Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 9

CIGRÉ-XXX 2012 CIGRÉ Canada Conference

21, rue d’Artois, F-75008 PARIS Hilton Montréal Bonaventure


http : //www.cigre.org Montréal, Québec, September 24-26, 2012

Environmental Statistical Approach to Evaluate Swing Angles,


Relative Insulation Strengths of Air Gaps and Risks of Failure
Based on Locally Acquired Data - Practical Cases

JIF, DA SILVA S, ANAUATE SJ, DA SILVA


CEPEL TACTA ENERCON UFRJ (UNIVERSITY)
Brazil Brazil Brazil
AN, MPALANTINOS M, FERNANDES SMM, FELIX
CEPEL TACTA ENERCON TRANSFER ENG
Brazil Brazil Brazil
AAJ, MENEZES
BATÁVIA ENG
Brazil

SUMMARY
Over the years, swing angles have been a subject of various calculation methods which usually led to a
wide range of parametric tower design and consequently to diverging operational performances.
Furthermore, while national and international analytical approaches abound, massive experimental
results are seldom reported with the rare exception of the widely known Hornisgrinde Experiment.
Reference [5] is another remarkable one. Not only rare field results are available to substantiate
models or equations on this issue but, in some instances, even the wind data base is somewhat
inaccurate to produce proper swing angle results, such inaccuracies being due to insufficient wind data
in time (few years of acquisition) or in space (sparse collecting networks). Wind speeds are here
specially evoked due to their governing role on the phenomenon in question.
Needless to say, the Gumbel distribution of maxima represents a so to speak point of no return in
what refers to forecasting wind loads on line components, even though alternate methods are
frequently described by the international technical literature or in scientific publications. But whenever
it comes down to evaluate a risk of failure in a swing angle situation, an “overall time wind speed
distribution” is remarkably required to produce wind speeds more appropriate to handle this specific
phenomenon. The relevant document “Tower Top Geometry and Mid-Span Clearances - Working
Group B2.06, June 2008) suggests general values for the WEIBULL distribution (its shape factors,
mostly) whenever 10 minute averages are continuously measured to build up an “all-the-time”
distribution. Such values can actually be assumed as the best bet or choice for design purposes if no
extensive measurements, in both time and space, are available to locally infer a more precise
parameter.
As a further step to nationally model the risk of an outage during an insulator string swing, this paper
statistically approaches:
 Wind speeds to cause swing angles, derived from GUMBEL and WEIBULL distributions, their
respective parameters being locally reevaluated and the criteria used to advance from the former
to the latter being clarified
 Resulting swing angles, according to up-to-date analytical models (CIGRÉ and Hornisgrinde)
 Relative insulation strength of air gaps between conductors and towers, for both 60 Hz voltages
and switching surges, based on hourly observations
 Risk of failure of the whole transmission line
1
Though not aiming at setting up any possible overall pattern whatsoever, this paper places due
emphasis on the importance that local data be acquired and statistically treated in order to achieve
best field performances and profit from statistical advances overhead lines state of art has already
found as a benchmark. A sensibility analysis of attained results for transmission tower weights
while comparing overall suggested parameters and local ones is presented as well aiming to
strengthen this remark. To achieve this, typical towers for different voltage levels are used.

KEYWORDS
swing angle, reliability, statistics, environment, test span

1 INTRODUCTION
It is almost amazing that a simple situation comprising an insulator string in its stand still
position and the same string pushed towards an overhead tower that holds it can be such a
source of random variables, partly due to the local environment that surrounds the system
and partly originated within electrical elements.
In fact, the random elements are very distinct from one
another and can be described as follows:
■ The wind speed that causes the displacement has a
largely varying intensity and direction along line spans
and will affect them in different ways, depending on the
nature of the phenomenon: most of the time, it is a cold
front or a local thunderstorm or still both altogether. All
this to focus attention on winds of definite statistical
characteristics, and leaving off discussion those so called
tornados, typhoons, cyclones, among others less prone
to statistical evaluations.
Figure 1.1 - A vertical insulator ■ To simulate the string displacement due to wind action
string, in its stand still position (from now on just swing angle) various mathematical

models are available, leading to dispersed results, different


line costs and conflicting performances.
■ The reduced air gap has withstand voltages that
intrinsically depend on the air around it, namely: its relative
air density (RAD), air temperature (tamb) and absolute
humidity (habs). All these variables have considerable
oscillations both in time and space, taken into account that
an Overhead Transmission Line (OHTL) normally has some
hundreds of miles in length and stretches out along some
number of altitudes and many degrees in latitude, meaning a
great deal of climatic conditions. In some instances, a
sparkover will occur, be it under normal voltage (worst and
most disturbing outcome) or be it under an overvoltage Figure 1.2 - The same
(switching surges), when a certain risk can be assumed. string under wind action
■ The switching surges themselves have their own statistical nature and are handled by
specific TNAs simulations.
At the very end, the theoretically simple situation previously described has in its core many
(maybe “too many”) specifics to be handled as some kind of a trifle challenge and deserves a
closer attention. This report tackles them separately but with different degrees of depth.

