worse performance.
individual tower to the number of
outages/100km.year line was performed. Figure 6 Assuming that the estimated performance without
(red line) shows the result. Additionally, an arresters (17.64 outages/100km.year) is a good
estimation of the OHL performance before the estimation for this line, the simulated failure rate
installation of the arresters was performed, reduction is 48% obtained using 140 arresters
considering the same line without any ZnO arrester. actually installed along the 103 towers. Therefore,
it can be inferred from Figure 7 that the same [1] IEEE Std. 1243-1997, “IEEE Guide for
performance improvement can be achieved by improving the lightning performance of
installing sets of tree arresters in 25% of the towers transmission lines,” 1997;
(78 arresters) or sets of two arresters in 60% of the
[2] A. R. Hileman, Insulation Coordination For
towers (124 arresters). In both cases, the total
number of ZnO arresters would be lower than the Power Systems, Marcel Dekker, 1999;
140 units actually installed. The installation of a [3] CIGRE WG 33-01, “Guide to Procedures for
single arrester per tower would not lead to an Estimating the Lightining Performance of
outage reduction of 48%. Transmission Lines”, CIGRE Brochure 63,
1991;
The curves presented in Figure 7 can be seen as a [4] Transmission Line Reference Book, 345kV
fingerprint of this specific OHL. They can be used to
and Above, 2º Ed., 1985;
find a cost effective solution on the configuration
and number of ZnO arresters needed to reach an [5] S. Sadovic, R. Joulie, S. Tartier and E. Brocard,
acceptable lightning performance of this OHL. "Use of line surge arresters for the
improvement of the lightning performance of 63
5 CONCLUSIONS kV and 90 kV shielded and unshielded
This paper presented a Monte Carlo methodology transmission lines," in IEEE Transactions on
to estimate the lightning performance of OHLs Power Delivery, vol. 12, no. 3, pp. 1232-1240,
protected by surge arresters. The methodology was Jul 1997;
implemented in MATLAB environment. The routine [6] H. W. Dommel, Electromagnetic Transients
uses the ATP to perform the electromagnetic Program. Reference Manual. Portland, OR:
transient calculations. Bonneville Power Administration, 1986;
[7] IEEE TF on Parameters of Lightning Strokes,
The methodology was compared with two well-
known lightning performance programs: IEEE Flash “Parameters of lightning
and Sigma SLP. In both cases, the results are in strokes: A review,” IEEE Trans. on Power
good agreement. Differently from IEEE Flash, the Delivery, vol. 20, no. 1, pp. 346-358, January
proposed method compares the resulting stress at 2005;
the insulator string with its supportability not only in [8] J.A. Martinez and F. Castro-Aranda, “Modeling
two points, but all the time and considers the of overhead transmission lines for line arrester
installation of line surge arresters for performance
studies,” IEEE PES General Meeting 2004,
improvement. On the other hand, the main
difference between Sigma SLP and the proposed June 2004, Denver;
method is the focus on individual tower performance [9] J.A. Martinez and F. Castro-Aranda, “Lightning
to search a better performance solution. performance analysis of overhead
Additionally, a real 138 kV OHL operated by CEMIG transmission lines using the EMTP,” IEEE
was studied. A total of 140 ZnO surge arresters Trans. on Power Delivery, vol. 20, no. 3, pp.
were installed along this line and its lightning
2200-2210, July 2005;
performance has been recorded from 2007 to 2015.
The actual performance data was compared with [10] J. A. Martinez and F. Castro-Aranda,
the proposed methodology and good results were "Lightning Flashover Rate of an Overhead
obtained. Transmission Line Protected by Surge
Arresters," Power Engineering Society
Finally, a tower by tower failure rate analysis was
proposed and performed. The approach allows the General Meeting, 2007. IEEE, Tampa, FL,
user to decide on how many ZnO arresters are 2007, pp. 1-6;
needed and in which towers the arresters are to be [11] Electro-Magnetic Transient Program (EMTP),
installed on a certain line to reach a desired lightning Theory Book;
performance. The proposed procedure can be used [12] J. A. Morales, E. A. Orduña, R. J. Cabral and
as a design tool for arrester installation to improve
A. S. Bretas, "Combined TACS-MODELS for
OHL lightning performance.
Footing Tower Resistance considering ground
ACKNOWLEDGMENTS ionization," 2014 North American Power
Symposium (NAPS), Pullman, WA, 2014, pp.
The authors thank the Brazilian agencies: CAPES,
1-5.
FAPEMIG and CNPq for the financial support
provided to this study. [13] S. C. Assis, W. C. Boaventura, J. O. S. Paulino,
R. L. Markiewicz, “Lightning Performance of
REFERENCES Transmission Line with and without Surge
Arresters: Comparison between a Monte Carlo
method and field experience”, Electric Power
Systems Research 149 (2017) 169-17;

Potrebbero piacerti anche