Вы находитесь на странице: 1из 18

Return Period for Environmental Loads Combination of Wind and

Wave Loads for Offshore Wind Turbines


Claus F. Christensen and Torben Arnbjerg-Nielsen
RAMBLL, Bredevej 2, DK-2830 Virum
Abstract: In order to reduce the CO2 emissions the focus on renewable energy from
wind turbines has increased in the last years. Due to the size of the wind turbines and
the large area a wind turbine park requires it is reasonable to place the parks at sea.
Building offshore wind farms generates needs for new knowledge about the
environmental conditions and how the different loads should be combined.
The present paper describes an approach for obtaining the environmental condition that
corresponds to the maximal response for given return period. The environmental
conditions are here the mean wind velocity and the significant wave height, but the
described procedure is general and can easily be expanded. Moreover the corresponding
partial safety factor is obtained for the operation design case and the extreme design
case.

1. Introduction
Design of offshore wind turbines is in general based on simulations of the response
corresponding to predefined time series. The present paper describes an approach for
the determination of these time series of the external environmental loads to be applied
in the calculation of the offshore wind turbine response.
The approach is a very general approach and it is illustrated by examples with two
correlated environmental parameters, i.e. mean wind velocity and significant wave
height for two different directions which corresponds to two different correlation levels.
The approach is based on inverse FORM. By means of inverse FORM environmental
contours can be determined independent of the structural response. Contours
representing extreme environmental conditions, e.g. 50-year return period can be
examined and the corresponding maximum structural response can be obtained.
The scope of the present paper is to establish the return period for the different
environmental loads that corresponds to an e.g. 50-year return period for the response
and the corresponding partial safety factor.

2. Code format
The proposed code format for the determination of the extreme response is illustrated in
Figure 1. The external loads are given in terms of combinations of time series for a
number of individual loads, e.g. wind, wave, current, etc. For each of these combined
time series of environmental loads the maximum response is determined.

The simulations are carried out for a number of time series, and the extreme response is
defined as the average value of the obtained extreme values for the various time
simulations.
Static and dynamic loads
F(t) = Fi(t)
Response
Max{R(t)}
Design extreme
Rd = f E[Max{R(t)}]
Resistance
Mc/m

Figure 1:

Design Extreme
Mc/m > Rd

Illustration of design process time simulation.

The time series of the external loads are to be defined so that the average value of the
extreme response found by the simulation is an estimate of the characteristic value of
the response (50-year return period). The design value is found by multiplying the
characteristic response with the partial safety factor and it is required to be larger than
the design value of the carrying capacity.
In the following sections an approach is outlined for the determination of the
environmental condition which leads to the 50-years return period of the response.
Further the corresponding partial safety factor f is estimated.
3. Contours of environmental parameters
Contours of environmental parameters are contours along which specified extreme
fractiles lie, ref. /1/. The contours are independent of the structure under consideration,
making the contours a general and practical way of illustration extreme combinations of
environmental parameters.
The approach is based on first-order reliability method (FORM), or in fact inverse
FORM in the sense that the environmental contours are identified corresponding to
hyperspheres with a given radius in the normalised space, ref. /1/ and /2/.
The extreme response of a given return period is then identified as the maximum
response along the contour for the corresponding to the required return period. In
addition the point on the contour corresponding to the maximum response, identifies the
combination of the environmental parameter, which yields the maximum response for
the wanted return period.

The basis for the evaluation of the environmental contour is the instantaneous
distributions for the stochastic environmental parameters.
In the present case two parameters are considered, wind and waves, but the approach is
applicable for higher dimensions as well. The instantaneous distributions are given as
the distribution of the mean wind velocity V and the significant wave height Hs
conditioned on the mean wind velocity
FV (v) and FHs V (h v)

(1)

The mean wind velocity and wave height may also depend on the direction and the
distributions could therefore also be conditioned on direction. Especially the significant
wave height is sensitive to the fetch length and water depth and may therefore vary a lot
with the direction.
Utilising inverse FORM, ref. /1/, the basis is the reliability index , corresponding to
the required return period. The following relationship exists between the reliability
index , the probability pf, and the required return period Tr

