henderson isope2004 loading

8/13/2019 Henderson ISOPE2004 Loading

1/8

Paper No. 2004-ARH-01 Henderson 1

Hydrodynamic Loading on Offshore Wind Turbines

A.R. Henderson

Ceasa

Stroud, United Kingdom

M.B. Zaaijer

Section Wind Energy, Delft University of TechnologyDelft, The Netherlands

ABSTRACT

In the shallow seas that are the favoured locations for offshore windfarms, the limited water depths can result in highly non-linear waves.

The determination of the design wave loads involves the selection ofappropriate models of wave kinematics as well as force and structuraldynamics models. Each selection will involve a compromise between

accuracy and usability (speed, ease of use and comprehension of theresults). Currently, the offshore oil and gas industry is focusing on everdeeper waters as much of the available hydrocarbon resource inaccessible medium depth waters is already being exploited. In contrast,

offshore wind energy is being developed in shallower waters, though

often at sites exposed to extreme weather such as the European NorthSea. The approach needed here will be subtly different but equallydemanding, with economic reasons being more prominent. Electricityis a low-value commodity in a highly competitive market and the costsof generation using offshore wind farms are approaching the costs of

conventional generation. It is vitally important that inappropriate andexcessively conservative design approaches do not sabotage this trend.Although existing offshore design methods can undoubtedly result in a

durable structure, there may be excessive cost penalties. On the otherhand, non-linear and breaking waves experienced in shallow watersmay mean that the design methods based on experience in deeper waterare unconservative. The coastal engineering branch does have

substantial experience in designing in shallow water conditions, albeitagain to much more stringent durability criteria than are appropriatehere.

This paper gives a final report on the work undertaken on this subjectwithin the OWTES 'Design Methods for Offshore Wind Turbines at

Exposed Sitesresearch project and examines the following aspects:

different recommendations for slender (monopiles) structures,

the affect of the shallow water on wave climate,

investigation of the uncertainties due to the selection ofappropriate models and the associated parameters regarding:(i) wave kinematics models, (ii) wave load models, (iii) and

structural models.For the evaluation of proposed engineering models for slenderstructures, the paper draws prominently on data collected at the Blyth

offshore windfarm, where one turbine is comprehensively instrumentedand an extensive collection of measurements, including extreme waves,

have been recorded. Further details are available in the reports, Camp,T.R. (2003), which describes the whole project, and Henderson, A.R.

(2003), which focuses on hydrodynamic loads, written as part of thisresearch project.

INTRODUCTION

The calculation and determination of design wave loads on offshore

structures is a complex undertaking involving different wave models,load-calculation methods and probability analyses. It is however ofvital importance if a cost-effective and durable structure is to bedesigned. Both the extreme and fatigue load cases need to be

considered and the actual approach may differ for these two cases and

for different support structures. The key to the problem is to determinethe nature of the waves: their distribution and their hydrodynamic

properties.The procedures necessary to calculate the critical wave loading, foreither the fatigue or extreme cases, can be divided into three stages:

(i) determining the design wave or wave climate(ii) selecting an appropriate wave load calculation procedure

(iii) determining the effect on the structure

Each stage is of importance for achieving an appropriately designsolution and cannot be considered in isolation, as they are interrelated:for instance, the design wave can depend on the structural responsewhen a larger wave at a frequency away from the structure's natural

frequency can be less critical than a smaller wave close to the naturalfrequency. Hence an important aspect in the prediction of extreme- and

fatigue loading of the support structure of an offshore windturbine canbe its dynamic response. The predictability of this dynamic responsediffers in some important aspects from that of platforms for theoffshore oil industry and of onshore wind energy converters. Thenatural frequency of an offshore windturbine can be wedged between

different excitation frequencies, whereas the natural frequency of afixed platform for the offshore oil industry is usually designed to bewell above the wave excitation frequencies. The geometry and

dimensions of offshore foundations differ from typical onshoresolutions, resulting particularly in a larger influence of soilcharacteristics for the slender monopile foundation.At this moment, the size of the offshore windenergy market does not

warrant intensive research on developing new and bespoke methods,

ISOPE Conference, Toulon, France, 23-28 May 2004.


