IWAIS XI, Montréal, June 2005

Collecting and using weather data for the design of overhead lines according
to IEC 60826

Lars Rolfseng Svein Fikke Elias Ghannoum

Sintef Energy Research AS Meteorological consultant Consultant
NO-7034 Trondheim, Norway NO-1470 Lørenskog, Norway 76 Claude Champagne, Montreal,
Quebec, Canada H2V 2X1

Abstract—IEC TR 60826 introduced the probability concept in economical benefits of such programs of recording ice and
the design of overhead electric power lines. While statistical wind data are demonstrated.
distributions of wind speeds are available at national
meteorological services, similar information on ice loadings can
only be provided by utilities themselves. This paper describes the
III. EXAMPLES OF WEATHER DATA FOR ICE AND WIND
consequences for investment costs in power lines due to For the purpose of comparisons, the same set of
uncertainties in ice load assessments, as well as the importance of observations on ice and wind loads are used in this paper as
long time series of both wind and ice data. those presented in CIGRÉ Technical Brochure 109: Review of
IEC 826 "Loading and Strength of Overhead Lines" [2]. The
I. INTRODUCTION
ice and wind load data are described in the following

T he International Electrotechnical Commission (IEC) has

published Technical Report 60826 which was the first
international recommendation on design of overhead lines
paragraphs.
A. Ice
based on reliability concepts [1]. This report deals with Time series of ice loadings are according the following
methods to calculate wind and ice load on overhead lines recorded annual maxima of ice loadings1:
starting with measurements and field observations of wind
speed and ice accretion. Year Ice load Year Ice load
Ice- and wind loads on conductors and earth wires are the (N/m) (N/m)
most important environmental factors for the mechanical 1980 10 1986 8
design of overhead lines. Wind has by itself a universal 1981 9 1987 6
meteorological interest and wind loads have a wide common 1982 8 1988 7
interest for the design of all types of structures as well as for 1983 11 1989 34
1984 32 1990 4
transportation on roads, at sea and in the air.
1985 3 1991 23
Ice loads on structures and overhead power lines are, on
the other hand, mainly of interest for the electric power
From the observations above the average ice load value is
utilities and telecommunication companies in areas exposed to
calculated to 12,9 N/m and the standard deviation to 10,7
ice and snow accretions. This fact reflects the large N/m, i.e. a coefficient of variation equal to 0,82.
differences in quality of the statistical databases for wind and The number of years with observations, the average value
ice loads. Wind data are generally available with sufficient and the standard deviation are the necessary parameters for
quality from most meteorological offices, while acquisitions making the statistical analyses of design wind and ice loads
of ice data are quite rare. for an overhead line.
This paper addresses the importance of long-term
recording/measurements of weather data and describes how B. Wind
these may be utilized according the 60826 report for the The time series of the maximum annual 10 min. wind
design of overhead lines. speeds (perpendicular to the line) is measured 13 m above
ground in terrain with ground roughness B for 29 years:
II. DESIGN METHODOLOGY IN IEC 60826
Year Wind speed Year Wind speed
It is demonstrated in IEC 60826 that if a load QT (load (m/s) (m/s)
having a return period T) is associated with a strength of 1956 16,5 1971 15,4
components corresponding to the 10% exclusion limit, the 1957 15,4 1972 15,4
resulting reliability is almost constant and equal to 1/(2T) in 1958 17,0 1973 20,6
the normal range of variation of Q and R. Thus, the accuracy 1959 15,9 1974 15,4
of the estimates of QT becomes quite important if one aims to 1960 25,7 1975 13,4
design lines according to a target reliability level.
In the following examples, the importance of increasing the 1
In this example, ice data is assumed to be measured as a unit weight per
number of years of wind and ice data is demonstrated and the conductor length. alternatively, ice data can also be expressed in radial ice
thickness
IWAIS XI, Montréal, June 2005