= 1 (1 p f ) = 1 (1

TSS
)
365 24Tr

(2)

where TSS in hours is the duration of the environmental conditions, e.g. mean wind
periods or seastates and Tr is the return period in years.
In the standard normal space the interesting combinations are thus located on a
hypersphere of radius corresponding to the required return period Tr, described by
U 12 + U 22 =

(3)

where U1 and U2 are the coordinates in the standard normal space, from which the
physical environmental parameters are found from transformation, see ref. /3/,
V = FV1 ( (U 1 ))
Hs = FHs1V ( (U 2 ))

(4)

3.1 Example: TSS for a 10-minute environmental condition


In case an n-years return period is required for a 10-minute environmental condition,
e.g. mean wind velocity, the corresponding effective period TSS must be obtained. Due
to the large correlation between two subsequent mean wind velocities this period is not
10 minutes.

Based on a time series of 17 year containing the 10 minutes mean wind velocity
measured 45 m above ground, ref. /9/, the correlation function is estimated to

(t ) = exp( t / T )
where T = 20.16 hours.
Based on the given correlation function the effective number of observations nef in a
time series containing n, successive 10 minutes wind velocities can be found as, ref. /8/
1 2
nef = + 2
n n

(n j ) j

j =1

n 1

(5)

where j is the correlation between to mean wind periods with the distance j times 10
minutes.
The value of Tss is then obtained as 10 minutes times n divided by nef, which yields a Tss
of about 40 hours or 241 periods of 10 minutes.
In case a 50-year return period is required for the environmental condition the
corresponding reliability index reads

= 1 (1 p f ) = 1 (1

40
) = 3.74
50 365 24

(6)

For other return periods the corresponding reliability index is shown in Table 1 for the
considered environmental condition.
Return period Reliability index
(year)
()
5
3.12
10
3.32
50
3.74
100
3.91
200
4,08
500
4,28
1000
4,44
10000
4,91

Probability
(pf)
9.1310-4
4.5710-4
9.1310-5
4.5710-5
2.2810-5
9.1310-6
4.5710-6
4.5710-7

Table 1: Return period and corresponding reliability index and probability for the
considered environmental condition.

3.2 Example: Contours at Horns Rev


The distribution at Horns Rev is based on 3 month of measurements covering all
directions. The significant wave height is measured every hour and the 10 minutes mean
wind velocity is measured once each hour.
The mean wind velocity V in a height of 60 meters is assumed to follow a twoparameter Weibull distribution given as
v
FV (v) = 1 exp(( ) k )
a

(7)

where the parameters a is 11 m/s and k is 1.8. The distribution is used independent of
direction and is in general found to fit well to the Danish wind climate.
The used distribution for the significant wave height conditioned on the mean wind
velocity is based on approximately two months data. The 10 minutes average wind
velocity is measured once each hour together with the significant wave height measured
every hour. The distribution is obtained from two directions. The first direction Case 1
covers the angle space 240 to 320, where north corresponds to 0. The other
direction Case 2 corresponds to the angle space 40 to 110. Due to the very limited
data only the obtained distribution for case 1 is reasonable, whereas the distribution for
case 2 is very uncertain and serves only for illustrative purpose.
The significant wave height Hs, conditioned on the mean wind velocity, V, is assumed
to follow a Gaussian distribution with the parameters given in Table 2 below
Hs conditioned on V
Case 1 (240 to 320)
Case 2 (40 to 110)

Mean Value
0.13V
0.07V

Standard Deviation
0.24 m
0.50 m

Table 2: Mean and standard deviation for the Gaussian distributed significant wave
height HS conditioned on the mean wind velocity in meter per second.
Based on the assumed distributions the correlation between the mean wind velocity and
the significant wave height can be calculated theoretically to 0.95 for case 1 and 0.62
for case 2, respectively.
It is noted that the Gaussian distribution for the significant wave height includes
negative values for small mean wind velocities. However as the present analysis
addresses extreme values this is of no importance.
In the standard normal space the environmental condition along the contours
represented by
U 12 + U 22 =

(8)

can be transformed into contours in the physical space utilising the above equations and
the following transformations
V = FV1 ( (U 1 )) = a( Ln(1 (U 1 ))1 / k