2/8


and hence judgment of appropriateness and applicability of existingmethods, which can easily be a very subjective process, is needed.

Points of concern with the application of the existing offshoreengineering methods include:

uncertainties in the wave modelling, because of increasednon-linearities

increased occurrence and importance of (near-) breaking

waves inappropriate safety margins

When considering the hydrodynamic loading aspects of the designprocess, both the extremeand the fatiguecases need to be considered,each of which can require a different approach. Defining the design

procedure consists of selecting an appropriate wave or wave climate,kinematic model, loading model and structural model. Only a limitednumber of permutations of these models are possible (an example beingthat diffraction cannot be applied to high order wave theories) and the

more sophisticated or accurate the model, the greater the demands onthe engineers and the computers time.

Determination of the Wave Kinematics

First the external loads need to be defined and for windturbine support

structures the hydrodynamic wave forces are of great interest (Incontrast, hydrodynamic loads have limited impact on the design of the

rotor and nacelle). When the windturbine is located in a sea, it willencounter a lifetime of waves of varying sizes and forms. How can this

be distilled into a limited number of cases that can be dealt with in atimely and cost-effective manner and yet represent the full-life

experience of the structure?Several wave kinematics models have been compared with the

measurements and these are briefly described below:(i) Airy; or linear wave theorys, Airy, Sir G. B. (1845), relative

simplicity is both its strength and weakness, in that what it lacks inaccuracy itself, it can compensate by being integrated with otheraspects of the wave-load calculation process, such as corrections,

stochastic waves, diffraction etc. Its primary weakness is that itcuts off wave peaks and troughs

(ii) Wheeler Stretching Wheeler, J.D. (1969) is an example of theflexibility of the Airy wave in that the kinematics as calculated at

the mean-water level are applied up to the true surface with thedistribution down to the seabed being stretched accordingly; thisgives an improved prediction of the wave kinematics and this

correction can also be applied to simulations of stochastic seas(iii) Deans Stream Function Dean, R.G., (1965); has largely

superseded all other wave-theories for regular waves where Airyis insufficient. It is a numerical solution, which can be applied up

to near breaking wave height.Stokes and Cnoidal theories found favour in the past however are nowsuperseded by stream function theory. Boussinesq theory Boussinesq,

J (1872) is receiving renewed research attention Madsen, P.A., (2000)but within the engineering profession its use is still limited. Othercorrections to linear theory include constant and extrapolated crest;these can be excessively conservative in deeper waters (but are

satisfactory in the water depths of interest here) and so are in limiteduse by engineers.

Fig. 1 compares the wave profile for the above wave models for anextreme wave

1 (7th order stream function recommended by Barltrop,

N.D.P. (1991)) all derivatives of the linear wave theory (i.e. Wheeler,constant and extrapolated crest) assume the same sinusoidal profile. It

can be seen that, relative to linear theory, the non-linear theories predictthat waves have (i) sharper crest and flatter trough profiles and(ii) higher crest and trough elevations. This is also clearly visible in

1for 10m 15s wave in 21m water

stochastic seas as the surface recording from Blyth in Fig. 2 shows. Aweakness of low order theories when applied to extreme waves is

apparent in Fig. 1 in that the profile of the second order Stokes theoryincludes an erroneous higher order harmonic and in that the exact wavelength for the fifth order Stokes theory could not be calculated because

the algorithm used here failed to converge.

0 50 100 150 200 2505

0

5

10

horizontal [m]

vertical[m]

Fig. 1: Wave Surface Profile

-2

-1

0

1

2

3

4

0 20 40 60 80 100 120 140Time [s]

WaveSurface[m]

Fig. 2: Recorded Sea Surface (Blyth)

Fig. 3: Recommended Wave Model, Barltrop, N.D.P. (1991)

Finally Fig. 3 shows which wave model would typically be

recommended for traditional offshore engineering versus offshorewindfarm engineering for both fatigue and extreme cases. For the case

of traditional offshore engineering, the fatigue region is still sufficientlyclose to the linear zone that a modified linear method, i.e. Wheeler or

Chakrabarti stretching, is acceptable, while stream function theory isrecommended for the extreme case. However, for an offshorewindfarm, depending on the site, this might not be the case and a

LinearStokes 2nd OrderStokes 5th order

Stream Function

Windfarm

Extreme

& Fatigue

Deep Water

Extreme &

Fatigue



3/8


modified linear method might underestimate the fatigue wave loads aswell as the extreme wave loads. Since waves will generally be an

important but not the dominant source of fatigue loading, this is notcritical though the designer should be aware of this issue.

