Professional Documents
Culture Documents
b damping matrix [kNs/m, kNms/rad] The linear frequency domain approach does not allow
B matrix of retardation functions [kN/m, kNm/rad] taking variations in time into account nor non-linear motion
c, C hydrostatic restoring forces matrix [kN/m, kNm/rad] response effects. In [3] a method is described to make (1)
F external force in the k-th mode [kN] suitable for analysis in the time domain by a normalization of
t time [s] the potential Φ. This method is described in [4]. Taking the
Cummins-equation into account (1) and following the approach
k,j hydrodynamic response in the k-mode
of [3] results in:
due to motion in the j-mode [-]
ρw water density [kg/m3] 6 t
study:
130.0 m
127.9 m
Motions
9.40 m
Fxyz TURBINE
Mxyz TURBINE Axyz LOW
X
Axyz MID
Axyz MID 4.40 m
1.50 m
200
• Surge is a forward motion (along the x-axis).
•
15
195
14
Sway is a lateral motion to portside (along the x-
190
13
axis).
185 222.5 m 445.0 m
12
• Heave is a upward motion (along the z-axis).
• Roll is a rotation around the x-axis.
180
11
385.4 m
10
170
Waves 180°
165
Current 180°
9
• Yaw is a rotation around the z-axis.
160
Wind 180°
8
The same definition is used for the measured motions and
7
the simulated motions. Figure 5 shows how the motions and
385.4 m
155
6
150
5
rotations are defined.
145
4
140
3
135
2
130
1
125
U T S R P O N M L K I H G F E D C B A
5°
22
120
es
av
W
115
113
112 110 105 100 95 90 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 10 5
Designation
Symbol Unit
Roll radius of gyration in air kxx m 53.54 Figure 5. Definition of the motions of the Spar
31.5
A. Wave loads
Natural pitch period (moored) Tθ s
The first order (and second order low frequency)
Natural heave period (moored) Tz s 31.0
hydrodynamics are calculated in the frequency domain, using a
Mooring connection height KFz m 50.0 three-dimensional source distribution technique. The mean
wetted part of the spar is modeled by a large number of panels.
Mooring connection radius r m 4.7
The distribution of source singularities on these panels forms
Mooring line azimuth μ Deg 120 the velocity potential describing the fluid flow around the spar.
The pressure distribution on the immersed part of the spar is
V. DESCRIPTION OF THE NUMERICAL MODEL calculated from the velocity potential. Fig. 6-a shows the panel
distribution for the spar. The transfer functions of the three
The floating structure consists of the wind turbine specified
component forces and three component moments result from
and delivered by the DeepCwind Consortium mounted on a
the integration of this pressure distribution for every frequency.
spar-buoy type floater. The Spar floating wind turbine
described in the previous chapter was used to carry out this
numerical study. The Spar is balanced is such way that it stays
vertical when no external loads act on it.
Figure 5-a Figure 5-b
Spar under water geometry Side view of geometry on which
diffraction/radiation wave forces are Figure 7. 3D-view of the mooring system
calculated
The overall stiffness of the mooring system of the
Figure 6. Under water geometry.
numerical model is compared to the stiffness of the physical
model in the next chapter.
B. Viscous loads
In addition to the radiation contribution of the potential- D. Wind loads
flow theory, some viscous loads are added to the hydrodynamic The turbulent wind is described as a time series of the
loading on the spar-buoy. This viscous damping is added horizontal wind velocity taken as uniform over the entire rotor
thanks to linear damping coefficients for the three translation plane. Wind load coefficients are commonly used to calculate
velocities and the three rotation velocities. the wind loads on a floating structure. These coefficients are
defined around the centre of gravity (Cog) of the floater. This
C. Mooring loads approach is not meant for a rotor of a wind turbine. Therefore,
In this simulation the seabed is horizontal and the water at best this standard model can give the expected average surge
depth is 200 m. The Spar is moored using three taut lines with offset and average heel angle of a wind turbine in a steady and
azimuth angles of 0 deg, 120 deg and 240 deg. Line 1 is the uniform wind field provided that the coefficients for the surge
line connected to the stern of the Spar. Line 2 is the line mode and for the pitch mode are appropriately derived. The
connected at starboard. Line 3 is the line connected at portside. formulation for the force in the surge direction and the pitch
Each line is modeled individually as a spring with constant moment around the y-axis passing by the CoG are:
stiffness and a pre-tension. Fig. 7 shows a picture of the • Surge wind force: Fx = ½ ρa CFx U2 Af
mooring system when no environment acts on the Spar.
