This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

On the failure probability of offshore wind turbines in the

China coastal waters due to typhoons: A case study using
the OC4-DeepCwind semisubmersible
Yichao Liu, Sunwei Li, P.W. Chan, and Daoyi Chen

 megacities along the China coastline and to reduce greenhouse

Abstract—The paper introduces a set of modelling gas emissions [2]. Before offshore wind farms can be
methods to predict the failure probability of offshore wind deployed in the China coastal areas, several technical
turbines in the China coastal waters. In detail, a series of challenges should be addressed [3]. Among the challenges, the
full-set three-dimensional meteorology simulations are threats posted by typhoons are our concern in the present
conducted using the Weather Research and Forecast study. It is noted that the offshore areas along the China
(WRF) model to provide the extreme wind fields while the coastline are exposed to the hazards of typhoons, who bring
extreme wave fields are predicted according to the strong winds and waves and exert a destructive influence on
conditional probability model. After the extreme wind and the offshore wind turbine [4]. The assessment on the
wave fields are prepared, the ultimate loads on the three long-term reliability of the offshore wind turbine, which is
critical parts of an OC4-DeepCwind semisubmersible wind strongly influenced by typhoons, should hence be considered
turbine are simulated using the Fatigue, Aerodynamics, in planning offshore wind farms in the China coastal waters.
Structures, and Turbulence (FAST) code, which in turn It is common to assume that the long-term reliability, which
leads to the prediction of the failure probability. It has is essentially the complement of the failure probability, of an
been found that the OC4-DeepCwind semisubmersible offshore wind turbine can be estimated through calculating the
constructed in the Taiwan Strait has higher chance to fail ultimate loads under the most severe 10-min mean wind and
under typhoon conditions than in other parts of the China wave conditions in a life cycle of the wind turbine [5]. Given a
coastal waters, showing a maximum failure probability of specific offshore wind turbine design, the limit states of three
0.6. In addition, the failure probabilities in the South and critical parts, namely the rotor blade, tower base and mooring
East China Sea are in the range of 0.3-0.45. The lines, under the extreme wind-wave conditions are analyzed.
probability densities of the ultimate loads show, in some Following the philosophy of the reliability analysis, Ronold
cases, a bimodal shape, which indicates that the loads as and Larsen [6] proposed a probabilistic model to predict the
well as the failure probability are not in a simple linear failure of rotor blades and proved that the failure due to
relationship with the extreme wind speeds and wave flapwise bending moments should be given priority in the
heights/periods. blade design. Using the similar model, Toft and Sørensen [7, 8]
have discussed the design of wind turbine blades. As for the
Index Terms—Extreme wind and wave fields; Failure tower base of an offshore wind turbine, Sørensen and
probability; FAST code; Offshore wind turbine; WRF model. Johansen [9] applied the probabilistic model in the estimation
of the tower base failure. Based on the model, Rose et al. [10],
I. INTRODUCTION Garciano and Koike [11] quantified the hurricane/typhoon
The China coastal waters have been identified as suitable hazards for offshore wind turbines erected off the shores of the
for the construction of offshore wind farms thanks to its United States and the Philippines, respectively. In addition, the
ambient wind resources, high frequencies of strong wind failure of mooring lines has drawn attentions from scholars for
observations and good wind power stabilities [1]. Hence, various types of offshore structures [12-15], and their analysis
offshore wind energy exploration has been included in the technique could be applied to the offshore wind turbine design.
national development plan and therefore is strongly supported Even though these published researches provide a series of
by the government to ease the power shortage situation in reasonable methods to estimate the structural failure of
offshore structures, it is still challenging to provide a reliable
This work was supported in part by National Natural Science Foundation failure probability map of offshore wind turbines for a
of China under Grant 51608302, Shenzhen Special Funds for Future Industry specified sea area. It is because a) the environmental loads
Development under Grant 201411201645511650, the Economy, trade and acting on the offshore wind turbine are modelled purely as
Information Commission of Shenzhen Municipality under Grant random variables with certain probability distributions in the
SZHY2014-B01-001 and the Development and Reform Commission of
Shenzhen Municipality under Grant DCF-2018-64. previous studies. In other words, the physical natures of the
Y. Liu, S. Li and D. Chen are with the Division of Ocean Science and loads are not included in the failure probability assessment. b)
Technology, Graduate School at Shenzhen, Tsinghua University, Shenzhen There is no systematic study, to the best of authors’
518055, Guangdong, China (e-mail: liuyc14@mails.tsinghua.edu.cn; knowledge, tried to put together an investigation considering
li.sunwei@sz.tsinghua.edu.cn; chen.daoyi@sz.tsinghua.edu.cn).
P.W. Chan is with Hong Kong Observatory, 134A Nathan Road, Kowloon, different failure mechanisms systematically (failure of blades,
Hong Kong (e-mail: pwchan@hko.gov.hk). towers and mooring system). More specifically, the published

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

researches generally focus on a single failure mechanism of mooring lines are derived from the dynamics simulation. The
the floating wind turbine. For instance, Rose et al. [10] comprehensive failure probability including abovementioned
quantified the failure probability of offshore wind turbines three critical parts of offshore wind turbines using the damage
considering only the tower bending moment in the absence of model is illustrated in Fig. 1.
ocean wave loads. Therefore, the ultimate loads on each part
of the offshore wind turbine, such as the blade, the tower and
the mooring lines are calculated and assessed independently.
In reality, an offshore wind turbine with a floating foundation
should be considered fail if one of the three critical parts fails.
In such cases, the ultimate loads and the corresponding
resistances in each critical part should be systematically
included in a comprehensive model predicting the failure of
the floating offshore wind turbine. More importantly, the
probabilistic models specified for predicting the failure of
offshore structures in China coastal areas have not been
developed, mainly due to the lack of studies focusing on the
estimation of extreme wind and wave conditions in the sea
areas boarding China [16].
In order to estimate failure probabilities of floating offshore
wind farms in China coastal waters, a numerical framework
including a set of modelling methods is introduced. In the
Fig. 1 Flowchart of the numerical simulation processes in terms of
approach described in the present research, three major failure predicting the failure probability of offshore wind turbines.
mechanisms, not only the tower base buckling, but also the
blade root buckling and mooring line tensions are extracted
from the numerical dynamics simulations and then combined A. Study area
to assess the failure probability of offshore wind turbines. In the meteorology simulation of extreme wind fields, a set
More importantly, the extreme wind/wave loads are of grid points covering the coastal waters with the depth
numerically produced based on a series of full-set varying from 20m to 1000m are established (latitudes
meteorology simulations, which certainly show the relations spanning from 17 to 40 and longitudes spanning from
between ultimate wind/wave loads and the geographic and 105.5 to 125 ) with the horizontal spacing varying from
meteorological characteristics along the coast of China. 10m to 30m, to formulate the study area, as shown in Fig. 2.
The remainder of the paper is organized as follows. Section
II briefly summarizes the numerical simulation processes, in
which the modelling methods of the three-dimensional
typhoon wind fields as well as the wave fields are articulated.
Following the philosophy of Monte-Carlo simulations,
numerical simulations are conducted to predict the ultimate
loads on the three critical parts of the offshore wind turbine in
section III. Through summarizing the numerical simulation
results statistically, the failure probability of the offshore wind
turbine is derived in section IV. Section V presents the
conclusion remarks.