1961 13,4 1976 18,5 period of n years, the value zi is calculated as follows:
1962 15,4 1977 12,9
1963 14,4 1978 13,4 i
1964 16,5 1979 14,4 zi = − ln(− ln ) where 1 ≤ i ≤ n (4)
n +1
1965 15,4 1980 14,9
1966 22,6 1981 14,4 n
1
1967
1968
17,5
15,4
1982
1983
13,9
17,0
C1 = σ z =
n
∑ z i2 − z m2 (5)
i =1
1969 13,4 1984 14,9 i.e. C1 is the standard deviation of the zi values
1970 17,5
n
1
From the observations the average wind load value is
calculated to 16,1 m/s and the standard deviation to 2,8 m/s,
C2 = z =
n
∑ zi (6)
i =1
i.e. a coefficient of variation equal to 0,18. i.e. C2 is the average of the zi values.

IV. CALCULATION OF CLIMATIC DESIGN LOADS FOR OVERHEAD It is noted that the above form of the Gumbel distribution
LINES ACCORDING IEC 60826 incorporates a bias for the number of years of data. This form
is the one preferred in IEC 60826. In some other cases, the
A. Reliability levels general form (C1=0.57722, C2= 1.28225) have also been
used.
The design loads are calculated according to the chosen
reliability level given by the return period of the climatic load. C. Design load with return period T
In IEC 60826 the reliability levels are defined in Table 1
below. Equation (1) gives the probability F(x) of not exceeding the
value x. The probability of not exceeding the value x(T) i.e.
TABLE 1. RELIABILITY LEVELS FOR OVERHEAD LINES the value with a return period equal to T is:
Reliability levels 1 2 3
Return period, T, of climatic F(x(T)) = 1 – 1/T (7)
design loads, in years 50 150 500
If the natural logarithm is taken twice on both sides of the
equal sign of equation (1) the result is:
B. The Gumbel distribution function
x(T ) − p1
The climatic design loads according to these reliability ln(-ln F(x(T))= (8)
p2
levels may be calculated using the Gumbel cumulative
distribution function. Equation (1) below gives the probability
Inserting the expressions for p1 (equation 2), p2 (equation
F(x) of not exceeding the value x.
3) and F(x(T)) (equation 7), we get the following relation for
x− p
the design value and return period T:
− 1
p2
F( x ) = e−e (1) x(T) = x −
C2
σ−
σ ⎡ 1 ⎤
ln − ln(1 − )⎥ (9)
C1 C 2 ⎢⎣ T ⎦

Where
Equation (9) shows that the climatic design value
calculated by the Gumbel distribution function, only depends
C2
p1 = x − σ (2) on the chosen return period, T, average value, x , standard
C1
deviation, σ, and number of observations (i.e. the constants C1
σ and C2).
p2 = (3)
C1
V. NUMERICAL EXAMPLE OF DESIGN LOADS FOR ICE AND
where WIND OBSERVATIONS IN CIGRÉ BROCHURE 109

x = average value A. Ice loads

σ = standard deviation
The average value for the 12 annual maximum ice load
The factors C1 and C2 depend on the number of values is 12,9 N/m, and the standard deviation is calculated to
observations in a measurement series. For a measurement 10,7 N/m that means a variation coefficient of 0,8 (80%). This
is somewhat higher than the indicated variation of 0,7 in IEC
IWAIS XI, Montréal, June 2005

