You are on page 1of 10

Optimization of a Monopile Support Structure using Offshore-Specific

Wind Turbine Controls


1*

Fischer, T.1*, Rainey, P.2, Bossanyi2 and Khn, M.1


Endowed Chair of Wind Energy, Universitt Stuttgart, Germany, tim.fischer@ifb.uni-stuttgart.de
2
Garrad Hassan & Partner Ltd., Bristol, United Kingdom

Summary
For current offshore wind farms, monopiles are by far the most popular support structure type.
However, for deeper water and/or larger turbines, the fatigue loading is becoming critical and
the monopile dimensions are exceeding the economical feasibility. Thus, further effort should be
made to lower the loading on monopiles and bring them back to be a cost-effective solution.
Since the industry aims to use standardized rotor-nacelle-assembly (RNA) designs, the goal of
this paper is to show the effectiveness of using the RNA as adjustable controller device for site
specific offshore support structure configurations by using both, dynamic control and operational
control devices. In general, the study showed that offshore-specific controls can be effective in
reducing hydrodynamic-induced loading on monopile support structures. Here the amount of
mitigation is very much depended on the importance of hydrodynamic loading with respect to
the overall fatigue.
1 Introduction
Since support structures are one of the main cost drivers offshore and since offshore other rules
apply with regards to design and site conditions, different methodologies have to be developed
to mitigate the loads on the support structure and therefore to reduce the associated component
costs. Detailed onshore studies and field tests have already shown that advanced control
algorithms like individual pitch control (IPC) and tower feedback or operational control strategies
like a softer cut-out can be an effective way to reduce fatigue and extreme loading on a wind
turbine. Still, the overall cost-effectiveness for large offshore wind farms has not been studied in
detail yet. In the scope of the integrated EU project UpWind, studies of new controller concepts
are integrated within the design process for offshore support structures in order to stretch the
applicability of monopiles to larger water depths or simply to reduce the structural weight and
associated costs.
2 Background
Reduction of the aerodynamic and hydrodynamic loads and of the associated dynamic
response is an obvious way to achieve optimized design, alongside optimised manufacturing
and installation logistics. This is achieved by integrating the design of the RNA and support
structure in the design process of the support structure. Hence, the RNA is considered as an
active component to mitigate the loads on the support structure. For this load mitigation of the
support structure, different concepts are possible and can be distinguished at three different
levels. These three levels are: considerations on the design level, implementations on the
operational control level or finally implementations on the dynamic control level. Figure 1
illustrates some examples for these different levels of load mitigation [1]. The goal is to identify
the best selection of different options for given site and design conditions in order to finally
obtain an optimised offshore wind turbine design.
For monopile support structures, as studied in this paper, the overturning bending moment at
mudline is the most critical one in cases of fatigue. Here the major impact is of course in the

fore-aft motion. In cases of strong wind- and wave-misalignment the sidewise loading becomes
more important and can even be design-driving. From a controller point of view, mitigation of the
fore-aft overturning moment should be possible both in theory and practice. Different studies
[2] have already shown the potential of damping the fore-aft motion of the tower top with simple
collective pitch controls, but mainly for onshore applications. This paper will therefore
concentrate on the damping of the fore-aft mode in an offshore application. As the emphasis is
on the fore-aft motion only, in all later discussed studies and simulations, the wind and wave
direction is assumed to be co-directional. Furthermore the focus will only be on fatigue loading,
as for the given site the design of the monopile is mainly driven by fatigue instead of extreme
loads.

Extreme Event
Control

Feedback
Control
Individual Pitch
Control

Active Mass
Damper

DYNAMIC CONTROL
LEVEL

Passive Mass
Damper

2-Bladed
Turbine
Truss Tower

vc

ut-

at

vr

ed

vc

3
1

ut-

1P Compliant
Design

OPERATIONAL CONTROL
LEVEL
Cut-out Control

DESIGN LEVEL

Figure 1: Levels and possible implementations of load mitigation


Two of the concepts illustrated in Figure 1were priori identified to reduce the fore-aft loading on
monopile structures namely the operational control concept of a softer cut-out and a tower
feedback controller. However, other options would have been possible as well, such as tower
damper devices, but are not considered here.
Of course, the use of additional control solutions for load mitigation imposes a number of
general requirements to other components and the full system, which have to be fulfilled. Those
are:
Potential additional loading of other components of the RNA especially pitch drives,
blades and sensitive drive train components like the gear box should be minimised.