Calculation of the Wave Loads

Of the available methods for calculating the wave loads listed in Table1, the two most widely used are:

Morison's method, usually in the time domain, used forslender structures, such as monopiles and tripods,

Diffraction theory, used for massive structures, such asgravity base supports

In addition the Froude-Krylov (or pressure integration) method offersthe advantage of being able to model massive and complex structuralgeometries with any wave model Chakrabarti, S.K. (1987); diffractionhas to be estimated in a similar manner as for Morison's method but themore complicated geometries makes this harder to perform. In the

situations where this method could offer the most beneficial results, i.e.gravity base structures in shallow water, the wind loads on the turbinetend to dominate the design process Srensen, H. C. (2000). In the

longer term, CFDoffers promising benefits of being able to model all

aspects, though at undoubted penalties of time and clarity.For the determination of waveloads on monopiles, the Morison method

is undoubtedly most suitable in the design process though diffractionand Froude-Krylov may offer insights for researchers.

Table 1: Wave load calculation Methods

Morison

Diffraction

Froude-

Krylov

CFD

Time / Frequency Domain TD FD FD TD TD

inertia Transverse

drag 5 X X

Lateral (drag) 5 X X Forces

Pressure X 1 X 1

Diffraction X 2 X 2 X3

1D X SurfaceEffects 6 3D X X X

Geometry

Massive Structures X X

Non-linear & extrapol. waves X X WaveModel Stochastic (Linear) X

4 X

4

Commercial Availability *** *** *** ** *

Ease of Use *** ** ** ** *Applicability

(* = poor /*** = good) Calculation Speed ** *** *** * *

1= can be modelled relatively easily by adding an extra term2 = can be modelled using MacCamy-Fuchs MacCamy, R.C. (1954)

correction for simple shapes3= must be estimated4= high demands on computation power5= linearised6= non-linear surface effects between the structure and the wave-field:1D = in vertical direction only (i.e. wave height considered only at thevertical-axis of the structure)

3D = full geometric field (i.e. wave height at each surface element ofthe structure)

HYDRODYNAMIC LOADING EXAMINATION OF THEORY

In this section, the wave loads calculated using different kinematicmodels is examined. In each case, Morison theory is used, since this ismost appropriate for the geometry under consideration. A table of

waves with period ranging between 5 s and 15 s at 1 s intervals and theheight ranging between 1 m and 7 m at 1 m intervals is applied. The

structure was assumed to be a 4 m diameter monopile, with very highroughness due to age, hence CM=1.79 and CD=1.55 as recommended by

DNV (2000). The water depth was taken as 21 m. (Note theseparameters are different from those at Blyth since these are likely to be

more representative of future offshore windfarms). Using wave theoryselection charts (such as in Barltrop, N.D.P. (1991)) it is found thatnon-linear theory is recommended even for the smallest 1 m waves.Similar charts have been developed to aid the selection of the load-model Chakrabarti, S.K. (1987), which would show that diffraction

effects are straight-forward and that both drag and inertia are important,hence the Morison method should be utilised.

T

6 8 10 12 14

5

10 1 107

1 107

1 107

5 106

5 106

5 106

Fig. 4: Overturning Moment calculated using Airy Theory

T

6 8 10 12 14

5

10

2 107

2 107

1.5 1071.5 10

7

1 107

1 107

1 107

5 106

5 106

5 106

Fig. 5: Overturning Moment calculated using Wheeler Theory

This group of charts shows the calculated maximum values of thewave-induced overturning moment. Fig. 4 and Fig. 5 plot the values

calculated using Airy linear and Wheeler theories respectively andFig. 6 shows the results from stream function theory, which gives the

Wave Period [s]

WaveHeight[m]

Breaking

Waves

Wave Period [s]

WaveHeight[m]

Breaking

Waves



4/8


highest predictions for the loads. For small wave heights of up toperhaps 2 m, the differences are generally low but the errors increase

rapidly with height, Fig. 7. These charts reflect a particular case andother situations will result in different values however the overallconclusions will remain similar, being more extreme in shallower

waters and less so as the water depths increase. For wheeler-theory, theoverturning moment can be less than 45% of the more accurate value

determined using stream function theory (and for linear theory, lessthan 75%).