2 WIND SPEEDS COMING FROM THE GUMBLE DISTRIBUTION


Nowadays, and for transmission line design purposes, wind speeds and pressure on each
component have become a synonym of the GUMBEL Distribution, for which [1] is the
worldwide best point of reference. For decades, a major number of OHTL have had their
ultimate loads accordingly calculated, on the basis of 250 years for EHV Systems and 150

2
years for HV ones. This refers to Brazilian current practice and may subject to minor changes
when compared with what goes on in other countries. In reference [1], all details are given on
how to apply the GUMBEL statistics, provided some reliable data with yearly average wind
speeds is available. This kind of information will not be repeated here for clear-cut reasons.
Nevertheless, some clues thereof are shown in APPENDIX A, just for the sake of
completeness. What will be here stressed is the concept of confidence interval that is often
neglected and turns out to be an important detail concerning this kind of application.

One approach to GUMBEL’s


confidence interval is given in detail
in [5]. The point here is that for short
periods of data acquisition (5 to 10
years) the confidence interval is
wide (40 km/h, see Figure 2.1), even
though fitting procedures for
GUMBEL has a correction for such
cases and [1] is very specific on this
in its Appendix C.4.

Figure 2.1 - GUMBEL plot for sample size = 8


The shorter the sample size is, the
higher is the correction. A glimpse to
Figure 2.2 indicates that the
confidence interval decays to 20 km/h
when sample size is 25 elements.
This kind of evaluation emphasizes
the relevant role yearly maxima
sample size plays and points out in
the direction of a permanent search of
ever longer historical series. With the
help of a wind database and the
concepts outlined in [1], all kinds of
isotach maps can be drawn, either Figure 2.2 - GUMBEL plot for sample size = 25
using gust factors IEC recommends or, what is more realistic, one’s own local factors
resulted from specific measurements. Figure 2.3 shows the isotach map for wind speeds for
T = 2 years, as return period and tavg = 10 minutes for averaging time.

The state of São Paulo, Brazil


was chosen as an example since
that’s where a more dense and
outspread wind speed acquisition
network is available. Besides, this
map shows typical wind speeds
that are currently in use in Brazil
to design transmission towers
accounting for switching surges.

Figure 2.3 - Isotachs (m/s) for São Paulo state (Brazil)

3
2.1 What is odd with the 2 year wind speed?
Following GUMBEL rules, wind speeds of 250 or 150 years/10 minutes have since long been
used to properly determine ultimate loads on towers due to wind. As for swing angles under
60 Hz voltages, 50 year wind speeds are commonly used in Brazil, which has led to good
practical results since 60 Hz voltages must be related to rare or high wind speeds for obvious
safety reasons. As an extension of common usage of GUMBEL, the 2 year /10 minute
speeds (from now on just v2/10) were deemed appropriate to handle switching surges, in other
words, a rarer phenomenon associated with a lower wind speed..
In fact, this is the least speed that can be provided by GUMBEL
(see APPENDIX A.1) and yet it is too high to occur with any
switching surges, on a common sense basis. On this issue,
reference [3] offers the WEIBULL distribution to mathematically
assess lower speeds without the need to resort to some subjective
reduction factor.
Still according to [3], an overall WEIBULL shape factor (ß) of 2.0 is
sugested as the best call to handle this migration from GUMBEL to
WEIBULL. Nevertheless, when checked with what was obtained in
5 Brazilian anemographic stations, the β value oscillated from 0.67
to 1.63, results that are way distant from what was sugested (see
Figure 2.1). In other words, if local values are available, they should
replace theoretical values to avoid unecessary bias in the results
here targetted. It is worth emphasizing that with longer historical
series and more spreaded stations, a closer match to the β = 2.0
can be found in Brazilian territorry.
There is another relevant assumption before WEIBULL distribution
be used. It lies in the probability avlue that can be associated with
the v2/10 wind speed. Reference [3] handles the challenge by means
of the number of readings v2/10 comes from. This is a relevant
statistical feature, in other words, what is the probability this V 2/10
wind speed should assume on WEIBULL (from now on, “binding
probability”)? It depends on the sample size [3] and from a practical
stand point, the maximum value would be 2(years) x 365(days) x
24(hours) x 6(every 10 minutes) = 105,120 readings (say, 100,000),
assuming wind data deriving from a daily anemogram paper, which
is a common case in Brazil, to these days. This would lead to a
0.001 % probability (1 outcome within 2 years). It should be born in
mind that the shape factor and some probability for the v2/10 wind
speed mathematically suifice to define the appropriate WEIBULL
Figure 2.1 - WEIBULL parameters to represent an all-the-time wind speed distribution. An
fit in 5 wind speed adequate means to evaluate a wind speed to coincide with an
Brazilian stations eventual switching surge and having a lower value when compared
to v2/10 would in this way be provided to OHTL designers
Figures 2.2 and 2.3 show results for WEIBULL with ß = 1.0 and 2.0; binding probabilities
0.001 and 1.0%; wind speeds with 10 to 25% probability of being exceeded. Nevertheless,
this way to tackle the problem has some degree of clumsiness since collecting protocals
varies from station to station and sometimes from year to year within the same station.
Furthermore, in certain instances, the sample size of v2/10 comes from is an unknown variable
or it was lost in time. The following Table summarizes results for varying binding
probabilities, shape factors 1.0 and 2.0 and wind speeds on WEIBULL for probabilities
ranging from 10 to 25%.