Extra controller action will inherently reduce the energy yield of the offshore wind
turbine by operation outside of the aerodynamic optimum and by direct energy
consumption of the actuators. As a rule of thumb at least 4 5% cost reduction in the
total support structure costs (material, manufacturing and installation) is required for
compensation of each percent loss in energy yield, assuming a 20 25% proportion of
support structure related investment of the cost of the energy.

New control concepts require innovative control algorithms as well as robust load or
acceleration feedback sensors for structural response and possibly also for
environmental conditions like wind and wave parameters.

3 Reference design and site


The studies in this report are based on
a location in the Dutch North Sea,
close to the already built offshore wind
farm Egmond aan Zee (OWEZ). The
mean water depth of the site is 21.4m
and the average wind speed 9.6m/s at
hub
height.
A
more
detailed
description of the site is given in the
design basis [3].
Figure 2, shows the used wind and
wave data is used. The Weibull
parameters of the site are A=10.9m/s
and k=2.05. The turbulence is
according to current guidelines [4]. For
the wave data, a lumping is performed
Figure 2: Wind speed distribution, turbulence intensity and
to reduce the amount of load cases.
wave conditions of the study
Thus, for each wind speed bin a
damage-equivalent sea state is
derived as shown in Figure 2. For the soil conditions a
Blade Tip
142.9 m + MSL
distribution with dense sand layers is taken, which
results in stiff conditions. All data is then used in load
simulations for the fatigue load cases DLC1.2 and
DLC6.4 according to IEC 61400-3 [4]. 10min
simulations and an averaging over three different
Hub Height
79.9 m + MSL
seeds for wind and waves are employed.
In the scope of the concepts studied here, the UpWind
5MW reference turbine is used, which is a 3-bladed,
variable speed and pitch controlled design. The turbine
is equipped with a standard power production,
collective pitch controller [5] as explained in more
detail in the following section. Based on the sites wave
conditions, the platform level is found at 13.5m as
13.5 m + MSL
Platform Level
shown in Figure 3. By using a standard tower of 64m
Sea Level
0.0 m + MSL
(with a top diameter of 4.3m and a bottom diameter of
-5.5 m + MSL
TP Bottom
6.0m), the support structure design results finally in a
hub height of 79.9m above MSL. The vertical hub
Seabed Level
-21.4 m + MSL
offset is 2.4m.
The turbine is supported by a monopile, which has an
Pile Toe Level
outer diameter of 5.5m and a constant wall thickness
-46.4 m + MSL
of 60mm, which leads to a total mass of 220t. For the
Figure 3: Schematic dimensions of
the reference design for the shallow
water location

assumed stiff soil conditions a penetration depth of


25m provided the necessary stability. The added

diameter and mass of the transition piece (4t) as well as the reduced fatigue strength at the
tower base flange are taken into account. The given turbine and monopile configuration results
in a rather stiff support structure with a fundamental eigenfrequency of 0.32Hz. This value is
rather close to the lower end of the 3P excitation range of 0.35 Hz. A softer structure would
however attract even higher dynamic wave loads and would not achieve the required fatigue
lifetime of 20 years.
4 Controller development
As discussed in section 2, different concepts are available to reduce the fore-aft loading on
monopile structures. In the following the baseline power controller and two chosen controller
concepts will be identified and explained.
4.1