T

6 8 10 12 14

5

10

4.5 107

4 1074 10

7

3.5 107

3 107

3 107

2.5 107

2 107

2 107

1.5 107

1.5 107

1 107

1 107

1 107

5 106

5 106

5 106

Fig. 6: Overturning Moment calculated using Stream Function

T

6 8 10 12 14

5

10 0

5

5

5

5

10

10

10

10

15

15

15

20

20

20

25

25

25

30

30

30

35

35

35

40

40

40

45

45

45

50

50

50

55

55

Fig. 7: Error (in %) in OTM Calculated Using Wheeler Method

Turning to shallower 6 m water depths, such as at Blyth, if the totalfatigue damage is estimated by applying the Blyth sea climate, use of

the Wheeler wave model results in a lifetime fatigue of 43% of themore accurate stream function value (linear theory returns 35%). Thecorresponding figures for the deeper water (21 m depth) case is 97%and 95% assuming the same wave distribution, confirming the

supposition that using a non-linear wave model is important for theshallower waters only. This error does not include other inaccuraciesdue to ignoring surface effects for example but it will not matter if thefatigue damage is dominated by the rotor loads.

The extreme and fatigue response stresses depend strongly on thedynamic behaviour of the wind turbine structure. The peaked surfaceelevation of the non-linear waves shown in Fig. 1 results in harmonic

loads at integer multiples of the wave frequency, a feature that is lesspronounced if present at all in linear waves. When harmonics of the

wave frequency coincide with the structural natural frequency,resonance of the structure results in amplification of the response,Fig. 8. (Note that numerical errors cause the deviation of the low

amplitude wave curve from unity). The presented amplification is validfor a infinite sequence of regular and identical periodic waves. For a

single incoming wave, the relation between the phase of the wave andthe initial structural motion determines the effective amplification, as isshown below in Fig. 11 and Fig. 12 from the measurements taken at theBlyth turbine.

0.5

0.75

1

1.25

1.5

1.75

2

2.25

6 6.5 7 7.5 8 8.5 9Wave Period [s]

DynamicAmplification

oftheOTM

Extreme

Waves

Low

Amplitude

Waves

Fig. 8: Dynamic Amplification of Non-Linear Waves LoadsFig. 8 shows that there is significant dynamic amplification at wave

periods of around three (6.4 s) and four (8.5 s) times the towersfirst natural period and that this is only apparent for the extreme wave(the undulations in the small wave curve are due to insufficient run

times in the simulations) and also that no cancellation can be identifiedat the mid-natural period point (i.e. 3 natural period) but instead

being present immediately prior to the multiple value (i.e. 2.9 and 3.9 natural period).

Examining the dynamic amplification in terms of kinematics, if theratio is an integer, the structure will be moving in phase with the wave

when each rising crest reaches the monopile and severe dynamicamplifications will be produced, Fig. 11, while if it is an integer plus ahalf, it is expected that the structure would be moving into the wave

when it impacts, Fig. 12. This second phenomenon is not reproducedusing this computer model, BLADED, it being primarily a windturbinemodel but with appropriate offshore features.

HYDRODYNAMIC LOADING EVALUATION OF

MEASUREMENTS

This chapter evaluates wave load theory against measurements taken atthe instrumented offshore wind-turbine at Blyth. The campaigndatasets used to illustrate within this paper was campaign reference158, recorded on 9 Nov 01 at 02:33, when the significant wave height

was 4.63 m, the tide level was 1.53 m and the windspeed was 13.92m/s. The turbine was off. The initial strain gauges calibrations were

performed using the method described inAppendix Aand this has sincebeen confirmed by external calibrations.