II. EXTREME WIND AND WAVE FIELDS (a) (b)

With the purpose of illustrating the modelling methods to Fig. 2 Calculation domain from 17° to 40° and from 105.5° to 125° .
predict the failure probability of offshore wind turbines, a (a) Water depth (unit: m). (b) Distance from the shore (unite: km).
flowchart has been illustrated in the Fig. 1. It is evident from the figure that the water depth varies from
As shown in the Fig. 1, three components, namely the 20m to 200m and the distance from the shore ranges from
meteorology simulation, dynamics simulation and damage 50km to 200km in the most parts of the study area, and hence
model are combined in the numerical framework introduced in should be considered as suitable for the deployment of
the present study to predict the failure probability of offshore semisubmersible floating wind turbines [18].
wind turbines. It is noted that the extreme wind loads play a B. Extreme wind fields
dominate role in predicting the failure probability of offshore
1) Modelling techniques
wind turbines under typhoon conditions. In other words, the
In present study, the Weather Research and Forecast (WRF)
extreme wind speed, regarded as the critical independent
model [19, 20], is employed to simulate a series of artificial
variable, should be given priority in modelling the extreme
typhoons influencing the China coastal waters. Since there is
loads [17]. Hence, the meteorology simulation is employed to
not a single historical reanalysis database containing enough
obtain realistic typhoon wind fields while the ultimate loads
environmental fields suitable for the growth of the number of
on not only the tower base, but also the blade root and the
typhoons planed in the present study (>1500), a modelling

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

method has been developed by the authors [21] to prepare the

boundary and initial conditions for the artificial typhoon
simulation. Following the steps articulated in the authors’
previous publications, the historical reanalysis data provided
by the European Centre for Medium-Range Weather Forecast1
(ECMWF) is employed to initialize a WRF simulation, as
shown in Table I.
TABLE I
DETAILS OF ECMWF HISTORICAL REANALYSIS DATA FOR WRF SIMULATIONS.
ECMWF data Description
Time June to October from 1979-2015 years.
3D variables U and V component of wind, Temperature, Relative (a) (b)
humidity and geopotential height. Fig. 3 Joint probability distributions in relation to the marginal distributions
26 Levels (hPa) 10, 20, 30, 50, 70, 100, 150, 200, 250, 300, 350, 400, of the (a) Location (longitude and latitude) and (b) Intensity (central pressure
450, 500, 550, 600, 650, 700, 750, 800, 850, 900, and maximum sustainable wind speed) of historical typhoons.
925, 950, 975, 1000
2D variables 10-meter U and V components of wind, Surface 2) WRF Simulation
pressure, Mean sea level pressure, skin temperature, With the purpose of capturing the details of the
2-meter temperature, 2-meter relative humidity.
Areas 106 -150 , 2 -50
three-dimensional typhoon wind fields while keeping tracks of
Resolution 0.75 0.75 large-scale processes, the nest-domain configuration with
Interval 6 hours three inter-chained domains is employed in the WRF model.
In order to cover the general meteorological patterns of In detail, the outmost domain (D01) with fixed grid points
historical typhoons influencing China coastal waters, 37 covers the Northwestern Pacific Ocean and the coastal regions
typhoon seasons (June to October in each year) of the regional of China (latitudes spanning from 2 to 50 and
ECMWF data, whose latitude spans from 2 to 50 N and longitudes spanning from 105 to 150 ). The inner
longitude spans from 106 and 150 are used. According domains, including the intermedia (D02) and innermost (D03)
to the official descriptions for WRF model 2 , the basic domains, are configured to move with the vortex (artificial
variables (five 3D variables at 26 pressure levels and seven 2D typhoon) to obtain the detailed three-dimensional wind fields.
variables at surface level) are employed. The data is available The horizontal grid spacing of D01, D02 and D03 are ,
four times at 6-hour interval a day (at 00, 06, 12, 18 UTC). In and , respectively. In the vertical direction,
fact, Carvalho et al. [22] compared the performances of terrain-following layers, whose vertical spacing increases with
different reanalysis data in simulating the offshore winds and heights, as proposed by Laprise [26], are formulated to
concluded that the ECMWF production with is discretize the atmosphere spanning from to .
more remarkable. Hence, following Carvalho et al. [22], the The physical parameterizations of the WRF model are
regional ECMWF data with the resolution of summarized in Table II.
is selected for the initialization of the WRF model. TABLE II
CONFIGURATIONS OF ALL THE PHYSICAL PROCESSES FOR
Considering the small-scale perturbations necessary for the APPLICATION OF THE WRF SOFTWARE.
typhoon genesis are absent due to the relatively coarse
Domain D01 D02 D03
horizontal resolution of the regional ECMWF data, a
Configuration 3 nested domains (vortex following), Mercator projection
three-dimensional Rankine vortex with the user-defined key Grid points
parameters is inserted into the initial field as the “seed” of Time step Adaptive time step (CFL 1.6)
artificial typhoons using the numerical tool included in the Surface layer: MM5 scheme [27, 28]
WRF software package [23]. The inserted Rankine vortex is Boundary Layer: YSU scheme [29]
Land surface model: MM5 5-layer thermal diffusion [30]
defined by the longitude and latitude of the vortex center, the Physics Cumulus Parameterization: Kain-Fritsch scheme [31]
radius to maximum sustainable wind speed ( ) and the Microphysics: WSM3 scheme [32]
maximum sustainable wind speed. The locations and the Radiation Physics: RRTM scheme [33]
maximum wind speeds of the artificial typhoons are drawn 1-D ocean mixed layer model [34]
from the probability distributions constructed from the After the WRF settings are configured, 1510 artificial
marginal distributions, which are produced by the best track typhoons are simulated to provide the extreme wind fields.
data of historical typhoons maintained by the Hong Kong Based on the simulations of artificial typhoons, the probability
Observatory using the empirical Copula model [24], as shown distribution of extreme wind speeds is estimated. In detail, a)
in Fig. 3. all the simulated typhoon wind fields are interpolated onto the
As for the , it is estimated as a function of the central pre-established grid points of the study area (Fig. 2) using the
pressure ( ) according to Chang et al. [25], as cubic spline interpolation technique [35]. b) The circular
(1)
sub-region method [36] is then employed to extract the
extreme wind profile at each grid point. c) Afterwards, the
Generalized extreme value (GEV) distribution model [37] is
1
employed to fit the series of the extreme wind speeds
European Centre for Medium-Range Weather Forecasts: calculated corresponding to all the typhoon simulation cases at
http://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/
2
WRF model online tutorial: each fixed grid point using the maximum likelihood estimates
http://www2.mmm.ucar.edu/wrf/OnLineTutorial/Basics/UNGRIB/ungrib_r (MLE) method [38]. The GEV distribution is shown as,
eq_fields.htm