Baseline power controller

The turbine used in this study uses collective pitch to feather control above rated wind speed,
and has a variable generator speed. The torque controller is capable of achieving any
demanded torque (within limits) at the air gap with a short delay. The baseline controller takes
measured generator speed as the controller input, and returns a demanded generator torque
and a collective pitch angle demand.
During low wind speed, the generator torque control follows a quadratic torque-speed curve.
This ensures that the rotor speed is optimal for energy capture. In moderate wind speed, when
the rated rotor speed is reached, the generator torque demand is derived from the measured
generator speed error using a proportional plus integral controller. When rated wind speed is
reached, and the blades are pitched away from fine pitch angle, the torque is varied in inverse
proportion to measured generator speed; this minimises power fluctuations. In addition there is
a drive train damping algorithm which adds small amplitude variation in torque demand which
increases the damping of the drive train eigenmodes.
The pitch controller is also a proportional plus integral (PI) controller on measured generator
speed error. The proportional and integral gains are scheduled according to the pitch angle, as
the aerodynamic torque is much more sensitive to pitch angle changes at higher pitch angle
than around fine pitch.
The pitch angle is held at fine pitch while the generator torque is below rated to keep the pitch
and torque control loops decoupled.
4.2

Application of a tower-feedback controller

In this case the controller uses measured nacelle acceleration as an additional input above
rated wind speed. It works along side the pitch controller by calculating an additional pitch rate
demand. The pitch rate is derived from passing the acceleration signal through a lead
compensator to achieve optimal damping of the 1st tower fore-aft mode. The stability margins of
the original pitch-speed control loop are eroded by the addition of the nacelle acceleration
feedback. Therefore the gains on the pitch-speed PI are reduced slightly to allow the nacelle
acceleration feedback to operate. This has the negative effect of causing greater generator
speed fluctuations which could lead to a more expensive drive train.
The dynamics of pitch-speed control loop vary considerably across the wind speed range. As
the nacelle-feedback controller interacts strongly with the pitch-speed loop, it is important to
ensure that the lead compensator is working optimally at all wind speeds. This has been
achieved by tuning it separately at several wind speeds and using a schedule based on pitch
angle to vary the compensator parameters appropriately.

The exact implementation is shown in block diagram form in Figure 4. The controller is
implemented in discrete time, and represented in the diagram using the backward-shift
operator. Many compensator parameters (Gain, A and B ) were investigated at each wind
speed, however it was found that good performance could be achieved by using just a single
lookup table rather than one for each parameter. This approach simplifies the task of tuning and
implementing the controller.

a0 = T + A
a1 = T A
b0 = T + B

b1 = T B
c0 = T
c1 = T

T = PitchStep
z 1 = Backward shift operator

a0 + a1 z 1
b0 + b1 z 1

c0 + c1 z 1
b0 + b1 z 1
Figure 4: Nacelle-feedback block diagram

4.3

Application of a soft cut-out controller

When the wind turbine is operating in power production the tower fore-aft mode is damped
aerodynamically. The motivation for introducing a soft cut-out is to reduce the fatigue loading of
the support structure by operating the turbine up to higher than normal wind speeds. It is
successful if the benefit from added tower fore-aft damping of the wave loading compensates
for the added turbulence loading. The demanded generator speed is reduced (to 2/3 or normal)
for the soft cut-out regime. This reduces the aerodynamic loads which otherwise can lead to
design driving ultimate loads. This has the additional benefit that the generator speed does not
need to be controlled as tightly as there is a much increased margin to the maximum allowed
generator speed. For the soft cut-out regime the controller has been tuned to give maximum
tower damping (using the nacelle acceleration feedback). This is achieved by having much
lower gains in the pitch-speed loop, than would be acceptable if the turbine was operating at
normal speed.
In the following, both concepts are used in an integrated manner. This means that over the
normal power production range (4-25m/s) the tower feedback controller is activated with the
goal of adding additional damping to the support structure mode while keeping the power as
stable as possible. Besides, the soft cut-out is used for the extended power range (26-30m/s)
together with tower feedback control with the goal of maximum damping to the structure mode.

