The document discusses simulation requirements and relevant load conditions for designing floating offshore wind turbines. It summarizes findings from simulations of a reference 10MW turbine and floating platform design in the Gulf of Maine. Key findings include: (1) Initial conditions are important to reduce transient behavior, with platform heave taking longest to converge; (2) A run-in time of 1000 seconds is sufficient for loads to reach stationarity; (3) Sensitivity analysis found wind speed and wave height most influential on loads; (4) 8 simulations are needed to estimate fatigue loads within 5% accuracy; (5) Simulation length under 3 hours has little effect on load statistics if using multiple wind seeds. The methodology focused on identifying best practices for floating turbine simulations
Simulation requirements and relevant load conditions in the design of floating offshore wind turbines
1. Stuttgart Wind Energy (SWE)
@ Institute of Aircraft Design
Simulation requirements
and relevant load
conditions in the design
of floating offshore wind
turbines
Ricardo Faerron 1, Kolja Müller 1, Po Wen
Cheng 1, Luca Vita
1-University of Stuttgart
International Conference on Ocean, Offshore & Arctic Engineering
June 2018, Madrid
2. • The industry has little experience compared to fixed offshore and onshore turbines
• Limited specifics when it comes to guidelines for simulation
• Many different floating platform concepts
• Different dynamics than fixed offshore turbine, lead to longer simulation times
• 1 to 3 hours per seed
2
Why investigate the simulation requirements and
relevant load conditions for floating wind turbines?
Need to investigate methodologies and best practices for simulations
3. Overview
• Wind Energy Research at University of Stuttgart
• Introduction to the LIFES50+ Project
• Reference model
• Design requirements of floating wind turbines
3
4. Research Groups at SWE
1- Testing and Measurement
2- Control, Optimization and
Monitoring
3- Conceptual Design and
System Simulation
4
Wind Energy Research at the University of Stuttgart
• Aero-Servo-Hydro-
Elastic MBS
simulation of WT
• Detailed Modelling of
subcomponents
• Code development
• Floating platform
design
5. OBJECTIVES
• Optimize and qualify to Technology Readiness Level (TRL) of 5, two substructure designs for 10MW
turbines.
• Develop a streamlined and KPI (Key Performance Indicator) based methodology for the evaluation
and qualification process of floating substructures.
5
Investigation within the LIFES50+ Project
6. 6
Reference design turbine and platform
Set up of model
Publicly available turbine and platform models
• DTU 10MW reference wind turbine
• Dr. techn. Olav Olsen AS floating platform
• Anchored to the sea bed with 3 catenary steel
mooring lines with additional ballasting weights
• SWE controller
Time domain simulation with NREL’s FAST8
• Aerodyn v14 module for aerodynamics
• Hydrodyn module with first order potential flow
theory with Morison drag forces for hydrodynamic
• Dynamic model of mooring lines with Moordyn
• Periodic 10 min full-field turbulent wind fields
www.olavolsen.no/
7. What do we want to analyse?
• Ultimate loads
• Fatigue loads (damage equivalent loads or
DEL)
• Assumes a straight S-N curve
(stress vs number of cycles) on a
log-log scale
• Wohler coefficient 4
7
Blade
Flapwise
bending
moment
Fairlead
tension
Tower
fore-aft
bending
moment
www.olavolsen.no/
Load sensors
8. • Gulf of Maine
• East coast in the United States of North
America and closely linked to the
medium severity site as defined in
Lifes50+ [Krieger et al., 2015]
• Water depth 130m
• Wind environment (as derived by
measurement buoy)
• Weibull scale coefficient of 6.2
• shape coefficient of 1.7
8
Reference environmental conditions
Portland
https://oceanservice.noaa.gov/
Gulf of Maine, U.S.A
9. 1 - Pre-simulation initial conditions
2- Initial transient effects and run-in time
3 - Sensitivity to environmental parameters
4 - Effect of simulation length on DEL
5 - Number of seeds needed
6 - Influence of wave peak shape parameter (JONSWAP spectrum)
9
Simulation requirements for floating turbines
Note: due to the wide range of possible platform geometries and dynamics,
emphasis should be given to the methodology
11. • Initial conditions as inputs for the simulations are important to reduce the simulation times
by reducing the transient behaviour during start-up
11
1- Pre-simulation initial conditions
Simulation requirements for floating turbines
Simulation conditions for Power production
case (DLC 1.2)
• Deterministic wind: 4-24m/s
• No waves
Analysis of the mean value for different 600s
time frames
Example simulation results for platform heave for
determination of initial conditions values𝝁 𝑛𝑜𝑟𝑚 𝑡 𝑠𝑡𝑎𝑟𝑡, 𝑣 =
𝝁 𝒕 𝑠𝑡𝑎𝑟𝑡, 𝒗
𝝁 3001𝑠, 𝒗
𝒕𝑖,𝑠𝑡𝑎𝑟𝑡 = 1𝑠, 101𝑠, … , 2901𝑠
𝒕𝑖,𝑒𝑛𝑑 = 𝒕𝑖,𝑠𝑡𝑎𝑟𝑡 + 600𝑠
t1 = 0-600s
12. Important degrees of freedom:
• Rotor speed
• Generator speed
• Blade pitch angle
• Tower top fore-aft displacement
• Platform surge
• Platform heave
• Platform pitch
12
1- Pre-simulation initial conditions
Simulation requirements for floating turbines
Convergence of normalized mean value for platform
pitch DOF, all wind speeds (each colour is a different
wind speed)
13. • The platform heave takes the longest to converge
• This is due to the fact that the mean value of the
final time frame of some simulations is almost 0
• Already reaches a value with less than 10%
deviation from the expected value with a starting
time of 1101s for the time window.