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

𝑥 𝜇 −1/𝑘 As for the spectral peak period, the log-normal conditional

𝐹(𝑥|𝜇, 𝜎, 𝑘) 𝑒𝑥𝑝 { [ + 𝑘 ( )] } (2)
𝜎 probability distribution is recommended based on the
In equation (2), 𝑥 is the extreme wind speed, 𝑘 is the shape significant wave height estimated in the previous step [41],
parameter, 𝜇 is the location parameter and 𝜎 is the scale and the formula is shown as,
2
parameter. All the fitted parameters at 90m level are shown in (𝑙𝑛𝑇𝑝 𝛩)
Fig. 4. 𝑓(𝑇𝑝 |𝐻𝑠 ) 𝑒𝑥𝑝 [ ] (4.a)
√ 𝜋𝜙𝑇𝑝 𝜙2
In equation (4.a), 𝛩 and 𝜙 are the mean and the standard
deviation of the logarithmic spectral peak period 𝑙𝑛(𝑇𝑝 ),
which can be estimated according to the significant wave
height (𝐻𝑠 ) as,
𝛩 𝑝1 + 𝑝2 𝐻𝑠 𝑝3 (4.b)
𝜙2 𝑞1 + 𝑞2 𝑒 𝑞3 𝐻𝑠 (4.c)
In equation (4.b) and (4.c), 𝑝1 , 𝑝2 , 𝑝 , 𝑞1 , 𝑞2 and 𝑞 are
empirical coefficients whose values are suggested by Myrhaug
and Fouques [42] based on the observations obtained via wave
buoys for 29 years (from 1974 to 2002). Integrating the
conditional probability distributions shown in equations (3)
(a) (b) and (4), the probability distributions of the significant wave
height and peak period can be obtained given the probability
distributions of extreme wind speeds. Specifically, the
probabilities of 𝐻𝑠 and 𝑇𝑝 can be calculated as,
𝑓(𝐻𝑠 ) ∫ 𝑓(𝐻𝑠 |𝑥)𝑑𝐹(𝑥) (5.a)
(5.b)
𝑓(𝑇𝑝 ) ∫ 𝑓(𝑇𝑝 |𝐻𝑠 )𝑑𝐹(𝐻𝑠 )
In summary, using the GEV model predicting the probability
of extreme wind speeds shown in equation (2), the probability
densities of significant wave heights and wave peak periods
are estimated according to equations (3)-(5).
(c)
Fig. 4 Three parameters of the GEV distributions at 90m level in China III. NUMERICAL SIMULATIONS
coastal waters. (a) Shape parameter 𝑘. (b) Scale parameter 𝜎. (c) Location
parameter 𝜇. Provided vertical variations of wind speeds, significant
wave heights and wave peak periods, dynamics of an offshore
C. Extreme wave fields wind turbine can be numerically simulated via specially
Since the extreme wave fields are in connection with the designed tools. In the present study, the Fatigue,
extreme wind fields under typhoon conditions, the key Aerodynamics, Structures and Turbulence (FAST) code,
parameters, such as the significant wave height and the which is a comprehensive aero-hydro-servo-elastic
spectral peak period, used to describe the evolution of wave engineering simulator capable of predicting both the extreme
fields could be estimated according to a conditional and fatigue states of the offshore wind turbines, is employed
probability model from extreme wind speeds. According to [43].
Teng and Liu [39], who compared conditional probability The failure of a floating offshore wind turbine can be
models widely adopted by meteorologists to link the wind and assessed according to the maximum bending moment at the
wave fields, the log-normal distribution [40] has been found tower base under the influence of the most severe 10-min
provide reasonable estimates of the key parameters defining wind climate [9]. In addition to the base bending moment, the
wave fields. The log-normal model suggested by Teng and Liu FAST simulation also provides the ultimate loads on the
[39] is shown as, blades and mooring lines of the offshore floating wind turbine,
𝑓(𝐻𝑠 |𝑥𝑖 ) [ /(𝐻𝑠 𝛼 √ 𝜋)] 𝑒𝑥𝑝{( / ) [(𝑙𝑛(𝐻𝑠 ) 𝛽)/𝛼]} (3.a) both of which are utilized to check against the resistances in
Where the present study. In summary, the FAST code is employed to
2 2 2 2 simulate the wind turbine dynamics in 10 minutes’ time spans,
𝛼 √𝑙𝑛[(𝑆𝑖 / 𝑖) + ], 𝛽 𝑙𝑛 [ 𝑖 /√𝑆𝑖 + 𝑖 ] (3.b)
in which the wind and wave loads are calculated based on the
In equation (3), 𝐻𝑠 is the significant wave height, 𝑥𝑖 is the extreme wind speeds and extreme wave heights/periods
wind speed at the height of 10m. When the extreme wind randomly drawn from the probabilistic models derived in
speed series at the height of 10m are divided into pieces, section II. The maximum values of blade root bending
each containing same number of extreme wind speed moments [6, 8], tower base buckling moment [44] and
“samples”, the auxiliary parameters 𝛼 and 𝛽 are calculated mooring line tensions [15] in each of the 10-min simulation
according to equation (3.b) from the mean ( 𝑖 ) and standard case are the indicators assessing the failures of the blade, the
deviation (𝑆𝑖 ) of the “samples” of extreme wind speeds in the tower and the mooring system.
th piece.