Following the same approach as in [9], the yaw spring stiffness • Trim wind moment: My = ½ ρa CMya U2 Af Lpp
of the mooring system is increased thanks to a linear spring
stiffness coefficient to better account for the “crowfoot”
VI. CALIBRATION OF NUMERICAL MODEL
connection to the spar. Eventually the mooring loads on the
Spar are the superposition of the tension in the three lines at the The numerical model has some shortcomings:
connection points to the spar plus the moment resulting from
• The mooring system is simplified.
this extra yaw spring stiffness.
• Viscous damping must be added.
• Wind forces are based on coefficients.
• The whole floating turbine is a rigid body.
• It lacks a model of a wind rotor.
Fortunately it is possible to calibrate the model based on the
results of some model tests so that we can find our way around
some of these obstacles.
A. Mooring stiffness The similar decay tests as those in the model tests can also
The global stiffness of the mooring system in surge (x) and be simulated with the numerical model and by the way checked
sway (y) can be determined thanks to the results of static load against measurements. Fig. 10 and Fig. 11 show the
tests, which were carried out during the model tests. These comparison between measurements and simulations of the
static load tests can also be simulated with the numerical model decay tests for heave and pitch.
with the simplified version of the mooring system. Fig. 8 and
Fig. 9 show the resulting stiffness plots for both the model tests C. Rotor loads
and the simulations. As can be shown in Fig. 8 and Fig. 9, the All loads acting on the rotor are transferred to the floater.
stiffness in x and y direction of the simulated model are quite The following parameters come into play for the description of
comparable with those in the model test. the dynamics of a rigid rotor:
• Wind: a 3 dimension description of the wind field is
required for a good assessment of the wind load. The
rotation of each blade and the correct angle of attack
should be accounted for in this calculation.
• Power extraction and power loss: the purpose of the
turbine is to extract power. Some power is also lost in
this conversion process.
• Rotational speed: the rotation of the rotor induces in a
gyroscopic effect.
The measurements done on the rotor were not available for
this study. Moreover these loads are not described accurately
Figure 8. Static loads in x direction enough by the numerical model as it is in its standard version.
Therefore, these loads cannot be accounted for in this study. As
a consequence all model tests with wind are excluded from this
study. Even without wind, the rotating rotor feels a relative
wind and it generates a thrust; moreover it remains under the
control of the power take off system which may slow it down
or accelerate it. This dynamic behavior of the rotor cannot be
described by the standard simulation tool.
The average thrust will cause a drift offset. This drift offset
will be visible in the horizontal translations and in the rotations,
mainly in pitch. As we cannot include these effects, we will try
to limit our comparison to the motions caused by the waves.
measurement
0
aNySIM
are outside the wave frequency range. In this respect the
motion response amplitude operators (RAOs) are a rather fair
-5
2000 2050 2100 2150 2200 2250 way to compare the simulations with the model tests. The
5 comparison is done on a frequency range in which the wave
Acc-Mid-X [m/s2]
measurement
aNySIM
energy is large enough. This is achieved by calculating the
0 RAOs for the spectral density of the wave higher than 10 % of
the peak of the wave spectrum. As illustration the part of the
-5
2000 2050 2100 2150 2200 2250 spectrum that is considered to calculate the RAO is colored in
2 orange in Fig. 16. RAOs for the heave motion and the pitch
Acc-Low-X [m/s2]
measurement
aNySIM rotation are calculated. These RAOs can also be deduced
0
directly in frequency domain from the diffraction results by
-2 rewriting (1) for an sine wave excitation. These RAOs are
2000 2050 2100 2150 2200 2250
Time [s] plotted with a blue line in Fig. 14, Fig. 15, Fig. 17, Fig. 18,
Fig. 20 and Fig. 21.