In all figures in this paper, the wave height is defined as the differencebetween the crest and the mean of the previous and following troughs.This provides a generally close match for the stream function wave,Fig. 9 being a particularly good example.

Wave Period [s]

W

aveHeight[m]

Breaking

Waves

Wave Period [s]

WaveHeight[m

]

Breaking

Waves



5/8


Individual Waves

0 1 2 3 4 5 6 7 84

2

0

2

4

Time [s]

Surf

aceElevation[m]

Wave Profile [m]Linear

Stream Function Theory Fig. 9: Wave Profile

0 1 2 3 4 5 6 7 82 10

6

1 106

0

1 106

2 106

3 106

4 106

5 106

Time [s]

BendingMoment[Nm]

Fig. 10: Pile Mudline Moment

A high wave selected from the dataset is examined in Fig. 9 (surface

elevation) and Fig. 10 (pile mudline bending moment). It can be seenthat for this example the stream function theory predicts the same crest

elevation but that, as would be expected, linear theory does not. This isalso reflected in the bending moment traces in Fig. 10, which show thatall theories underestimate the maximum bending moment but that thestream function is closest. It can also be seen that the peaks in the

loading of both the recorded wave and the stream function solutionoccur where the wave surface is steepest, at approximately 2 s and3 s respectively.

5 0 5 10 15 205 10

6

0

5 106

1 107

Time [s]

B

endingMoment[Nm]

Fig. 11: Dynamics Amplification

10 5 0 5 10 15 202 10

6

1 106

0

1 106

2 106

3 106

Time [s]

Ben

dingMoment[Nm]

Fig. 12: Dynamics Cancellation

Note that in Fig. 10 measured internal bending stresses are beingcompared against theoretical external wave loads, i.e. the dynamics are

not taken into account in the theoretical traces. Inclusion of dynamicsin the theoretical trace will change the profile (by adding highfrequency oscillations due to modal response) but would probably not

change the maximum value that much for this example; wheredynamics is of particular importance is when the structure is alreadyoscillating when the wave impacts onto it. This is clearly shown in

Fig. 11 and Fig. 12, which illustrate amplification and cancellationrespectively.

Campaign (30 minute sea state)

If a comparison is made between the measurements and theory for allindividual waves in the campaign (30 minutes), Fig. 13, it can be seen

that all theories underestimate the maximum bending moment, withlinear and stretched-linear theories being lower than stream function. Acorrect prediction would place the points on to the diagonal line. Thereis a large amount of scatter, due to several reasons (identified in

Fig. 13):region (a) ringing induced by the previous wave resulting in

an apparent underestimation by theory of the loads,region (b) dynamic amplification resulting in high measured

loads in comparison with theory, Fig. 11, (i.e.underestimation by theory)

region (c) dynamic cancellation resulting in low measuredloads in comparison with the theory, Fig. 12.

0 1 106

2 106

3 106

4 106

5 106

6 106

7 106

8 106

0

1 106

2 106

3 106

4 106 Identical Crest Elevations

Measured Maximum Bending Moment [MNm]CalculatedMaxBendingMoment[MNm]

Fig. 13: Maximum Bending Moment - Calculated verses Measured

DISCUSSION

This paper has focused on understanding and predicting the

hydrodynamic loads and hence the structural response of offshorewindturbine support-structures. The principal problem identified is thewaves in shallow water are less linear than those that the currently used

methods were developed for (the methods being developed by the

region (a)

region (c)

region (b)

rms (X&Y) Strain gaugeAiry Linear

Wheeler Stretching

Chakrabarti Stretching

Constant CrestExtrapolated

Stream Function

rms (X&Y) Strain gaugAiry LinearWheeler StretchingStream Function



6/8


offshore industry which focuses on deeper waters). This affects theanalysis of slender structures in the following way:

slender structures respond dynamically to the loads, howeverno design approach is currently able to include this structuralresponse together with stochastic non-linear waves of anappropriately high order,

A long term solution to both these dilemmas will be CFD however we

await further development of theory as well as necessary increases incomputer power, both of which should be available in perhaps adecades time.