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

A. OC4-DeepCwind semisubmersible established in section II. In addition, the geophysical database 3

Considering that the water depth in a dominate portion of developed by the national oceanic atmospheric administration
the study area (90%) varies from 20m to 200m, the (NOAA) of the United States is used in the HydroDyn module
semisubmersible floating wind turbine is an appealing to estimate the influence of water depths on wave evaluations.
candidate for the exploration of offshore wind resources [45]. Parameterizations and settings used to run HydropDyn are
In present study, the OC4-DeepCwind semisubmersible wind summarized in Table IV.
TABLE IV
turbine [46] in accompany with the NREL 5-MW baseline DETAILED CONFIGURATIONS OF HYDRODYN MODULE IN FAST V8.
wind turbine developed by the National Renewable Energy HydroDyn Setting
Laboratory (NREL) is employed as an example illustrating the Water depth The geophysical database developed
proposed numerical framework to estimate the failure
by NOAA
probability of offshore wind turbines. The sketches of the
Wave model JONSWAP spectrum
OC4-DeepCWind submersible are presented in Fig. 5.
Low cut-off frequency of the wave 0.314159
spectrum
High cut-off frequency of the wave 1.570796
spectrum
Significant wave height 𝐻𝑠
Peak-spectral period 𝑇𝑝
Peak-shape parameter
Time step 0.2
3) Other models
(a) (b) (c) The Mooring dynamic model (MoorDyn version 1.00.02)
Fig. 5 Sketches of the OC4-DeepCwind semisubmersible wind turbine (a) provided by the FAST system is employed to simulate the
Front view (b) Side view (c) Bird view. mooring system of the OC4-DeepCwind semisubmersible via
B. Model configurations the lumped-mass modeling approach [48], aiming to capture
the nonlinear relationship between the mooring loads and the
1) Aerodynamic model
displacement of the floating platform. In order to assess the
As a pre-processing tool for the aerodynamic model of the
stability, the pre-tensions of the catenary lines are designed to
FAST system, the TurbSim software developed by NREL
be 25%-35% of the displacement of the floating platform,
(version 1.06) is used to produce time series of
according to the American Petroleum Institute (API) code [49].
three-dimensional turbulent wind speeds based on the extreme
In addition to the MoorDyn, the wind turbine is set parked in
wind speeds at the hub height. In detail, 200 wind speed time
the control and electrical model (SevoDyn version 1.05) to
series, each lasting 630-second, are generated corresponding
avoid the damages induced by extreme winds and waves.
to each extreme wind speed randomly drawn from the GEV
When the aerodynamic loads, hydrodynamic loads, mooring
model derived from fitting the WRF simulation results. The
loads calculated from the abovementioned modules are
first 30-second of the time series is employed to filter out
transmitted into the structural dynamic model (ElastoDyn
transient responses of the wind turbine. In the main simulation
version 1.03), the kinematics and dynamics of the offshore
lasting 600 seconds, the extreme wind loads are simulated
wind turbine are simulated for 10-minute simulation time.
continuously at a time interval of 0.0125 seconds. Other
settings used to run AeroDyn are summarized and presented in C. Post-processing the simulation results
Table III. Using the FAST system, a total number of
TABLE III
DETAILED CONFIGURATIONS OF AERODYNAMIC MODULE IN FAST V8.
simulation cases are carried out corresponding to the 1598
Aerodynamic module Setting fixed grid points in the study area established in section II. In
Flag in AeroDyn v15.03 the simulations, the maximum values of the three loads,
Blade airfoil aerodynamics model Steady model namely the flapwise bending moment at the blade root ( ),
Tower influence on wind around the tower Baseline potential flow
the tower base bending moment ( ) and the mooring line
Tower aerodynamics loads On
Air density 1.225 / tensions ( 𝑇 ), are extracted and checked against the
Kinematic air viscosity 1.464 − 2
/ corresponding resistance of the turbine [7, 8, 44] using the
Speed of sound 335m/s following equation (6),
2) Hydrodynamic model 𝐺(𝑟, ) 𝑟 (6)
On basis of the hydrodynamic model (HydroDyn In equation (6), 𝑟 denotes the bending moment resistance of
version2.03), the FAST code could calculate hydrodynamic the blade root (𝑟 ), the buckling-resistance of the tower base
loads on the offshore wind turbine. In the present work, the (𝑟 ) and the mooring line resistance (𝑟 ) while is the vector
popular Joint North Sea Wave Project (JONSWAP) spectrum of stochastic loads ( , and 𝑇 ). Then the failure
model [47] is employed to generate the irregular wave field. probability of the wind turbine is calculated as
The significant wave height and the spectral peak period, 𝑓 [𝐺(𝑟, ) ] (7)
which are required by HydroDyn to define the JONSWAP
spectrum, are drawn from the log-normal distributions
3
NOAA: https://www.ngdc.noaa.gov