(9)

Hs = FHs1V ( (U 2 )) = U 2 + V

where and are (0.13; 0.24) for case 1 and (0.07; 0.50) for case 2, and a is 11 m/s
and k is 1.8.
The obtained contours for the two cases are shown in Figure 2 and the correlation
between the environmental loads is clearly reflected in the shape of the contour.
6,0

4,0

Return period 5 year


Return period 10 year
Return period 50 year
Return period 200 year
Return period 1000 year

5,0

3,5
3,0

4,0
2,5

3,0

2,0
1,5

2,0
Return period 5 year
Return period 10 year
Return period 50 year
Return period 200 year
Return period 1000 year

1,0

1,0
0,5

0,0

0,0

10

15

20

25

30

35

Mean wind velocity [m/S]

40

45

10

15

20

25

30

35

Mean wind velocity [m/S]

Figure 2: Environmental contours Horns Rev. Left: Case 1: direction WNW with high
correlation. Right: Case 2: direction E with low correlation.

3.3 Example: Contours at Rdsand


Studies of joint distributions for significant wave height and wind velocity has been
reported in ref. /4/, based on an analysis of the environmental conditions at Rdsand.
In ref. /4/ the significant wave height is taken as the unconditioned parameter and the
mean wind velocity is then conditioned on the wave height. The significant wave
height, Hs, in an arbitrary sea state is assumed to follow a two-parameter Weibull
distribution
FH S (h) = 1 exp((

h h
) )
h0

(10)

where h0 = 0.863 m and h= 1.817. The 10-minute mean wind velocity V conditioned on
the significant wave height is assumed to be log-normal

40

45

ln(v) u
FV (v H S ) =
u

(11)

The parameters at Rdsand has been estimated as


D[V ] 2
) + 1) and u = ln( E[V ]) 0.5 u2
E[V ]

(12)

E [V ] = a 0 + a1 H S and D[V ] = 0.2889 ln z + 0.4918

(13)

u = ln((
with

where Hs is the significant wave height in meters and z denotes the height in meters
above sea level and a0 and a1 is given as
a 0 = 0.368 ln z + 1.7582 and a1 = 1.0667 ln z + 5.771
In ref. /4/ it is stated that the auto-correlation function for the significant wave height
can be adequately represented by a two-parameter exponential decay model as
a
(t ) = exp( at ) (cosh(bt ) + sinh(bt ))
b
where the parameters a and b are estimated to 0.837 and 0.793, respectively.
Based on the estimated correlation function and the formula given in equation (5) the
effective number of 1 hour sea-states in a time series with n observation can be
obtained. The duration between two independent 1 hour sea-states is then obtained to 46
hours.

= 1 (1 p f ) = 1 (1

46
)
Tr 365 24

(14)

where Tr is the return period. For a 50-year return period the target is 3.71.
Using the transformation given in equation (3) and (4) the environmental condition for
Rdsand can be obtained. The contours are given in Figure 3 for z equal to 70 m. It is
seen from the figure that the correlation between the mean wind velocity and the
significant wave height at Rdsand is high. Moreover it is seen that the environmental
parameters in general are somewhat smaller at Rdsand than at Horns Rev.

3,5
Return period 5 year
Return period 10 year
Return period 50 year
Return period 200 year
Return period 1000 year

3,0

2,5

2,0

1,5

1,0

0,5

0,0
0

10

15

20

25

30

35

40

Mean wind velocity [m/S]

Figure 3: Environmental contours for mean wind velocity and significant wave height
at Rdsand

4. Extreme response environmental parameters


In the previous section the relevant combinations of environmental parameters
corresponding to a given return period are identified. These combinations are to be
combined with the response of the structure under consideration in order to determine
where on the contour the extreme value lays for the given return period. This point will
further more define the combination of external environmental parameters to be applied
in the deterministic design approach.
The used response model is described in ref. /5/. The model calculates the mean value,
r, standard deviation, r, and the standard deviation on the response velocity, r ,
based on given wave and wind spectra. The spactra depend on the mean wind velocity
and significant waveheight.