5 Results
5.1

Results in load mitigation

moment at mudline [MNm]

60
50
40
Ref_Mx
30

TFC_Mx
Ref_My

20

TFC_My
10

moment at blade root [MNm]

The new features described above have been tested in dynamic simulations using GH Bladed
with three-component turbulent wind and irregular wave trains as input, both in co-directional
manner. Figure 5 shows the bending moment of the support structure at mudline and the
moment at the blade root. In both cases, the loads are illustrated as damage equivalent loads
(DEL) referring to a lifetime of 20 years and an equivalent frequency of 1Hz (N=6.31E8). The
distributions shown in Figure 5 are non-lifetime weighted, which means that each wind speed
bin DEL assumes a duration of 20years for the corresponding bin. By this, the relative results
can be analysed in more detail. In contrast, Figure 6 plots the same loads weighted with their
actual probability during 20 years.
8
7
6
5
Ref_Mflap

TFC_Mflap

Ref_Medge

TFC_Medge

1
0

0
2

8 10 12 14 16 18 20 22 24 26 28 30

8 10 12 14 16 18 20 22 24 26 28 30

wind speed [m/s]

wind speed [m/s]

Figure 5: Non-lifetime weighted DELs (with N=6E+8) for the support structure moments (m=4)
at mudline (left) and the blade moments (m=10) at blade root (right) as comparison of the
reference and the TFC-controlled case

moment at mudline [MNm]

4,0
3,5
3,0
2,5

Ref_Mx

2,0

TFC_Mx
Ref_My

1,5

TFC_My

1,0

moment at blade root [MNm]

For the support structure it can be seen that the TFC reduces the fore-aft loading (here
expressed as overturning moment, My) well and for the normal power production range (425m/s). The moment in the sidewise direction, Mx, is also slightly reduced. This is due to the
coupling in movement of the tubular structure in longitudinal and lateral direction, which
generally moves on an oval path. If the main contributor to the movement, the fore-aft direction,
is damped, this could also imply a damping to the sidewise direction.
1,7
1,6
1,5
1,4
1,3

Ref_Mflap

1,2

TFC_Mflap

1,1

Ref_Medge

1,0

TFC_Medge

0,9
0,8
0,7

0,5

2
2

8 10 12 14 16 18 20 22 24 26 28 30

8 10 12 14 16 18 20 22 24 26 28 30

wind speed [m/s]

wind speed [m/s]

Figure 6: Lifetime weighted DELs (with N=6E+8) for the support structure moments (m=4) at
mudline and the blade moments (m=10) at blade root as comparison of the reference and the
TFC-controlled case
The results for the blade loads, here expressed as flapwise and edgewise moments at the blade
root, are almost unchanged. This is due to the fact that the controller is smoothly integrated into
the overall power controller, which results in quite balanced amplification gains for the control

goals of maximising power and mitigating loads. Furthermore, the TFC is designed to damp the
first support structure eigenfrequency, where the blade loads are mainly affected by rotational
depended contributions, such as 1P or 3P.
Beside the activation of the TFC for the normal power production range (here the wind speed
bins 4-25m/s), the soft cut-out controller described above can be activated as well, which can
additionally damp the fore-aft loads at the support structure. For the given study the extended
cut-out decreases the fore-aft moments at mudline for the support structure significantly. This
result illustrates again the importance of providing aerodynamic damping to the system, as it is
an optimal way to reduce hydrodynamic induced vibrations. Of course, due to the extended
power range the blade loads are increased noticeable, as they are turned out of their idling
position into the wind and they are rotating, which adds gravity and gust slicing 1P loads. And
also the sidewise support structure moment, Mx, increases due to its connection to the
edgewise blade loads. However, the blade and the sidewise support structure loads are kept in
an acceptable level, especially if their load level is compared to the cases of the former power
production range (wind bins of 4-25m/s).

1st support structure


mode (0.32Hz)

Figure 7: Spectral density of the overturning support structure moment at mudline (left) and the
pitch rate (right) as comparison of the reference and the TFC-controlled case
The results are also compared on a fatigue-lifetime level. First of all the DEL for each wind
speed bin is connected to the corresponding probability of occurrence, as shown in Figure 6 for
the support structure and blade loads. Again, the DEL values are referring to a lifetime of 20
years and an equivalent frequency of 1Hz. In this representation, the large changes in loading
for the soft cut-out option are diminished out by the low occurrences of those high wind speeds.
Still, by looking at the figures a good mitigation in the dominant fore-aft loading can be seen,
especially around rated wind speed and for the extended cut-out. Figure 7 summarises the
mitigation effect for the fore-aft moment in the frequency domain for all simulated wind speeds
as comparison between the reference case and the case with an applied TFC. It can be seen
that the controller reduces well the first support structure eigenfrequency peak at 0.32Hz, while
the other major loading frequencies, such as the rotational depended 1P (0.20 Hz) and 3P and
the other support structure frequencies, such as the 2nd mode (1.6 Hz), are left unaffected. This
effect is expected, as the TFC is tuned on the first support structure eigenfrequency and
therefore it reduces mainly loads at this frequency and maybe some other connected fixed
frame loads. If a reduction in rotational depended loads, such as 1P or 3P, shall be achieved,
other control options like individual pitch control are giving good results [6].