60826. If a return period of 50 years is chosen (often used for increasing the number of years with observations is largest for
electric distribution systems) the design value according to the fewest number of years with observations. An increase of
equation (9) is calculated to be approximately 50 N/m. If now observations from 5 to 10 years will lead to an ice load
the design load is calculated for different number of years of reduction of about 14% while the corresponding reduction by
observations based on the same fixed average value of 12,9 increasing the observations from 55 to 60 years only is about
N/m and standard deviation of 10,7 N/m, the resulting design 0,5%. This increase can be simply explained as follows: When
loads for a return period of 50, 150 and 500 years are statistics are based on a large number of data, the prediction
calculated in table 2 below. of, say a 50 year load, will be more precise than if the same
was based on a reduced number of data. The above equation
TABLE 2. DESIGN ICE LOADS AS FUNCTION OF NUMBER OF OBSERVATION No 9 attempts to compensate the reduced number of years of
YEARS
Design ice loads (N/m)
data by increasing the predicted 50-year load.The shape of the
Number of C - coefficients Return time T years curves in Figure 1 shows that the larger part of the uncertainty
years is eliminated after 30 years of observations. If the number of
n C1 C2 50 150 500 observation years is increased from 5 to an infinite number of
5 0,458794 0,792778 59,2 74,0 90,2 years the total reduction of the design ice load is roughly 30%.
10 0,495207 0,949625 51,1 63,5 77,0
20 0,523552 1,062822 46,8 57,8 69,9
Figure 1 reveals that if you have made observations for 10
30 0,536221 1,112374 45,1 55,7 67,3 years, another 10 years of observations will reduce the design
40 0,54362 1,141315 44,2 54,6 65,8 load nearly 10%. An increase from 20 to 30 years of
50 0,548542 1,160661 43,7 53,8 64,9 observations will reduce the design loads with further about 3
60 0,552084 1,174665 43,3 53,3 64,2
100 0,560023 1,206489 42,4 52,2 62,8
%.
200 0,567153 1,235977 41,6 51,2 61,6
300 0,569926 1,247866 41,3 50,8 61,1
400 0,571437 1,254501 41,2 50,6 60,8 B. Wind loads
500 0,572398 1,2588 41,1 50,4 60,6
∞ 0,57722 1,28255 40,5 49,7 59,7 The average value and standard deviation for the 29 annual
maximum wind speed values are considerably lower than for
The ratio of the design ice load values based on only 5 the ice loads.Following the same procedure as for ice, the
observations to the design ice load values based on an infinite wind speed with a return period of 50 years is calculated to be
number of observations is about 1,5. This means that if you 24,7 m/s. If the design wind speed is calculated for different
have a statistical basis of only 5 years with observations, the numbers of observations based on the same fixed average
design value in this case is 50% higher than a design value value of 16,1 m/s and standard deviation of 2,8 m/s, the
based on infinite number of years with observations. The main resulting design loads for a return period of 50, 150 and 500
reason for the large uncertainty for a low number of years years are shown in Table 3.
with observations for ice loads is the high value of the
standard deviation of the observations. Even if you have TABLE 3. DESIGN WIND LOADS (M/S) AS FUNCTION OF NUMBER OF
observations for more than 30 years, a couple of years with OBSERVATION YEARS

unusual or low high ice loads may influence the design load. Number of Design wind loads [m/s]
years C - coefficients Return time T years
If 10 years of observations is chosen as reference, Figure 1
n C1 C2 50 150 500
shows the relative reduction in the design values as a function 5 0,458794 0,792778 28,4 32,4 36,7
of number of observations n, for return period, T, equal to 50, 10 0,495207 0,949625 26,3 29,6 33,2
150 and 500 years. 20 0,523552 1,062822 25,1 28,1 31,3
30 0,536221 1,112374 24,7 27,5 30,6
40 0,54362 1,141315 24,4 27,2 30,2
50 0,548542 1,160661 24,3 27,0 29,9
Ice load reduction as a function of years of observations 60 0,552084 1,174665 24,2 26,9 29,8
referred to 10 years of observations
100 0,560023 1,206489 23,9 26,5 29,4
200 0,567153 1,235977 23,7 26,3 29,1
20,0
300 0,569926 1,247866 23,7 26,2 28,9
15,0 400 0,571437 1,254501 23,6 26,1 28,9
10,0 500 0,572398 1,2588 23,6 26,1 28,8
50 Infinite 0,57722 1,28255 23,4 25,9 28,6
5,0
150
%