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

D. Summary of verifications IV. DISCUSSIONS

Before further investigating the numerical simulation results, A. Ultimate load maps
it is of necessity to discuss the reliability of abovementioned
Based on the FAST simulation results, the mean values of
modelling techniques. In fact, a series of verifications have
the simulated ultimate loads acting on the three critical parts
been conducted in the previous publication [21] to check the
of the OC4-DeepCwind semisubmersible ( , and 𝑇) are
reliability of the artificial typhoon simulation, as summarized
derived and presented in Fig. 6.
in the Table V.
In detail, the field data provided by the Hong Kong
Observatory is employed to compare with not only the
cumulative distribution function (CDF) of the simulated wind
speeds but also the simulated mean wind profile. The error
indicators of root mean square error (RMSE) and scatter index
(SI) are used to quantify the agreement, as shown in the Table
V.
TABLE V
SUMMARY OF VERIFICATIONS FOR ARTIFICIAL TYPHOON SIMULATION.
Verification Weather Location RMSE SI
type station (m/s) (m/s)
GI station ′ ′′ 0.85 0.098
′ ′′ (a) (b)
WGL station ′ ′′ 0.36 0.048
CDF
′ ′′
TC station ′ ′′ 1.02 0.092
′ ′′
HKO station 0.36 0.048
(𝑈10
Mean wind 𝑚/ ) ′ ′′
profile HKO station ′ ′′ 0.17 0.018
(𝑈10
𝑚/ )
The simulation results are considered as acceptable since
the RMSE and SI are much lower than the error criteria in
other relevant researches [50, 51]. Therefore, the simulated
wind speeds yielded from the artificial typhoon simulations (c)
are reliable, even though the simulation results in the low wind Fig. 6 Mean of the simulated ultimate loads of the OC4-DeepCwind
speed bin (4m/s-6m/s) slightly deviate from the observed data. semisubmersible wind turbine in China coastal waters. (a) (unite: K
As for the dynamics simulation of the OC4-DeepCwind ) (b) (unite: K ) (c) 𝑇 (unite: K )
semisubmersible, the natural periods of floating foundation It is found from the Fig. 6 that the maximum ultimate loads
discerned from the free decay tests, which are regarded as the occur in Taiwan Strait where high-valued extreme wind
essential verification criteria, are checked after comparing speeds are also observed. In fact, the ultimate loads exceed
with the results from a series of previous researches. In special, 5000K , 3.5 K and 10000K respectively for
three degree of freedom (DOF) motion responses with an , and 𝑇 in Taiwan Strait. In addition, the ultimate
initial displacement of 3m or 3 are tested and the loads in the South China Sea and the East China Sea are
comparisons are presented in the Table VI to indicate the significantly larger than in other sea areas (~4000K ,
reliability of simulation results. 2.5 K and 5000K for , and 𝑇), implying
TABLE VI the influences of the typhoons.
SUMMARY OF VERIFICATIONS FOR DYNAMICS SIMULATION.
Research Surge (s) Heave (s) Pitch (s) B. Area-wide failure probability
Coulling et al. (FAST) [52] 107.0 17.3 26.8 Provided the ultimate loads, the damage model (equation
Coulling et al. (Exp.) [52] 107.0 17.5 26.8
Tran and Kim (CFD) [53] 108.1 17.8 25.2
(7)) can be employed to estimate the failure probability. The
Liu et al. (CFD) [54] 107.2 17.5 27.4 rotor blade is currently made up of a fiber-reinforced polyester
The present study 110.1 17.5 25.8 laminate in the NREL baseline wind turbine design, whose
It is evident from the Table VI that the comparisons of the strength determines the flapwise bending resistance of blade
motion period between the present study and experimental root (𝑟 ). Based on a series of experimental verifications [6], it
data, FAST and CFD data show acceptable agreement, even is recommended that a normally distributed random variable
though the period of surge in the present simulation results reliably shows the statistics of 𝑟 , whose arithmetic
seems to be slightly larger. In summary, not only the artificial expectation and the coefficient of variation are
typhoon simulation, but also the dynamics simulation process / 2 and 0.1 [6]. The bending stress induced by the
introduced in the present work is verified according to the extreme loadings at the blade root, which is compared to the
Tables V and VI. resistance, is calculated as 𝑆 / , where is the
flapwise section modulus of the blade root. As for the strength

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7

of the tower base, the buckling resistance with uncertainty (𝑟 )

is modeled according to equation (8) [9-11, 44],
𝐷 𝑋𝑦,𝑠𝑠 𝐹𝑦
𝑟 ( ) [𝐷 (𝐷 𝑡) ]𝑋𝑦,𝑠𝑠 𝑋 𝑟 𝐹𝑦 (8)
𝑡 𝑋𝐸,𝑠𝑠 𝐸
The variables contained in equation (8) are summarized in
Table VII.
TABLE VII
PARAMETERS OF RESISTANCE TO BUCKLING AT THE BASE OF A NREL 5-MW
TURBINE TOWER [44]. LN=LOG-NORMAL DISTRIBUTION, COV=COEFFICIENT
OF VARIANCE.
Distribution Expected
Variable Description COV
type value
D 𝑎𝑠𝑒 Tower diameter (base) - 6.5m -
D 𝑜𝑝 Tower diameter (top) - 3.87m - (a) (b)
t Tower base thickness - 0.027m -
E Young’s modulus LN 0.02
MPa
F𝑦 yield stress LN 240MPa 0.05
X𝑦,𝑠𝑠 Model uncertainties due to LN 1 0.05
scale effects: yield stress
X𝐸,𝑠𝑠 Model uncertainties due to LN 1 0.02
scale effects: Young’s
modulus
X𝑟 Critical load capacity LN 1 0.10
The resistance of the mooring line (𝑟 ), on the other hand,
can be characterized by a Weibull extreme distribution [14, 15,
55]. The arithmetic expectation of the breaking strength is
empirically estimated according to nominal diameters of (c)
Fig. 8 Mean of the simulated extreme wind and wave fields. (a) Extreme
mooring lines while its coefficient of variation can be assumed wind speeds at 90m (unite: m/s). (b) Extreme significant wave height (unite:
as [56]. m). (c) Extreme spectral peak period (unite: s).
In summary, the resistances of 𝑟 , 𝑟 and 𝑟 are modelled From Fig. 7, it is found that the maximum failure
as random variables whose statistics are explicitly provided. probability, which is around 0.6, has been found in Taiwan
Given these stochastic resistance models and the ultimate Strait where the maximum value of extreme wind speeds and
loads, the failure probability of the OC4-DeepCwind wave heights are also observed as indicated in Fig. 8. The
semisubmersible can be estimated according to equation (9) failure probability in the South China Sea and the East China
as, Sea is around 0.3-0.45. The mean of the extreme wind speeds
[𝐺(𝑟 , 𝑆 ) | 𝐺(𝑟 , ) | 𝐺(𝑟 , 𝑇) ] (9) and wave heights in such areas are found less than in Taiwan
Equation (9) implies that the offshore wind turbine fails if one Strait by ~12.50%, which implies that the decreases in
of the three critical components (blade, tower and mooring extreme wind speeds and wave heights and the reductions in
line) fails. Based on the simulated loads and modelled the failure probability are not linearly related. In other words,
resistances, the failure probability of the OC4-DeepCwind in the spatial variations in failure probabilities can not be reliably
the China coastal waters is calculated and presented in Fig. 7. estimated based solely on the extreme wind and wave fields,
and the interactions between winds/waves and floating
foundations are important in the estimation of failure
probabilities.