4
Figure 2.2 - WEIBULL plot for v2/10 = 18 m/s, shape factor = 2, binding probabiities of
0.001and 1%; other wind speeds for x% probability of being exceeded

Figure 2.3 - WEIBULL plot for v2/10 = 18 m/s, shape factor = 1, binding probabiities of
0.001and 1%; other wind speeds for x% probability of being exceeded
Table 2.1 - Different Wind speeds to generate swing angles
based on WEIBULL distributions (shown in blue)
wind speeds (m/s) in blue referred to x % in the WEIBULL distribution
(1) (2) (3) (4) (5) (6) (7) (8) (9)
% in shape factor ß = 2 shape factor ß = 1
Weibull 10 % 15 % 20 % 25 % 10 % 15 % 20 % 25 %
0.001 8.0 7.3 6.7 6.2 3.6 3.0 2.5 2.2
0.01 9.0 8.2 7.5 7.0 4.5 3.7 3.1 2.7
0.1 10.4 9.4 8.7 8.1 6.0 4.9 4.2 3.6
1 12.7 11.6 10.6 9.9 9.0 7.4 6.3 5.4
(1) % in the WEIBULL distribution assumed for v2/10 = 18 m/s, all cases)
(2) to (5) wind speeds for x % on WEIBULL of being exceeded, ß= 2.0
(6) to (9) wind speeds for a risk of x % on WEIBULL of being exceeded, ß = 1.0

Some conclusions can be drawn from Table 2.1 that could be handy when choosing either
the shape factor or the “binding probability (bp)”:
► The higher the shape factor, the more conservative are wind speeds for any other
probability (10, 15, 20 or 25%). Therefore, in the absence of local measured values, 2.0
should be preferred.
► The higher the binding probability is, the higher are wind speeds, similarly to what was
previously stated. Therefore, unless otherwise ascertained, 1% would be suitable.
► The choice of the appropriate percent value (here assumed from 10 to 25%) on
WEIBULL to enable the determination of a wind speed to swing the insulator string back
and forth under switching surges perhaps remains as the weakest and most controversial

5
feature and eventually having wide oscillations, according to what is shown in Table 2.1.
Suggestions so far would be premature.
► Since the target here is to set up a reduction, and a realistic one, in the wind speed for
switching surges with respect to the former v2/10 value (around 18 m/s), it is feasible the
proper speed will lie somewhere in between 18 and some guessing figure around 10 m/s,
taken as reference. Under such premises, adopting β = 2.0 pu and bp = 10% seems to do
the job, even though the value for β is still to be investigated prior to any final acceptance.
► To improve matters, an inquiry related to the real statistical occurrence of switching
surges in OHTL’s and the weather conditions under which they take place in the field
seems to be prone to provide a good way out.

2.2 Wishful thinking


All issues described in 2.1 will be solved soon enough. Within the next decades, updated
wind speed acquisition systems and protocols will provide the worldwide electrical branch
with continuous readings in the rate of 1 second; sonic sensor data will be stored in huge
magnetic media and duly subjected to statistical analysis afterwards. Consequently, all
measuring stations will have their own GUMBEL and WEIBULL historical samples, 10 to 20
years long. No assumptions will then be needed concerning shape factors or binding
probabilities and the approach described in [3] will then become solider and wide spread.