The extreme value of the response is modelled by a filtered Poisson process, i.e.
Fmax (r , t ) = exp( t (1 F (r ))

(15)

where is the intensity of the Poisson process, t the time period over which the extreme
is considered and F(r) is the distribution of the pulses in the response process.
For a normal distributed response r as in ref. /5/ with the mean value and standard
deviation ( r, r) the expected number of up-crossings of the level is given by Rices
formula as
1
r
( ) = 0 exp
2 r

(16)

where 0 denotes the expected number of up crossings of the mean level r per unit of
the time t given by

0 =

1 r
2 r

(17)

Having the response given as a Gaussian process described by the normal distribution
function as stated above and by assuming that each up crossing of the level is
independent, the distribution for the extreme in the period t can be approximated by

1 r 2
Fmax ( ) = exp 0 t exp( (
) )
2 r

(18)

The median value of the maximum response, which is slightly lower than the average
value, in the period t is thus given as
ln(2)
+ r
= r 2 ln
0 t

(19)

4.1 Example: Response applied for illustration


The considered wind turbine used together with the environmental contours for Horns
Rev is an active stall-regulated turbine with a hub height of 63.5 m, placed on a water
depth of 9 m. The lowest natural frequency of the structure is 0.401 Hz and the rotor
disc are is around 3160 m2.
The sea roughness is put to 0.004 m and the turbulence intensity is assumed constant
with a value of 10.34 %. The waves are assumed to follow a Pierson-Moskowitz
spectrum where the peak period is a function of the significant wave height given as
T p = 190 H s / g

(20)

where g is the gravity. The response model is described in details in ref. /5/.
The design response for a wind turbine in operation is typically in the same range as the
design response for a parked turbine under extreme conditions. The established model
for the wind turbine response takes this automatically into account by changing the
wind load when the mean wind velocity becomes larger than the so-called cut-out mean
wind velocity, which in this case is put to 25 m/s. For larger mean wind velocities the
wind turbine will be parked and the wind response will hereby reduced. The considered
response is the overturning moment and the shear force at the seabed and the calculated
response is seen in Figure 4 where the change in response at V=25 m/s also is very
distinct.

Figure 4: Response for maximum expected shear force (left) and maximum expected
overturning moment (right).
From the figures it is seen that the mean wind velocity governs the overturning moment,
whereas the influence of the significant wave height is very limited. The shear force is
most sensitive to the significant wave height, but both environmental parameters are
important when the shear force is considered.

4.2 Example: Extreme response environmental parameters, Horns Rev


The maximal response for a given return period is then obtained by searching on the
contours shown in Figure 2. For the operation design case the mean wind velocity is
limited at 25 m/s and only the significant wave height increases when the return period
increases. The maximal shear force and overturning moment are in all cases obtained
for the same combination of wind and wave load. The results are given in Table 3 and
Table 4.
Operation situation (25 m/s) Extreme situation
Return Response Hs [m] Return pe- Response Hs [m]
period [MN]
riod for Hs [MN]
5
1.22
3.78
1
1.21
4.31
10
1.24
3.86
1.5
1.25
4.52
50
1.26
3.99
2
1.38
5.00
100
1.27
4.05
2.5
1.44
5.21
200
1.27
4.10
3
1.49
5.41
500
1.28
4.15
3.5
1.56
5.66
1000
1.29
4.20
4
1.61
5.84

V [m/s] Return pe- Return period for Hs riod for V


32.67
5
5
33.94
10
10
37.83
40
47
39.37
67
93
40.89
104
188
42.77
145
461
44.18
174
922

Table 3: Characteristic shear force response and return periods for the individual
environmental loads for the different response return periods for Horns Rev
case 1 (large correlation).