0,0
500
-5,0
5 10 15 20 25 30 35 40 45 50 55 60 In this case the ratio between the design wind values based
-10,0
on only 5 observations and the design wind values based on
-15,0
an infinite number of observations is about 1,25. This is half
Number of years
of the variation for design ice loads, due to the much lower
standard deviation for the annual maximum wind speeds.
Figure 1. Reduction in design ice loads as a function of number of observation
years referred to 10 years.
If 10 observations is again chosen as reference, Figure 2
Table 2 and the curves in Figure 1 show that the effect of
IWAIS XI, Montréal, June 2005

shows the relative reduction of the design wind values, as a corresponding economic value will be 15 million $. In some
function of number of observations n for return time, T, equal cases this would result in somewhat higher annual outage
to 50, 150 and 500 years. costs. It is obvious that the above economy assumes that the
average and standard deviation of ice data remains the same.
In order to have a better coordination of investment costs
Wind speed reduction as a funtion of years of observations and maintenance costs in the grid a ice measurement program
referred to 10 years of observations
on existing overhead lines was suggested to the network
utilities. The benefit of this would be either
10,0
ƒ reduced investment costs if the recommended and
5,0
50 highly uncertain design ice loads are to high, or
ƒ the damage to the transmission system and loss of
%

0,0 150

-5,0
5 15 25 35 45 55 500 customer supply due to ice storms could be reduced
if the recommend design ice value is too low.
-10,0
Number of years It is most likely that the first alternative is more relevant in
Norway. The long term experience is that extensive damage
Figure 2. Reduction design wind speed as function of number of observation and loss of supply during heavy icing periods are quite rare.
years referred to 10 years of observations.
Table 4 above indicates that to achieve a 10% reduction of the
If the observations are increased from 10 years to 20 years, uncertainty the Norwegian utilities could annually spend 3
the design wind speed will be reduced approximately 5%. If million $. The fact is that far less is probably needed to
the observation series is increased to 50 years the reduction is achieve such a reduction from where we stand today.
in the range of 8 – 10%. Norwegian transmission and distribution utilities are however
In this case it is observed that the relative reduction by strongly focused on short term economical results and even
increasing the number of observation years from 30 to 60 small long term investments have low priority. Attempts to
years is only giving a reduction of the design wind loads in initiate such ice monitoring and recording program in Norway
the range of 2 – 3%. have therefore failed.

VI. ECONOMY OF ICE LOAD REGISTRATIONS VII. THE STUDNICE MEASUREMENTS IN THE CZECH REPUBLIC.

Table 4 contains information on the current length and the All estimates on ice and wind statistics above were based
“as new” cost of the overhead lines at the different voltage on the assumption of a “constant climate”. This means that it
levels in Norway. The economical lifetime of the overhead would make no difference during which time period wind and
lines is set to 40 years and the marginal cost of ±1 kg ice load ice measurements were taken. In particular, the above
per km overhead line corresponds to about 6% of the overall estimates of the uncertainties in ice loads were based on this.
investment cost for a new overhead line. The uncertainty of This assumption will be discussed in this section.
the existing ice load recommendations in Norway is At the mountain Studnice approximately 800 m of altitude
conservatively estimated to 40%. Based on these assumptions in the Czech Republic, ice loads are measured since 1940 and
Table 4 also gives information on “price of uncertainty” of the annual maximum ice load continuously recorded. The results
design ice load for different voltage level in Norway. from the measurements up to 1998 are given in figure 3 below
[3].
TABLE 4. TOTAL LENGTH OF OVERHEAD LINES IN NORWAY AND ECONOMIC
EFFECTS OF UNCERTAINTIES IN DESIGN ICE LOADS IN US $
Iceload registrations at Studnice, Czech Republic
Voltage level <1 1 - 24 33 - 66 132 300 420 Sum
Total length of OHL [km] 116376 63856 10391 9483 5086 2104 207295
Average ice load = 5,8 kg/m Standard deviation = 5,2 kg/m
Value as new, [Billion $] 2,82 3,07 1,09 1,44 2,38 1,09 11,88
Annual replacement [km] 2909 1596 260 237 127 53 5182 25
Maximum ice load in kg/m

Annual replacement, [Million $] 70,4 76,7 27,2 36,0 59,5 27,1 296,9
Average ice load [kg] 1,5 3,5 5 5 7 7 - 20
+/- 40% cost of uncertainty, [Millon $] 2,53 6,44 3,27 4,32 9,99 4,56 31,1
15