Currently, the design process for slender offshore windturbine supportstructures takes two approaches: regular non-linear waves andstochastic linear seas. The problem with using regular non-linear wave

approach is that the structural motion at incidence of the wavedetermine the dynamic response, (either amplification or cancellation).

Evaluation of Current Methods

Regarding the stochastic or probabilistic approach, this appears to

underestimate the structural response since it does not model thehighest waves accurately (exclude harmonics) and does not takeaccount of impact loading Henderson, A.R. (2002). These excluded

aspects may cause damage disproportionate to their size because theypotentially act near to the structures natural frequencies resulting inboth extreme but also fatigue loading being underestimated. Togetherwith other aspects identified within this paper, this may lead to thehydrodynamic loading being relatively more important in the overall

design of the support structure than expected. Fig. 14 and Fig. 15 showthe estimated fatigue damage derived from a selection of measurements

campaigns (both 4 and 30 minute), plotted against significant waveheight (left) and wind speed (right). The turbine status (off, switchingor on) is identified by the marking used in the charts. From the leftchart, it can be seen that when the turbine is off, fatigue damagecorrelates very well with significant wave height, while from the right

chart, it can be seen that when the turbine is on, fatigue damagecorrelates reasonable well with windspeed. Apart from the worst

fatigue occurring when the turbine is switched on and off, both wind

and wave loads appear to be important.

0 1 2 3 4 5 61 10

3

0.01

0.1

1

10

100

Wave Height [m]

FatigueDamage[arbitraryunits]

Fig. 14: Campaign Tower Mudline Fatigue Damage vs. Wave Height

0 5 10 15 201 10

3

0.01

0.1

1

10

100

Wind Speed [m/s]

FatigueDamage

[arbitraryunits]

Fig. 15: Campaign Tower Mudline Fatigue Damage vs. Wind Speed

Examining the relationship between extreme loads recorded withineach campaign against the wave height, Fig. 16, and windspeed,Fig. 17, again shows a stronger correlation with the waves when theturbine is off and the wind when the turbine is on.

0 1 2 3 4 5 60

5

10

15

20

Wave Height [m]

ExtremeMudlineOTM

[MNm]

Fig. 16: Campaign Tower Mudline Extreme Loads vs. Wave Height

0 5 10 15 200

5

10

15

20

Wind Speed [m/s]

ExtremeMudlineOTM

[MNm]

Fig. 17: Campaign Tower Mudline Extreme Loads vs. Wind Speed

These four figures also show that it is of some importance to minimizethe number of times the windturbine is switched on and off in order to

maximize fatigue life.Meanwhile, the main conclusions regarding fatigue loading at themonopile mudline from Camp, T.R. (2003) were that (i) the fatiguelifetime damage was underestimated by 50% and overestimated by 8%

along the two axes, the first being closer to the prevailing wavedirection and that (ii) according to BLADED, wave loads only added

turbine ON

turbine SWITCHINGturbine OFF

turbine ON


turbine ONturbine SWITCHINGturbine OFF

turbine ON




7/8


14% to the fatigue damage however potential limitations to BLADEDhave been noted above (note that these figures are in terms of fatigue

lifetime damage and not fatigue equivalent loads).

Alternative Approach

A potential solution, the full investigation of which was beyond the

scope of this present work, would be to use linear stochastic models todetermine a preliminary estimation of the motion response distributionand to apply the initial conditions to regular non-linear wave analysis.An approach that could be taken to implement this might be:

use existing linear-modified stochastic models to determinean initial estimate of the probability distribution of the

response of the structure within a particular sea state

determine that probability distribution of waves within thatparticular sea state

apply a range of regular non-linear waves to a matrix ofinitial conditions and determine the structural response

apply probability distribution of the waves and the structuresinitial conditions to calculate the actual response to the seastate

This is based on the following assumptions:

that damping of the first mode is low (around 5% predictedby BLADED Camp, T.R. (2003); also see Fig. 11 andFig. 12), so the motion response probability distribution

derived from the complete sea state is assumed to apply at thepoint of wave impact

that the wave periods are significantly longer than the firstmode natural period, hence there being limited correlation ofthe motion of the structure with the moment that a wave

makes impact

that there is limited correlation between the size ofconsecutive waves (this is incorrect; it might be possible tomake modifications )

that the wave direction and structural response is uni-directional ( an additional structural response dimension

would make the analysis unwieldy).The main difficulties encountered were that BLADED does not allowthe initial conditions to be specified and that the last assumption is not

correct, i.e. see Fig. 18: it is difficult to define the initial conditions in asimple manner.