10

Operation situation (25 m/s)


Extreme situation
Return Response Hs [m] Return pe- Response Hs [m]
period [MNm]
riod for Hs [MNm]
5
34.2
3.78
1.0
25.0
4.31
10
34.2
3.86
1.5
27.0
4.50
50
34.3
3.99
2.0
33.0
5.00
100
34.3
4.05
2.5
35.6
5.21
200
34.3
4.10
3.0
38.4
5.41
500
34.4
4.15
3.5
41.9
5.66
1000
34.4
4.20
4.0
44.7
5.84

V [m/s] Return pe- Return period for Hs riod for V


32.67
5
5
34.05
9
10
37.83
40
47
39.37
67
93
40.89
104
188
42.77
145
461
44.18
174
922

Table 4: Characteristic overturning moment and return periods for the individual
environmental loads for the different response return periods for Horns Rev
case 1 (large correlation).
From Table 3 and Table 4 it is seen that the characteristic overturning moment is largest
in the operation situation for return periods up to 50 years, whereas the characteristic
shear force in the operation situation is largest up to a return period of 10 years.
Moreover it is seen that when the correlation between the loads is high, the return period
for the loads in the extreme situation is close to the return period for the response. In the
operation situation the wave height is limited due to the high correlation and becomes
almost constant for any return period. This emphasises that two design situation are to
be considered operation and extreme environmental conditions.
Results for case 2 with low correlation are given in Table 5 and Table 6.
Operation situation (25 m/s) Extreme situation
Return Response Hs [m] Return pe- Response Hs [m]
period [MN]
riod for Hs [MN]
5
1.07
2.83
5
0.91
2.71
10
1.09
2.97
9
0.96
2.85
50
1.14
3.25
32
1.07
3.13
100
1.16
3.35
48
1.12
3.25
200
1.17
3.45
67
1.17
3.36
500
1.19
3.57
102
1.23
3.47
1000
1.21
3.66
123
1.27
3.59

V [m/s] Return pe- Return period for Hs riod for V


30.48
3
2
32.31
5
5
36.22
19
23
37.81
32
47
39.40
50
95
41.51
74
252
42.86
107
482

Table 5: Characteristic shear force response and return periods for the individual
environmental loads for the different response return periods for Horns Rev
case 2 (low correlation).

11

Operation situation (25 m/s)


Extreme situation
Return Response Hs [m] Return pe- Response Hs [m]
period [MNm]
riod for Hs [MNm]
5
33.8
2.83
5
23.4
2.40
10
33.8
2.97
9
25.9
2.53
50
34.0
3.25
32
32.0
2.82
100
34.0
3.35
48
34.7
2.93
200
34.0
3.45
67
37.4
3.04
500
34.1
3.57
102
41.0
3.19
1000
34.1
3.66
123
43.8
3.29

V [m/s] Return pe- Return period for Hs riod for V


32.31
1
5
34.05
1.4
9.5
37.83
5
47
39.37
8
93
40.89
13
188
42.77
24
461
44.18
37
922

Table 6: Characteristic overturning moment and return periods for the individual
environmental loads for the different response return periods for Horns Rev
case 2 (low correlation).
In the second case where the correlation between the loads is more limited, it is seen
that the operation situation is governing the design for return periods up to around 50 to
100 years. Moreover it is seen that for the shear force response, which is slightly wave
dominated the return periods for the environmental loads are only about half of the
return period for the response. For the overturning moment, which is dominated by the
wind load due to the long internal arm, the reduction in return period is not that
significant. It is also seen that when the correlation between the loads is limited the
significant wave height increases in the operation situation for increased return periods.
4,0

6,0

Max response
Return period 5 year
Return period 10 year
Return period 50 year
Return period 200 year
Return period 1000 year

5,0

4,0

3,5
3,0
2,5

Max. operation response

Operation

Extreme

2,0

3,0

1,5

2,0

1,0
1,0

0,5

Max shear force


Max overturning moment
Return period 5 year
Return period 10 year
Return period 50 year
Return period 200 year
Return period 1000 year

0,0

0,0
0

10

15

20

25

30

Mean wind velocity [m/S]

35

40

45

10

15

20

25

30

35

Mean wind velocity [m/S]

Figure 5: Location of the maximum response shown on the environmental contours


from Figure 3. Left: Horns Rev case 1: direction WNW with high correlation.
Right: Horns Rev case 2: direction E with low correlation.
From Figure 5 it is seen that if the correlation between the loads is low, the load
combination causing the maximum response is not obvious and the combination may
depend on the type of response. Figure 5 right, shows that the maximum expected shear
force (the full black circles) will occur for a larger significant wave height than the
overturning moment (the stars), which is dominated by the wind load.