C. Point failure probability

With the purpose of further discussing the failure
probability of the OC4-DeepCwind floating wind turbine, 4
typical locations in different sea areas are selected where the
ultimate loads and resistances in different parts of the floating
wind turbines are discussed in details. The longitudes,
latitudes and other relevant information on the selected
Fig. 7 Failure probability of the OC4-DeepCwind semisubmersible wind locations are listed in Table VIII.
turbine in the China coastal waters. TABLE VIII
COORDINATES OF THE SELECTED 4 TYPICAL LOCATIONS IN CHINA COASTAL
WATERS.
Water Distance off
Longitude Latitude
Location depth the coast Sea areas
° °
(m) (km)
A 119.45 24.92 34.23 32.34 Taiwan Strait 1
B 68.66 115.82 Taiwan Strait 2
C 22.02 243.25 Bohai Gulf

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8

D 150.74 217.29 The South FAILURE PROBABILITY FOR OFFSHORE WIND TURBINES AT SELECTED 4
China Sea LOCATIONS. (𝑥10 IS THE MEAN OF THE EXTREME WIND SPEED SERIES AT 10M.
Fig. 9 shows box-whisker plots illustrating the dispersals in 𝑠, AND ARE THE FAILURE PROBABILITY FOR 𝑆 , AND 𝑇. P IS THE
COMPREHENSIVE FAILURE PROBABILITY).
our dataset concerning the ultimate loads and the resistances
Index A B C D Resistance
of blade root, tower base and mooring line at the selected 𝑥10 (m/s) 14.73 14.31 11.80 14.27 -
locations. 𝐻𝑠 (m) 9.48 7.81 5.27 6.93 -
𝑇𝑝 (s) 16.23 17.15 12.06 14.71 -
𝑆 (K / 2 ) 6.34 5.84 5.30 5.81 5.18
(K ) 3.28 2.46 1.46 2.09 2.08
𝑇 (K ) 9.82 5.31 4.24 3.64 4.95
0.53 0.34 0.16 0.29 -
0.52 0.36 0.16 0.30 -
0.52 0.35 0.21 0.29 -
P 0.55 0.37 0.23 0.32 -
It is evident that the structural failure at the locations A, B
(a) (b) and D could be resulted from the failures in any of the three
critical parts or their combination. The failure at the location C,
on the other hand, is dominated by the mooring line break due
to the fact that is much larger than and . From
Table IX, it is found that even though the mean of extreme
wind speed series at the locations A, B and D are similar, not
only the failure probabilities of the critical parts ( 𝑠 , and
) but also the comprehensive failure probability ( ) at
(c) (d) location A is much larger than at the other two locations. One
plausible explanation might be that the probability density
distributions of the ultimate loads are different at the location
A and at the locations of B and D in spite of the similar mean,
as shown in Fig. 10.

(e) (f)
Fig. 9 Box-whisker plot of wind speed series at 10m 𝑥, significant wave
height 𝐻𝑠 , spectral peak period 𝑇𝑝 , flapwise bending stress at the blade root
S , tower base bending moment , the mooring line tensions 𝑇 and the
resistances (𝑟 , 𝑟 and 𝑟 ). The dashed line represents the mean of the
resistances.
(a) (b)
The dispersals, indicated by the Inter Quartile Range (IQR)
of the box-whisker plot shown in Fig. 9, of the ultimate loads
at the location C (Bohai Gulf) is less than other locations.
Such an observation indicates that the sea state in Bohai Gulf
is less influenced by typhoons and hence is stable. In
particular, the high upper quartiles (Q3) corresponding to the
locations A, B and D imply that the extreme wind and wave
loads in the Taiwan Strait and South China Sea could be
substantially higher than the mean values shown in Fig. 8, and
consequently corresponds to a higher failure probability. (c) (d)
Given that the IQR of the resistances of 𝑟 , 𝑟 and 𝑟 are Fig. 10 Probability densities of 𝑆 and 𝑟 at locations A-D (Unite:
included in Fig. 9, the failure of the turbine can be illustrated KN/ 2 ). (a) Location A. (b) Location B. (c) Location C. (d) Location D.
through comparing the boxes of resistances to the boxes of From Fig. 10, it is evident that the probability density
loads. For example, the load boxes of and 𝑇 distributions of the ultimate loads are bimodal, which are not
corresponding to the location A are found higher than the revealed by the mean presented in Table IX. The first peaks
resistances boxes (2.08 KN m for 𝑟 and 4.95 KN are found close to the concentrations of the probability
for 𝑟 ), which implies that the structural failure can be largely distribution corresponding to 𝑟 . Consequently, the second
attributed to the failure of the tower base and the mooring peaks contribute to the failure of the turbine because the
lines. In addition to the box-whisker plots of the data, all the chance for the resistance (𝑟 ) exceeding the second peak is
failure probabilities of the offshore wind turbines and the slim. In addition, Fig.10 implies that the turbine blade at the
associated wind speeds, significant wave heights, spectral location A has higher chance to fail as the second peak in Fig.
peak periods and the simulated ultimate loads are listed in 10(a) is broader than in Fig. 10(b-d). Other ultimate load
Table IX. probability density distributions, such as the and T,
TABLE IX

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9

sharing similar variation patterns are presented in Fig. 11 for Moreover, the failure probability in the South China Sea and
references. the East China Sea is around 0.3-0.45 due to the influence of
the numerous typhoons. Although the failure probability is
connected to the intensity of the extreme wind and wave fields,
the assessment of the failure of the offshore wind turbine
relies heavily on the interactions between the wind turbine and
the extreme wind/wave environment. Such a conclusion is
substantiated by the point failure probability predictions. More
specifically, the bimodal probability density distributions of
the ultimate loads reveal that ultimate loads, and hence the
(a) (b) failure probability, are not in a simple linear relationship with
the extreme wind speeds and wave heights/periods.