CONCLUSIONS

In this paper, an attempt has been made to equate the measured waveloads with those calculated from the measured waves at a windturbinein an exposed North Sea location. The wave climate is severe and with

the shallowness of the water, breaking waves are likely to be a frequentoccurrence during storms. (It should be noted that these conditions are

not necessarily going to be typical for future offshore windfarms).By comparing the loads predicted by linear and non-linear wave

theories, it can be seen that the use of non-linear theory of a high orderis recommended. Currently, only regular non-linear waves can be

modelled using widely available and computer-efficient theories. Onthe other hand, a perfunctory examination of the strain gauge output

indicates that structural dynamics are important, hence the seas shouldbe modelled in a stochastic manner. Hence the lack of availability of awave model that is both non-linear and stochastic is an importanthindrance in the case of Blyth.Bearing this in mind, it was found that in storm seas, the measured

loads where on average higher than those predicted, hence the fatiguedamage would be higher than predicted by stochastic linear or regularnon-linear analysis.

It should be noted that the general case for support structures in such

shallow water is that turbine loads dominate over wave loads for boththe fatigue and extreme cases. However, this conclusion is based ontheoretical analyses undertaken using the wave theories shown to beunconservative. It seems likely that in cases such as Blyth, wave loads

are of greater importance than is generally assumed, possibly up to thepoint of being a similar importance, though wind loads are still likely to

at least slightly larger.To summarise, the key conclusion of this work is that the measuredwave loads at Blyth are higher than any of the available theories

predict. It is expected that this be partly due to breaking wave impactloads but the fact that waves in the non-breaking seas will be of an

extreme form, which cannot be adequately modelled, must also beimportant.

RECOMMENDATIONS

Future work on hydrodynamic loading of offshore windturbines shouldfocus on achieving the following in an efficient and effective manner:

To examine how stochastic non-linear wave loading of ahigher order than at present can be applied to slender supportstructures

In the meantime, it is recommended that analyses involving acombination of linear stochastic seas and non-linear regular waves be

performed.

ACKNOWLEDGEMENTS

The work reported here forms part of the European Commission

supported research project entitled 'Design Methods for Offshore WindTurbines at Exposed Sites (OWTES) and is being undertaken byAMEC Border Wind, Delft University of Technology, GermanischerLloyd WindEnergie, PowerGen Renewables Developments and Vestas

Wind Systems under the co-ordination of Garrad Hassan and Partners.Support from the owners of the windturbine, Shell, PowerGen, Nuonand AMEC, is also acknowledged for allowing the work to proceed on

their windturbine.

APPENDIX A - CALIBRATION OF STRAIN GAUGES

This appendix presents a description of how preliminary calibrations ofthe strain gauges in the tower and pile were made against the periodic

first mode oscillations during short periods (of a few seconds) ofrelative calm in between the waves that were generating the motion

response. This method depends on the fact that extreme non-linearwaves have a particular shape: they consist of short and sharp peaks in

between long regions of relatively smooth water where the waveloadings are minimal, see Fig. 1 and Fig. 2.Since the external calibrations of the strain gauges were undertaken at alate stage in the project, in order to start analysing the data earlier on

and hence gain maximum benefit from the measurement programme, itwas necessary to perform a preliminary calibration. Examining the

data for the nacelle accelerometers and the strain gauges, it could beseen that at certain times, when the turbine was turned off, there was

good correlation between the signals, Fig. 18.