12

40

45

In order to evaluate the benefit of the given procedure, the response is calculated for the
50-year return period for both environmental loads for the extreme situation, and a 50year return period for the significant wave height together with a mean wind velocity of
25 m/s for the operation situation. The 50-year return period corresponds to a mean
wind velocity of 38.29 m/s and significant wave height of 5.09 m for the first case with
large correlation and 3.36 m for the case with low correlation.
Operation situation
Correlation Hs [m] V [m/s] Shear force Return period Overturning
Return period
V [MN]
for V
moment M [MNm]
for M
Large
5.09
25.0
1.41
990,000
34.80
2,800,000
Low
3.36
25.0
1.16
79
34.00
86
Table 7: Extreme responses for a 50 year return period for both environmental loads.
Extreme situation
Correlation Hs [m] V [m/s] Shear force Return period Overturning
Return period
V [MN]
for V
moment M [MNm]
for M
Large
5.09 37.97
1.40
70
33.26
68
Low
3.36 37.97
1.14
128
32.48
58
Table 8: Operation responses for a 50 year return period for the significant wave height
and a mean wind velocity of 25 m/s.
From Table 7 and Table 8 it appears that if the correlation is large in the operation
situation a 50-years significant wave height is extremely conservative. The 25 m/s
corresponds to a very low return period (approximately 0.37) and due to the strong
correlation the significant wave height will be in the range of return period. In the
extreme situation and when the correlation is low the benefit is more limited, but though
still considerable.
It should be noted that the given procedure assumes that the maximal response cannot
occur for a lower return period than the considered, which is a non-conservative
assumption. There is a minor probability of that the largest response can occur for a load
combination that corresponds to a lower return period, but this probability is minor and
will decrease with the length of the considered period.

5. Partial safety factor response


In most codes it is the individual environmental loads that are multiplied by a partial
safety factor and not the response as suggested in this case. For an offshore wind turbine
the relation between the environmental loads and the response is non linear in some
cases and the suggested procedure is therefore be more suitable.
According to the code format, see Section 2, the maximal expected response should be
multiplied by the partial safety factor, and the safety factor does therefore depend on the
coefficient of variation of the extreme response.

13

The total coefficient of variation for the structural response contains the following
contributions. A contribution due to the statistical uncertainty caused by the random
nature of the load process, a contribution covering the physical uncertainty in how the
flow converts into load on the structure (for the wind load this term is around 25 %
according to ref. /7/) and finally a contribution covering the model uncertainty. The
coefficient of variation for the model uncertainty for the wind is according to ref. /7/
around 14 %.
The variation of the extreme response obtained by the procedure described in Section 4
can be established by determining the maximal response for different return periods, i.e.
different fractiles in the upper tail of the response distribution.
Using the obtained response values for the chosen fractiles an extreme value distribution
can be fitted to these values. In this case a Gumbel distribution is fitted by use of a least
square fit, and the Gumbel distribution seems to fit very well to the obtained values. The
Gumble distribution fits very well because the dependence between the response and the
Weibull distributed mean wind is almost linear, especially for the case with large
correlation. Knowing the parameters in the distribution fitted to the upper tail the
coefficient of variation of the responds can be calculated.
In Figure 6 the fitted Gumble distribution is shown together with the calculated values
for the extreme situation and large correlation between the environmental loads.
0,98

0,98

0,94

0,94

0,90

0,90

0,86

0,86

0,82

Emperical

0,82

Emperical

F-Gumbel
0,78

F-Gumbel
0,78

1,2

1,3

1,4

1,5

Max Response: Shear force [MN]

1,6

24,0

28,0

32,0

36,0

40,0

44,0

Max Response: Overturning moment [MNm]

Figure 6: Empirical and fitted extreme distribution for the extreme situation with
large correlation between the environmental loads.
The corresponding figures for the operation situation for the case with large correlation
is given in Figure 7.