3 MATHEMATICAL MODELS TO APPROACH SWING ANGLES


So far, there is no unanimity on which mathematical is to be used for swing angles
calculations. International ones are the CIGRÉ model and that developed in Hornisgrinde, to
mention those well known. Other national applications that accommodate local needs or
preferences
In Brazil, the national code uses an
analytical model similar to Hornisgrinde’s
but with a different manipulation without
any field measurement support. Results
deriving from all these models vary
widely as reported in [3], Figure 3.1
being a copy of what is published there.
Precisely noticing and stressing, on the
20 m/s mark, mean differences range
from 16 to 38º, values which are large
enough to cause worries on the present
state-of-art and places the whole
approach at stake..
Figure 3.1 - Swing angle deviations
between calculated and measured values
The 20 m/s wind speed is close to what is currently used in association with switching surges
on EHV systems

4 RELATIVE INSULATION STRENGTH OF AIR GAPS


Reference [2] thorough and painstakingly describes all protocols and details to assess any
value of RIS (relative Insulation Strength) of air gaps, under 60 Hz, surges and even direct
current. Long weather database containing ambient temperature, air pressure and absolute
humidity having an hourly acquisition rate can give means to a statistical evaluation of RIS
values along an OHTL, provided measuring stations are located close to the line
environment. To accomplish this, the following assumptions can be adopted:
■ Calculations can be carried out for a 3 m gap. Gaps higher than this value are on the safe
side while shorter gaps will have an error not greater than 2%.

6
■ For 60 Hz voltages, it is
recommended that OHTL
design uses a RIS minimum
value of µ ± 3.00 σ (guaranteed
99.99% of the time) or µ ± 2.33
σ (guaranteed 99.00% of the
time), depending on the
required performance.
■ Gap factor can assume a 1.3
pu value, which is appropriate
for conductors to tower. Gap
factors of 1.3 x (±1.2) will not
influence results over 3%
■ For any RIS mean value, their
respective coefficients of
variation will not exceed 3%,
according to Brazilian
Figure 4.1 - RIS values exceeded 99% geographical situation stretched
of the time (60 Hz) out from equatorial up to tropical
zones and altitudes rarely
exceeding 1500 m, considering
OHTL’s applications or design.

Figure 4.2 - RIS values exceeded 90%


of the time (switching surges)

5 RISKS OF FAILURE
Under operating voltage, even if the withstand voltage of an air gap has an intrinsic statistical
value defined by a mean value and a standard deviation, no risk is admissible. Therefore, the
withstand voltage for any practical purpose will be corrected by some low mean value of the
RIS along the line less 3 σ (exceeded 99.99% of the time) or 2.33 σ (99% safe). Both seem
to guarantee a proper performance for swing angles related to 50 y / 10 minute wind speeds.
As for switching surges, the best way to tackle to problem has already been recommended in
many publications [6] and deals with the concept of risk of failure. Figure 5.1 represents the
numerical integration which includes both surges and withstand voltages distributions; and
the number of gaps along a certain physical extension of an OHTL under the influence of
adverse weather conditions (cold fronts or thunderstorms, mostly). Further details can be
deepened with [6] and shall be omitted here, except that for EHV systems rough values of
10-4 to 10-5 have been used in Brazil to set up a satisfactory grid operation.

7
The feasible number of gaps
plays a relevant role and must
account for the number of
vertical strings per tower and the
full spacial magnitude of the
weather phenomenon: thunder
storms will simultaneously affect
some few spans (say, 10) and
cold fronts will affect many more
spans (say, 50 to 100).

Figure 5.1 -

6 PRACTICAL CASES
TACTA (table)

swing angles (º)


tower 20 25 30 35
types tower weights (tons)
230 kV
self - supporting
345 kV
(whatever)
500 kV
self - supporting
500 kV
guyed “V” tower

Any number of towers will do. The more, the better!


7 COMPARISON AMONG ATTAINED RESULTS

8 NEEED OF ENHANCEMENTS VIA TEST SPANS - MANGEMENT AND


TECHNICAL FEATURES

SJS (text)

9 CONCLUSIONS

8
End of text

BIBLIOGRAPHY

[1] IEC Publication 60826 -Design Criteria of Overhead Lines, 2003


[2] IEC Publication 60-1 - High Voltage Test Techniques,- Part 1 . General Definitions and
Test Requirements, 1989
[3] CIGRÉ - Working Group B2.06, Tower Top Geometry and Mid Span Clearances, 2008
[4] JIF, da Silva, RM, Azevedo et alli, Fatores de Correção Atmosféricos Aplicados ao
Dimensionamento de Isolamentos em Ar - Nova Metodologia de Cálculo (Relative
Insulation Strength Applied to the Design of Air Gaps - A New Approach), XVIII
SNPTEE, 2003
[5] GR, Kendall, Statistical Analysis of Extreme Values, First Canadian Conference
Symposium on Spillway Design Floods, Ottawa, November 4-5, 1959
[6] EPRI - Report Summary, L, Shen, Evaluation of the Results of Several Full-Scale
Conductor Wind Loading Experiments, TR-104479, December 1994
[7] IEC Publication 60826 -Design Criteria of Overhead Lines, 2003
[1] IEC Publication 60826 -Design Criteria of Overhead Lines, 2003
[1] IEC Publication 60826 -Design Criteria of Overhead Lines, 2003

You might also like