REFERENCES
[1] Li W, Yao H, Wang H, Wang Z. Latest development status of offshore
wind power in China—The perspective of developers. Journal of Renewable
& Sustainable Energy. 2014;6:786-803.
[2] Hong L, Möller B. Feasibility study of China’s offshore wind target by
2020. Energy. 2012;48:268-77.
[3] Liu Y, Li S, Chan PW, Chen D. Empirical correction ratio and scale factor
(c) (d) to project the extreme wind speed profile for offshore wind energy
Fig. 11 Probability densities of the ultimate loads and resistances at location exploitation. IEEE Transactions on Sustainable Energy. 2017;PP:1.
B and D. (a) and 𝑟 at location B (unite: KN ). (b) and 𝑟 at [4] Hong L, Möller B. Risks of tropical cyclones on offshore wind farms in
location D (unite: KN ). (c) 𝑇 and 𝑟 at location B (unite: KN). (d) 𝑇 China. 6th Dubrovnik Conference on Sustainable Development of Energy,
and 𝑟 at location D (unite: KN). Water and Environment Systems. 2011.
In general, the probability distributions of and 𝑇 are [5] Madsen HO, Krenk S, Lind NC. Methods of Structural Safety:
Prentice-Hall, Inc; 1986.
bimodal, which are similar with the distribution of 𝑆 , as [6] Ronold KO, Larsen GC. Reliability-based design of wind-turbine rotor
shown in Fig. 11. In addition, the first peaks of and 𝑇 blades against failure in ultimate loading. Engineering Structures.
are different from those of S in terms of showing the value 2000;22:565-74.
quite smaller than the resistances (𝑟 and 𝑟 ). Such finding [7] Toft HS, Sørensen JD. Stochastic Models for Strength of Wind Turbine
Blades using Tests. European Wind Energy Association. 2008.
indicates that the second peaks observed in the probability [8] Toft HS, Sørensen JD. Reliability-based design of wind turbine blades.
density distributions corresponding to the ultimate loads ( Structural Safety. 2011;33:333-42.
and 𝑇) determine, to a large extent, the failure probability of [9] Sørensen JD, Tarp-Johansen NJ. Reliability-based optimization and
the tower base and the mooring lines. The bimodal optimal reliability level of offshore wind turbines. International Journal of
Offshore & Polar Engineering. 2005;15:141-6.
distributions shown in Figs. 10 and 11 imply that the extreme [10] Rose S, Jaramillo P, Small MJ, Grossmann I, Apt J. Quantifying the
wind and wave loads simultaneously influence the mooring hurricane risk to offshore wind turbines. Proceedings of the National
line tensions and the buckling moments at the blade rotor and Academy of Sciences of the United States of America. 2012;109:3247-52.
tower base. The combination of the extreme winds and waves [11] Garciano LEO, Koike T. New Reference Wind Speed for Wind Turbines
in Typhoon-Prone Areas in the Philippines. Journal of Structural Engineering.
make the ultimate loads concentrated on two major probability 2010;136:463-7.
intervals. In other word, the bimodal probability distributions [12] Luo Y, Ahilan R. Mooring Safety Assessment Using Reliability
shown in Figs. 10 and 11 indicate that the ultimate loads in the Techniques. Moorings. 1991.
three critical parts of the floating wind turbine are not in a [13] Larsen K, Mathisen J. Reliability-based mooring design for a drilling
semisubmersible. Proc of ISOPE conference1996. p. 457-66.
linear relationship with the extreme winds and waves. [14] Vazquez-Hernandez AO, Ellwanger GB, Sagrilo LVS. Reliability-based
comparative study for mooring lines design criteria. Applied Ocean Research.
V. CONCLUSION REMARKS 2006;28:398-406.
[15] Montes-Iturrizaga R, Heredia-Zavoni E, Silva-González F. On the
In present study, a numerical framework is articulated, estimation of mooring line characteristic resistance for reliability analysis.
which couples the meteorology simulation, dynamics Applied Ocean Research. 2007;29:239-41.
simulation and damage model to predict the failure probability [16] Jiang D, Zhuang D, Huang Y, Wang J, Fu J. Evaluating the
spatio-temporal variation of China's offshore wind resources based on
of the OC4-DeepCwind offshore wind turbine in the China remotely sensed wind field data. Renewable & Sustainable Energy Reviews.
coastal waters. 2013;24:142-8.
Based on the simulation results, the failure probability of a [17] Jha A, Dolan D, Musial W, Smith C. On Hurricane Risk to Offshore
site-specific wind turbine is estimated by checking the Wind Turbines in US Waters. Offshore Technology Conference2010.
[18] Bayati I, Gueydon S, Belloli M. Study of the Effect of Water Depth on
maximum ultimate loads acting on the three critical parts of Potential Flow Solution of the OC4 Semisubmersible Floating Offshore Wind
the floating wind turbine against the resistances. Take the Turbine. Energy Procedia. 2015;80:168-76.
OC4-DeepCwind semisubmersible as an example, the [19] Grell GA, Peckham SE, Schmitz R, McKeen SA, Frost G, Skamarock
area-wide and point failure probability predictions are derived WC, et al. Fully coupled “online” chemistry within the WRF model.
Atmospheric Environment. 2005;39:6957-75.
from comparing the ultimate loads acting on the blade root, [20] Michalakes J, Dudhia J, Gill D, Klemp J, Skamarock W. Design of a
tower base and mooring line to the structural resistance. It is next-generation regional weather research and forecast model. Towards
found that the maximum failure probability (~0.6) occurs in teracomputing. 1998:117-24.
Taiwan Strait where the maximum values of extreme wind [21] Liu Y, Chen D, Li S, Chan PW. Revised power-law model to estimate the
vertical variations of extreme wind speeds in China coastal regions. Journal of
speeds and significant wave heights are also observed. Wind Engineering and Industrial Aerodynamics. 2018;173:227-40.