14

0,98

0,98

0,94

0,94

0,90

0,90

0,86

0,86

0,82

0,82

Emperical

Emperical
F-Gumbel

F-Gumbel
0,78

0,78
1,20

1,22

1,24

1,26

1,28

1,30

Max Response: Shear force [MN]

34,10

34,15

34,20

34,25

34,30

34,35

34,40

34,45

Max Response: Overturning moment [MNm]

Figure 7: Empirical and fitted extreme distribution for the operation situation with
large correlation between the environmental loads.
For the operation situation the extreme distribution fit is not unique, but the coefficient
of variation on the operation response is very low because the mean wind velocity is
constant (25 m/s) and only the significant wave height increases when the return period
increases. The obtained coefficient of variation on the response in the different
situations is given in Table 9.

Correlation
Large
Low

Operation situation
Extreme situation
C.o.V. Shear
C.o.V. Moment C.o.V. Shear
C.o.V. Moment
1.6 %
0.2 %
8.3 %
19.0 %
4.1 %
0.3 %
10.8 %
24.6 %

Table 9: Obtained coefficient of variations.


For an offshore wind turbine the physical uncertainty may be lower than stated in ref.
/7/ which covers normal land structures, because the external pressure coefficient, the
terrain topography and roughness is well determined. Moreover the uncertainty in the
load transformation from the waves may be lower than the wind and a contribution
around 18 to 20 % instead of the 25 % may be more realistic. (The 25 % reduce to 18 %
if the uncertainty contribution from the terrain topography and roughness is neglected).
The total uncertainty used to determine the safety factor is then calculated as shown in
the following example taken from ref. /7/.
Considering a structure only exposed to wind load the total coefficient of variation can
be calculated as
(1 + Vtot2 ) = (1 + Vq2 )(1 + Vc2e )(1 + Vc2p )(1 + V J2 )

(21)

where Vq , Vce , Vc p and VJ are the coefficients of variation of the extreme wind, exposure
coefficient, external pressure coefficient and the model uncertainty.
The partial safety factor is then a function of the obtained total coefficient of variation
Vtot for the wind load.

15

34,50

The safety factors in the Danish code DS 409 ref. /6/ for environmental loads are
obtained by calibration at two values of the coefficient of variation i.e. 20 % and 40 %
where the corresponding safety factors are found to be 1.3 and 1.5. Assuming a linear
variation around and between these to values the connection between the total
coefficient of variation and the partial safety factor is obtained and shown in Figure 8.
1,6
1,5
1,4
= C.o.V. + 1,1
1,3
1,2
1,1
10%

Figure 8:

20%
30%
40%
Total coefficient of variation

50%

Safety factor as a function of the coefficient of variation.

In order to use the relationship shown in Figure 8 the coefficient of variation for the
response must be obtained and the contributions form the uncertainty in the
environmental parameters, transformation from flow (wind and wave) to pressure and
the model uncertainty must be included.
In the purposed code format the total coefficient of variation is obtained as
(1 + Vtot2 ) = (1 + Vq2 )(1 + Vc2 )(1 + V J2 )

(22)

where Vc is the total uncertainty for the transformation of the loads to response
approximately around 18 to 20 %.
The obtained partial safety factors for the operation and extreme situation are given in
Table 10 for the two environmental cases obtained for Horns Rev.
Variable
Shear
Moment
Shear
Moment
Shear
Moment
Shear
Moment

Design Correlation CoV. CoV. TransSituation


Response formation
Extreme
Large
8.3%
20.0%
Extreme
Large
19.0%
20.0%
Extreme
Low
10.8%
20.0%
Extreme
Low
24.6%
20.0%
Operation
Large
1.6%
20.0%
Operation
Large
0.2%
20.0%
Operation
Low
4.1%
20.0%
Operation
Low
0.3%
20.0%

CoV.
Model
14.0%
14.0%
14.0%
14.0%
14.0%
14.0%
14.0%
14.0%

CoV.
Total
26.0%
31.4%
27.0%
35.3%
24.6%
24.6%
24.9%
24.6%

Safety
factor
1.36
1.41
1.37
1.45
1.35
1.35
1.35
1.35

Table 10: Obtained partial safety factors.