8/8


0.3 0.2 0.1 0 0.10.2

0.1

0

0.1

0.2

X axis

Yaxis

Acceleration at Nacelle [m/s2]

start of cycleStrain Gauge Pile MWL [mV]

Fig. 18: Correlation of Nacelle Accelerations and Strain GaugeReadings

When all such cycles within a 30 minute campaign are evaluated, andthe most appropriate selected (i.e. with little higher mode noise) a

calibration of the strain gauge voltage reading against the nacelleacceleration can be made, Fig. 19. Making an assumption for the modeshape allows an estimation of the calibration against the moments at the

strain gauge to be made in both X and Y axes, Fig. 20. Thesecalibration factors were in good agreement with the externallycalibrated values: in particular, the differences for the mudline straingauges were significantly less than the accepted uncertainty of the

external calibration process itself (a few percent), see Table 2.

RecValue

0 2 104

4 104

6 104

8 104

0.001 0.0012 0.0014 0.0016

0

0.1

0.2

0.3

Estimated Strain Gauge Calibration

Probability

Fig. 19: Preliminary Calibration of the Strain Gauges m/s2per V

1 2 3 4 5 61 10

8

1 109

1 1010

1 1011 Recommended Strain Gauge Calibratation

Gauge Location

Str

ainGaugeCalibration[NmvsV]

Tower

Top

Tower

Base

Pile

MWL

Pile

dep 1

Pile

dep 2

Mud-

line

X

Y

Fig. 20: Calibration Factors(Moment per Volt output) for Tower andPile Strain Gauges, also showing the + standard deviation

Table 2: Evaluation of Strain Gauge Calibration Method

Calibration Factor [Nm per V]

AxisCalculated Henderson, A.R.

(2002)Externally Measured

X 1.27 1010 1.251 1010

Y 1.21 1010 1.208 1010

References

Airy, Sir G. B. (1845), Tides and waves, Encycl. Metrop., Art 192,.

Barltrop, N.D.P. (1991) & Adams A.J., Dynamics of Fixed OffshoreStructures, Butterworth-Heinemann

Boussinesq, J (1872), Thorie des ondes et des ramous qui se propagentle long dn canal rectangulaire horizontal, en cmmuniquant au liquidecontenu dans ce canal des vitesses sensiblement pareilles de la surfaceau fond, J. Math. Pures Appl., Vol. 17, pp 55-108

Camp, T.R. (2003) et al, Design Methods For Offshore Wind TurbinesAt Exposed Sites (OWTES) Final Report 2317/BR/22, Garrad Hassan,

November 2003

Chakrabarti, S.K. (1987), Hydrodynamics of Offshore Structures, WIT

Press, ISBN 0-905451-66-X, .Cheng, P.W. (2002), A reliability based design methodology for extremeresponses of offshore wind turbines, PhD Thesis, Delft University of

TechnologyDean, R.G., (1965), Stream Function Representation of Nonlinear Ocean

Waves, J. Geophys. Res., Vol 70, No. 18, pp 4561-4572DNV (2000) (Det Norske Veritas), Rules for the classification of fixed

offshore installationsHenderson, A.R. (2002) & Cheng, P.W., Wave loads on slender offshore

structures: Comparison of Theory and Measurement, German Wind

Energy Conference DEWEK, Wilhelmshaven, GermanyHenderson, A.R. (2003), et al., Hydrodynamic Loading on Offshore

Wind Turbines, Report SW-0218x, Section Wind Energy, TechnicalUniversity of Delft, February 2003.

Madsen, P.A., (2000), Bingham, H., Liu, H., The ultimate Boussinesq

formulation for highly dispersive and highly nonlinear water waves,Coastal Engineering Conference, p 176-189,

MacCamy, R.C. (1954) and Fuchs, R.A., Wave Forces on Piles: a

Diffraction Theory, Tech. Memo No. 69, US Army Corps ofEngineers, Beach Erosion Board,

Neumann, G. (1963) and Pierson W. J., Principles of physical

oceanography; Prentice-Hall..Srensen, H. C. (2000) et al, Middelgrunden 40 MW offshore wind farm,

a prestudy for the Danish offshore 750 mw wind program, ISOPEConference.

Wheeler, J.D. (1969), Method for calculating forces produced byirregular waves, First Offshore Technology Conference, Houston,

Texas, OTC 1006


henderson isope2004 loading

Documents