• Implies total simulation time for starting
conditions would be of 1701
13
1- Pre-simulation initial conditions: Results
Simulation requirements for floating turbines
Convergence of absolute maximum deviation for all
analysed sensors taking into account all wind
speeds
15. 𝜎 𝑛𝑜𝑟𝑚(𝑡) =
𝜎 𝑡
𝜎 𝑡 = 2000
t T
t = 2000 T
𝜎 = 𝑠𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛
• Check for stationarity of the load sensor: Initial transients time need to be disregarded
• Tested different approaches: moving average, convergence of frequency response,
forward statistic convergence,
• Standard deviation 𝜎 of the remaining time series was chosen
15
2- Initial Transient effects Run-in Time
Simulation study
• 13698 Simulation
• 8 varying environmental conditions (case
of severe sea state as in DLC1.6)
• Total simulation length T = 11800 seconds
• Check convergence of the normalized
standard deviation
16. Longest to settle at a stable standard deviation
• Platform surge
• Platform sway
Results
Stationarity can be seen after 600s, were the
STD of the simulations is within 5% of the
expected value, with 1000s being a conservative
value.
Further investigation could include Heidelberger
and Welch's convergence diagnostic
2- Initial Transient effects Run-in Time
PtFm=platform,TT=Towertop,Dsp=displacement
Run-in-time evaluation for DLC 1.6.
18. • In order to determine the critical environmental conditions across a wide range of values, a sensitivity
analysis can be used
• Sampling methods help reduce amount of simulations: Sobol sampling method is used as described in
[Müller,2017]
18
3- Sensitivity to environmental parameters
Simulation conditions for power production case (here
power production for fatigue design DLC1.2)
Three load ranges differentiate between controller
mode:
• LR1: below rated wind speed
• LR2: around rated wind speed
• LR3: above rated wind speed
3 seeds simulated with average used for evaluation Scatter plot of environmental
yellow: LR1, red: LR2, green: LR3.
19. Environmental condition Values
LR1
wind speed [m/s] 4 : 0.1 : 10.2
wave height [m] 0.3 : 0.1 : 3.1
wave period [s] 2.5 : 0.1 : 10.7
LR2
wind speed [m/s] 10.2 : 0.1 : 13.8
wave height [m] 0.3 : 0.1 : 4
wave period [s] 2.5: 0.1 : 10.7
LR3
wind speed [m/s] 13.8 : 0.1 : 25
wave height [m] 1.2 : 0.1 : 6.6
wave period [s] 4 : 0.1 : 10.7
Turbulence intensity [%] Class C
Wind direction [°] 0
Wind seeds / wave seeds 3
Wave direction [°] 0
Current speed [m/s] 0
Current direction [°] 0
Simulation length
4,600 (3,600 for
statistics)
Total number of simulations 2799 19
3- Sensitivity to environmental parameters
Note: methodology could be
expanded to include more
environmental parameters
that vary
Environmental boundary conditions for sensitivity study
based on DLC1.2
20. 20
3- Sensitivity to environmental parameters
Exemplary scatterplots : tower base bending moment DEL
21. 21
3- Sensitivity to environmental parameters
• The chi-squared test is calculated for each
environmental parameter and load range
• The test results in a p-value which gives the
probability that the correlation between two
observed variables is random
• Data can then be ranked
Ranking table
• Importance
• Useful for early design stages to understand
the system
• When used for more parameters one can
determine the most important ones which
drive the loads
23. • Wind and wave environment are stochastic: how many seeds need to be simulated to
obtain an accurate fatigue estimation?