Table 1: Comparison of results between the reference and the TFC-controlled case
Change in energy output
and power fluctuations
from reference case

Loads as DEL change from reference case


[N=6E+8, m=4 (support structure) and m=10 (blades)]
Support Structure at
mudline ( -21.4 m )
My
Mx

Blades
Mflap

Medge

AEP

Pstd

Reference case

4.6 MNm

18.8 MNm

4.7 MNm

6.0 MNm

22.6 GWh

82 kW

TFC 4-24m/s

-2.8 %

-3.4 %

+1.0 %

0.0 %

-0.3 %

+0.4 %

TFC 4-30m/s

+45.8 %

-8.8 %

+1.5 %

+0.1 %

+0.3 %

+0.4 %

Table 1 summarises the results as changes in lifetime weighted equivalent fatigue loads for the
support structure and the blades. It shows that the TFC reduces the dominant support structure
fore-aft moment up to 9% while keeping the blade loads almost unchanged. Only the effect of
extended cut-out increases the flapwise moment slightly and the sidewise support structure
moment significantly from a relative point of view, but not from an absolute in comparison to the
dominant fore-aft loading.
In addition to the analysis of the loads of different components of the offshore wind turbine,
other system effects have been evaluated in order to judge the proposed controller concepts.
Table 1 shows the effect on the power production. The power output is slightly reduced if the
TFC is activated for the normal power production range (wind speed bins 4-24m/s) and the
power fluctuations are slightly higher. However, if the power production range is extended (here
up to 30m/s), the penalties in power output are cancelled and even a slight increase in annual
energy production (AEP) is possible. The only mentionable influence of the TFC concept
proposed here is the effect on the pitch system. Here, the influence is measured by the
standard deviation of the pitch rate, which gives an indication of the increased pitch actuator
duty. On average, the pitch rate is in increased by approximately 25% for the rated power
cases. For the region before rated power, the pitch rate is of course increased, as before no
pitch action is given. Figure 7 illustrates the increased pitch rate in terms of a spectral density.
5.2

Results in structural optimization

After the performance and analysis of the new control


concepts, the achieved load reductions are used to redesign the given monopile. The optimization is based on
depth for max
moment using
a simplified fatigue analysis according to [7]. A simplified
linear approach
fatigue analysis is generally applied in cases of safety
margin considerations. Here the safety margin shall be
the basis of comparing two different monopile concepts
depth for max
with each other. The analysis is based on damage
moment using
equivalent loads (DEL) according to the Palmgren/Miner
non-linear
approach
rule. For the following monopile stress analysis, an S/N
curve slope parameter of m=4 (for welded steel) is
assumed.
As a first step, the section with the highest loading
has to be determined. Generally this is below mudline
Figure 8: Determination of the depth for
and very much dependant on the soil stiffness. As the maximum bending moment in the soil
using a linear and non-linear approach

shown in Figure 8, the maximum loading, here illustrated for a case under ultimate loading, is
found -10m below mudline using LPILE, a software to simulate piles under lateral loadings
and for non-linear soil responses. To estimate the maximal loading in a linear, simplified
manner, an equivalent soil depth is derived by applying a linear curve fit from mudline
downwards. In Figure 8 it can be seen that at a soil depth position of -5m an equivalent maximal
loading is obtained. Based on this, the maximal overturning moment can be found by adding the
product of damage equivalent shear force at mudline with the lever of 5m to the damage
equivalent overturning moment at mudline. The damage equivalent moments and shear forces
at mudline are generated with simulations using GH Bladed and for a number of stress cycles of
N=2E6.
Table 2: Comparison of results between the original and optimized monopile structural design
Original monopile