The total “price of the uncertainty” in the design ice loads 10

for the total electric overhead line network in Norway is about 5

30 million $. This represents about ±10% of the annual 0

replacement value of the grid. If this uncertainty could be
19 41

19 44

19 47

19 0

19 53

19 56

19 9

19 62

19 65

19 8

19 71

19 74

19 77

19 80

19 83

19 86

19 9

19 92

19 5

8
-5

-5

-6

-8

-9

-9
-

-
40

43

46

49

52

55

58

61

64

67

70

73

76

79

82

85

88

91

94

97

reduced by 10% (i.e. from 40% to 36%), this would imply that
19

Season
about 3 million $ less would have been invested in an
unknown, and probably in most cases undesired, safety. If the Figure 3. Annual maximum ice load registrations at Studnice.
uncertainty in ice loads could be reduced to 20%, the
IWAIS XI, Montréal, June 2005

The average maximum ice load is 5,8 kg/m and the Measurements of wind loads in a coordinated manner are
standard deviation is 5,2 kg/m giving a variation factor of 0,9. common worldwide and the data basis for the calculation of
From these measurements the ice loads with a return period of design wind loads is normally good and much better than the
50 years are calculated for each period of 10 years by using data basis for ice loads.
equation (9) above and corresponding C-factors. In addition In all cases there is an economic incentive to measure and
the ice loads with a return period of 50 years are calculated for collect for a long time ice and wind data for the design of
the accumulated period of 10, 20, etc. up 60 years by using overhead lines. Such long periods of records will provide the
equation (9). The result is given in figure 4 below. best estimates of design ice and wind loads. If the design loads
are underestimated due to short periods of data, the failure rate
Ice loads loads with return period of 50 years
at Studnice
of lines will increase and cause serious economical
consequences. If on the other hand, the design loads are
50 overestimated, the initial cost will be much higher and may
40 reduce funds available for other projects.
30 10 year period
kg/m

20 Accumulated IX. REFERENCES

10
0
[1] IEC 60826 Ed. 3: ”Design criteria of overhead
transmission lines”. 2003
19

19

19

19

19

19
40

50

60

70

80

90
-5

-6

-7

-8

-9

-9
0

Period
[2] CIGRÉ, Technical Brochure 109: Review of IEC 826
Figure 4. Design ice load for a return period of 50 years based on each
"Loading and Strength of Overhead Lines". 1996
separate 10 year period and the total number of years.
[3] CIGRÉ Session 1998. Paper 22-105: “Ice monitoring at
As seen in Figure 4, predicting ice loads based on a stand Studnice. Tuned vibration control of overhead line
reduced number of years, such as 10 years, may lead to under- conductors” F. Popolansky, J. Kruzik, J. Hrabanek, and
or overdesigning overhead lines. J. Lago.
For example, if the 50 year load prediction was based on
the 1980-1990 period, the 50 year ice load would have been
near 10 kg/m, while if the 1990-1997 period was selected, the
50 year load would have been 40 kg/m.
If, on the opposite, the total period of ice observations from
1940 to 1997 were selected, the predicted 50 year load would
have been 20 kg/m.
The above findings confirm the importance of relying on a
long series of climatic data records in order to estimate ice
loads corresponding to return periods of 50, 150 or 500 years,
particularly in countries where ice loads do not occur every
year.

VIII. CONCLUSIONS

The paper addresses the procedure to calculate design

weather loads by applying the Gumbel cumulative distribution
function as described in IEC 60826. The procedure seems to
be adequate for its purpose and it is quite simple to apply. The
IEC report also contains curves and tables to further simplify
its use.
The chosen example of ice load measurements and records
made at Studnice show that the standard deviation of
maximum annual ice loads is large (a variation factor in the
range of 0,8 to 0,9). Due to this, calculations of design ice
loads based on few years of observations may be imprecise.
Even simple ice load records may improve the data basis for
the calculation of the design ice loads substantially and
acquiring such data is very cost effective and highly
recommended.