23
4- Number of seeds needed
One set of environmental conditions for Power production
case with large DEL was chosen (rated wind)
1000 simulations were performed
Example distribution of tower base
bending moment DEL values.
Case: change wind
and wave seeds
Run in time later
removed [s]
600
Total simulation length
[s]
4,200
Simulations 1,000
Wind speed [ms-1] 12
Turbulence intensity [%] 14.6
Wave height [m] 6.3
Wave period [s] 7.8
24. • Bootstrap analysis with increasing number of seed
based on 5000 draws with replacement
• The median, 1st, 5th, 95th and 99th percentile
values are determined
• Largest scattering: blade root flapwise bending
moment
• For it, in order to be within 5% of the expected
mean 95% of the time, one needs to perform 8
simulations (4 seeds to be within 6%)
4- Number of seeds needed: Results
Plot of Lifetime DELs bootstrap evaluation
(Red horizontal lines indicate median, box borders 95th
percentile and whiskers 99th percentile values. Normalized
against the Lifetime DEL of all 1,000 simulation)
26. • DNV offshore standard for design of floating wind
turbines [Det Norske Veritas, 2013]
• For operating conditions, a minimum of 3 hours are
to be simulated “to adequately capture effects such
as nonlinearities, second order effects, and slowly
varying responses, and to properly establish the
design load effects.”
• But it may good to split a simulation into a larger
number of shorter simulations
• E.g. parallelise computation
• 3 hours wind files are too large
26
5- Effect of simulation length
Case: change wind
and wave seeds
Run in time later
removed[s]
600
Total simulation length [s] 4,200
Simulations 1,000
Wind speed [ms-1] 12
Turbulence intensity [%] 14.6
Wave height [m] 6.3
Wave period [s] 7.8
Simulation settings
• Simulation study: 1000 simulations, each with
different wind and wave seed ( periodic wind files)
• statistical value of one simulation is compared to the
option of combining a multitude of shorter
simulations.
27. 27
5- Effect of simulation length
2- Box with 1000
simulations of same length
(t seconds)
3- Pick X number of
simulations from box to
make up 3600s
( e.g. 6 * 600s)
5- Repeat process
5000 times to obtain
statistics of DEL
4- Calculate Lifetime
DEL of combined X
simulations
1- For each of the 1000 simulation,
different time frames are taken
t1
t2
t3
t…
28. 28
5- Effect of simulation length: results
5th and 95th percentile of Lifetime DEL with varying
simulations length. (normalized to the lifetime DEL of
the 1000 simulations with 3600s)
Result show:
• Decrease on simulation length does not effect
statistics of the sampled DELs for the fairlead
tension and the tower base fore-aft bending
moment
• Changes in statistics of blade root flapwise DEL
• Remembering that wind field is cycled every
10 min (i.e. periodic)
• Using shorter simulation times means more
wind seeds are considered in the individual
DEL calculation
• Lead to less spreading
29. • Simulation setting should not be overlooked as important factors for calculation of
ultimate and fatigue loads
• Good interpretation and sometimes further review of results is important to come up with
an appropriate analysis of the data
• Findings can be platform concept dependent (excitation frequencies and hydrodynamics)
and focus should be given to the methodology
• Continuing work
• expansion of the considered environmental conditions
• analysis of second platform concept (Nautilus platform)
• Future report: Deliverable 7.7 Lives50+
29
Summary and Remarks
30. e-mail
phone +49 (0) 711 685-
fax +49 (0) 711 685-
University of Stuttgart
Thank you
Questions?
Ricardo Faerron
68243
faerron@ifb.uni-stuttgart.de
The research has received partial funding from the European Union’s Horizon 2020 research and
innovation programme under grant agreement No. 640741 (LIFES50+). We are grateful to Dr. techn.
Olav Olsen AS for the permission and contribution to the public semi-submersible design based on
their OO Star Wind Floater (www.olavolsen.no).
31. • A. Krieger, G. K. V. K. Ramachandran, L. Vita, P. G. Alonso, G. G. Almeria, J. Berque and G. Aguirre,
"LIFES50+ D7.2 Design Basis," 2015.
• Det Norske Veritas, “Design of Floating Wind Turbine: DNV-OS-J103”, Det Norske Veritas AS, 2013.
• K. Müller and P.W. Cheng, "Application of a Monte Carlo Procedure for Probabilistic Fatigue Design of
Floating Offshore Wind Turbines," Wind Energ. Sci., 3, pp. 149-162, 2018
• K. Müller et. al., "LIFES50+ D7.7 Identification of critical environmental conditions and design load
cases," 2018, to be published.
31
References