Figure 12. Comparison of accelerations at 3 locations (Low, Mid & Top)
between aNySIM and measurements. A. White noise tests simulation results
A ‘white noise’ kind of wave spectrum is chosen so that the
In a second stage the motions of the spar during the decay
response of the spar can be plotted in terms of RAOs over a
model tests are forced into the numerical model. In this way the
wide frequency range. Fig. 14 and Fig. 15 shows the RAOs of
calculated values of the tensions in the lines can be compared
the heave and pitch motions as calculated from the
to the measured values. This tension check is done with feeding
measurements and from the simulations.
the numerical model with the motions resulting from the pitch
decay. Fig. 13 shows that the match is not perfect, but the
average values and the trend of the tensions in the three lines
are found back by the numerical model. Note that the simulated
tensions are disturbed by numerical noise, which is a
consequence of the way the motion are forced into the model.
The numerical model is fed in with wave time series in order
to compare the simulation results with the model tests. Table
II shows the statistics of the simulated motions. Table III
shows the statistics of the measured motions. In this sea state
the Spar does not move much. However the differences
between the simulation results and the test results are
noticeable. In the test, the drift horizontal drift motions is
clearly visible, as well as the trim angle. More remarkably the
Spar moves sideways and she rolls in head sea. The numerical
model is incapable of capturing this behavior.
Signals Standard
Unit Average
deviation
Surge m -0.0 0.08
Signals Standard
Unit Average
deviation
Surge m 0.15 0.20
ACKNOWLEDGMENTS
The authors would like to thank their colleagues of MARIN
for their enthusiast collaboration in the execution of this work
and in the redaction of this paper. The authors are very obliged
for the DeepCWind consortium, which provided them with
very valuable data for the comparison of simulations with
measurements.
REFERENCES
[1] A.R. Henderson, B. Bulder, R.H.M. Huijsmans, J.M. Peeringa, J.T.G.
Pierik, E.J.B. Snijders, M.Th. van Hees, G.H. Wijnants and M.J. Wolf,
2003, “Floating Windfarms for Shallow Offshore Sites”. Proc. Of
OWEMES Conference, Naples, Italy.
[2] B. Bulder, A.R. Henderson, R.H.M. Huijsmans, J.M. Peeringa, J.T.G.
Figure 21. Pitch RAO in survival sea-state Pierik, E.J.B. Snijders, M.Th. van Hees, G.H. Wijnants and M.J. Wolf,
2003, “Floating Wind Turbines for Shallow Waters”. EWEC 2003,
Madrid, Spain.
VIII. CONCLUSION [3] W.E. Cummins, “The Impulse Response Function of Ship Motions”,
International Symposium on Ship Theory, Hamburg, number 8, 1962.
In this study a numerical model of the Spar has been made [4] G. van Oortmerssen, “The Motions of a Moored Ship in Waves”. NSMB
and this model has been calibrated against the model test Publication No. 510, 1973.
results. The simulation results have been compared with the [5] M. Naciri, O.J. Waals and J.J. de Wilde, “Time Domain Simulations of
model test results for two tests without wind: an operational Side-by-Side Moored Vessels, Lessons Learnt from a Benchmark Test”,
OMAE2007-29756, San Diego, California.
sea state, and a design sea state. Clearly the numerical model
[6] J.J. de Wilde, A.W. van Dijk , A.J. van den Berg and J. Dekker, “Direct
without a model for the rotor failed to restitute the behavior of Time Domain Downtime Assessment for LNG Operations using
the Spar in the operational sea-state. Nevertheless the work Computer Cluster”, ISOPE 2009.
was continued to look specifically at the motion response of [7] J. Serraris, 2009, “Time Domain Analysis for DP Simulations”. OMAE
the Spar to the waves. Heave RAOs and pitch RAOs were 2009-79587, Honolulu, Hawaii.
used for this comparison. This comparison was done for [8] C. Lindenburg, “PHATAS Release “APR-2005” USER’S MANUAL,
Program for Horizontal Axis wind Turbine Analysis and Simulation”,
frequencies where the wave energy is big enough. Despite that ECN-I—05-005 Rev.7, ECN, Petten, December 2010.
the numerical tool is not fit for the simulation of a wind [9] J.M. Jonkman, “Phase IV of the IEA Annex XXIII Offshore Code
turbine including the dynamics of the rotor, the first order Comparison Collaboration”, 2010, Technical Report NREL/TP-500-
wave responses in heave and pitch are comparable for these 47535.
tests. However the numerical tool slightly underestimates the [10] User manual of aNySIM, http://wiki.marin.nl/index.php/ANYwiki.
heave and pitch responses for the most severe sea-state.