16

From Table 10 it is seen that the partial safety factor for the operation situation should
be lower than in the extreme situation. Moreover the variation in the extreme situation is
larger than in the operation situation where the governing uncertainty is the model
uncertainty and the physical uncertainty. These two uncertainty contributions can be
reduced when a precise and accurate load model is used and when the knowledge about
the load process is increased.
Applying the safety factors in Table 10 together with the obtained characteristic
response given in Table 3 to Table 6 the design loads shown in Table 11 can be
obtained.
Response variable
Shear force
Overturning moment
Shear force
Overturning moment
Shear force
Overturning moment
Shear force
Overturning moment

Design Correlation Safety Characteristic


Situation
factor
response
Extreme
Large
1.36
1.38 MN
Extreme
Large
1.41
33.0 MNm
Extreme
Low
1.37
1.07 MN
Extreme
Low
1.45
32.0 MNm
Operation
Large
1.35
1.26 MN
Operation
Large
1.35
34.3 MNm
Operation
Low
1.35
1.14 MN
Operation
Low
1.35
34.0 MNm

Design
response
1.88 MN
46.7 MNm
1.47 MN
46.5 MNm
1.69 MN
46.1 MNm
1.54 MN
45.7 MNm

Table 11: Design loads for the operation and extreme situation for the case with large
correlation and low correlation between the environmental loads.
From Table 11 it is seen that for the case with large correlation the largest design load is
obtained for the extreme situation, where as for the case with low correlation the
operation situation is governing the design shear force and the overturning moment is
governing by the extreme situation. The variation in the design loads is though limited.

6. Conclusions
The environmental contours for the mean wind velocity and significant wave height are
established and the maximal structural response is obtained for the operation situation
and the extreme situation. Moreover it is found that the auto-correlation is important in
order to estimate the effective number of environmental conditions for the considered
return period.
Furthermore the correlation between the environmental loads are important. It is found
that in case of large correlation the benefit of the described procedure is very large for
the operation situation and limited for the extreme situation. For a low correlation
between the loads the benefit is in general more limited, but still considerable.
Based on the calculated maximal response for different return periods an extreme
distribution is estimated and the partial safety factor can be obtained. The obtained
safety factors for the extreme design situations are in general in the same range as the
partial safety factors given in the Eurocode and the Danish code, ref. /6/ and /10/. The
partial safety factors obtained for the operation situation are a bit lower due to the

17

limited uncertainty, and it seams reasonably to use smaller partial safety factors for the
design operation situation.

7. Acknowledgements
The paper form part of EFP J.nr. 1363/99-0007 Designgrundlag for vindmlleparker
p havet (Design Regulations For Offshore Windfarms) which is financial supported
by the Danish Ministry of Environment and Energy and the power supply companies
ELSAM and SEAS.

8. References
/1/

Winterstein, S.R, T.C.Ude, C.A.Cornell, P.Bjerager: Environmental parameters


for extreme response: Inverse form with omission factors, Structural Safety &
Reliability, 1994 Balkerna, Rotterdam.

/2/

Madsen, Krenk, Lind: Methods of Structural Safety, Prentice-Hall, 1986.

/3/

Ditlevsen, Madsen: Structural Reliability Methods, Wiley, 1996.

/4/

Ronold, Knut: Probabilistic Stability Analysis of Wind Turbine Foundation on


Clay in Cyclic Loading, EFP99, December 2000.

/5/

Tarp-Johansen, N. J., Frandsen, S.: A Simple Offshore Wind Turbine Model for
Foundation Design, EFP99, December 2000.

/6/

DS 409, Danish code of Practice for the Safety of Structures, 1998.

/7/

NKB Committee and Work Reports 1999:01 E, BASIS OF DESIGN OF


STRUCTURES Proposals for Modification of Partial Safety Factors in
Eurocodes.

/8/

Conradsen, K., Nielsen, L. B. and Prahm, I. P.: Review of Weibull Statistics for
Estimation of Wind Speed Distributions, Journal of Climate and applied
Meteorology, Vol. 23, No., 8, August 1984.

/9/

Kristensen, L., Jensen, G., Hansen, A. and Kirkegaard, P.: Field Calibration of
Cup Anemometers, Ris National Laboratory, January 2001.

/10/ Draft prEN 1990, Basis of Design, European Standard, September 2000.

18