Optimized monopile

Baseline
power
controller

TFC and extended cutout controller

TFC and extended cutout controller

Pile outer diameter and wall


thickness

5.5 m and 60 mm

5.5 m and 60 mm

5.5 m and 57 mm

Total pile mass

220 t

220 t

205 t

st

1 eigenfrequency

0.323 Hz

0.323 Hz

0.317 Hz

DEL of overturning moment at


mudline N=6.3E8

95.3 MNm

88.6 MNm

89.0 MNm

Damage equivalent stress range


at maximum loaded cross section

69.1 N/mm

67.6 N/mm

67.9 N/mm

Fatigue strength utilisation

86 %

85 %

85 %

In Table 2, the results of the fatigue analysis are shown. Several monopile optimization
iterations have determined that the wall thickness of the pile can be reduced by 3mm over the
full length, which leads to a 7% lighter structure. This could be achieved through the application
of the new control concepts, which reduce the damage equivalent bending moment, here at -5m
below mudline. Further reductions in wall thickness are not possible, as the support structure is
then becoming too soft. A final fatigue safety margin check shows that the new monopile design
satisfies the same utilization value as before, by what the new monopile design together with
the new controller concept is proven to be successful from a structural optimization point of
view. To assess the material savings in a better way, a cost factor of 2/kg manufactured
monopile is assumed, which results in cost savings of 30k per monopile structure. For an large
offshore wind farm with 80 units, as it is commonly planned in the European North Sea, this
would save costs of about 2.5M.
6 Conclusion and outlook
The mitigation of aerodynamic and hydrodynamic loads is essential for future developments of
offshore wind turbines. In this report the prospects of different controller concepts are briefly
discussed and two concepts are investigated in a detailed design study.
The study performed here considers a standard 5MW turbine design on a monopile support
structure in about 21m water depth. The focus is on the reduction of the dominant fore-aft
motion of the support structure. Thus, wind and waves are considered co-directional. The
implemented load mitigation concepts leads to valuable reductions in loading, which could be
transferred to an optimized and therefore more cost-effective structural design. Besides, other

system loadings and characteristics could be kept in an acceptable range; hence a clear tradeoff can be seen for the given study.
In general, the study showed that offshore-specific controls can be effective in reducing
hydrodynamic-induced loading on monopile support structures. Here the amount of mitigation is
very much depended on the importance of hydrodynamic loading with respect to the overall
fatigue. However, other loading effects could be additionally studied and can influence the
optimization process of monopiles. The misalignment between wind and waves can have a
significant impact to the overall fatigue. Furthermore low availabilities of the offshore wind
turbine can cause high loading to the support structure, as the important effect of aerodynamic
damping is not available. Here, other concepts can be a solution for mitigation loadings. Some
of them are proposed in Section 2 of this paper, such as tower mass dampers or active control
of sidewise motions via pitch or generator torque, and will be studied in the further scope of the
European Integrated Research Project UpWind.
Acknowledgements
The presented work was funded by the Commission of the European Communities, Research
Directorate-General within the scope of the Integrated Project UpWind Integrated Wind
Turbine Design (Project No. 019945 (SES6).

References
[1]
[2]
[3]
[4]
[5]
[6]
[7]

Fischer, T. et al: Report on design integration studies on bottom-mounted support


structures, UpWind Report 4.1.3, Endowed Chair of Wind Energy, Stuttgart, 2009
Stdler, M.: Controls for Load Reduction, DEWEK, Bremen, 2008
Fischer, T, et al.: Design Basis - UpWind shallow water site, Endowed Chair of Wind
Energy, Stuttgart, 2009
IEC 61400-3, Wind turbines Part 3: Design requirements for offshore wind turbines, 1st
edition, CDV 2007
Bossanyi, E., A state-of-the-art-controller for the UpWind reference wind turbine,
EWEC, Marseille, 2009
Bossanyi, E.: Individual blade pitch control for load reduction, Wind Energy, Volume 8
Issue 4, Pages 481 - 48, 2005
Germanischer Lloyd: Rules and regulations; Guideline for the certification of offshore
wind turbines, Hamburg, 2005

You might also like