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 10

[22] Carvalho D, Rocha A, Gómez-Gesteira M. Ocean surface wind reanalyses: Comparison with observed data for Portugal. Applied Energy.
simulation forced by different reanalyses: Comparison with observed data 2014;117:116-26.
along the Iberian Peninsula coast. Ocean Modelling. 2012;56:31-42. [52] Coulling AJ, Goupee AJ, Robertson AN, Jonkman JM, Dagher HJ.
[23] Davis CA, Low-Nam S. The NCAR-AFWA Tropical Cyclone Bogussing Validation of a FAST semi-submersible floating wind turbine numerical model
Scheme. Nam. 2001. with DeepCwind test data. Journal of Renewable and Sustainable Energy.
[24] BouyéE, Durrleman V, Nikeghbali A, Riboulet G, Roncalli T. Copulas 2013;5:023116.
for Finance - A Reading Guide and Some Applications. Ssrn Electronic [53] Tran TT, Kim DH. Fully coupled aero-hydrodynamic analysis of a
Journal. 2000. semi-submersible FOWT using a dynamic fluid body interaction approach.
[25] Chang CM, Fang HM, Chen YW, Chuang SH. Discussion on the Renewable Energy. 2016;92:244-61.
Maximum Storm Radius Equations When Calculating Typhoon Waves. [54] Liu Y, Xiao Q, Incecik A, Peyrard C, Wan D. Establishing a fully coupled
Journal of Marine Science & Technology. 2015;23. CFD analysis tool for floating offshore wind turbines. Renewable Energy.
[26] Laprise R. The Euler equations of motion with hydrostatic pressure as an 2017;112:280-301.
independent variable. Monthly weather review. 1992;120:197-207. [55] Ang HS, Tang WH. Probability Concepts in Engineering Planning and
[27] Paulson CA. The mathematical representation of wind speed and Design, Vol. II. Decision, Risk, and Reliability: Wiley; 1984.
temperature profiles in the unstable atmospheric surface layer. Journal of [56] Kim H, Choung J, Jeon GY. Design of mooring lines of floating offshore
Applied Meteorology. 1970;9:857-61. wind turbine in Jeju offshore area. Journal of the Society of Naval Architects
[28] Skamarock WC, Klemp JB, Dudhia J, Gill DO, Barker DM, Wang W, et of Korea. 2014;9:V09ATA042-V09AT09A.
al. A Description of the Advanced Research WRF Version 3. 2008. p. 7--25.
[29] Hong S-Y, Pan H-L. Nonlocal boundary layer vertical diffusion in a
medium-range forecast model. Monthly weather review. 1996;124:2322-39.
[30] Dudhia J. A Multi-layer Soil Temperature Model for MM5. Psu/ncar
Mesoscale Model Users’ Workshop1996.
[31] Kain JS. The Kain-Fritsch convective parameterization: an update.
Journal of Applied Meteorology. 2004;43:170-81.
[32] Hong S-Y, Dudhia J, Chen S-H. A revised approach to ice microphysical
processes for the bulk parameterization of clouds and precipitation. Monthly
Weather Review. 2004;132:103-20.
[33] Mlawer EJ, Taubman SJ, Brown PD, Iacono MJ, Clough SA. Radiative
transfer for inhomogeneous atmospheres: RRTM, a validated correlated‐k
model for the longwave. Journal of Geophysical Research: Atmospheres
(1984–2012). 1997;102:16663-82.
[34] Pollard RT, Rhines PB, Thompson RORY. The deepening of the
wind-Mixed layer. Geophysical Fluid Dynamics. 1972;4:381-404.
[35] Mckinley S, Levine M. Cubic Spline Interpolation. Numermathjchinese
Univ. 1999;64:44-56.
[36] Li SH, Hong HP. Typhoon wind hazard estimation for China using an
empirical track model. Natural Hazards. 2016;82:1009-29.
[37] Fisher RA, Tippett LHC. Limiting forms of the frequency distribution of
the largest or smallest member of a sample. Mathematical Proceedings of the
Cambridge Philosophical Society. 1928;24:180-90.
[38] Cam LL. Maximum likelihood: An introduction. International Statistical
Review. 1990;58:153--71.
[39] Teng, Liu PC. Estimating Wave Height Distributions from Wind Speed
Distributions. Internationa Conference on Coastal Engineering2000.
[40] Isaacson MDSQ, Mackenzie NG. Long-term distributions of ocean
waves: a review. Journal of the Waterway Port Coastal & Ocean Division.
1981;107:93-109.
[41] Moan T, Gao Z, Ayala-Uraga E. Uncertainty of wave-induced response of
marine structures due to long-term variation of extratropical wave conditions.
Marine Structures. 2005;18:359-82.
[42] Myrhaug D, Fouques S. A joint distribution of significant wave height
and characteristic surf parameter. Coastal Engineering. 2010;57:948-52.
[43] Jonkman JM, Buhl ML. FAST User's Guide SciTech Connect: FAST
User's Guide. 2005.
[44] Sørensen JD, Toft HS. Probabilistic Design of Wind Turbines. Energies.
2010;3:241-57.
[45] Bayati I, Gueydon S, Belloli M. Study of the Effect of Water Depth on
Potential Flow Solution of the OC4 Semisubmersible Floating Offshore Wind
Turbine. Energy Procedia. 2015;80:168-76.
[46] Robertson A, Jonkman J, Masciola M, Song H, Goupee A, Coulling A, et
al. Definition of the Semisubmersible Floating System for Phase II of OC4.
Scitech Connect Definition of the Floating System for Phase IV of Oc3. 2014.
[47] Hasselmann K, Barnett T, Bouws E, Carlson H, Cartwright D, Enke K, et
al. Measurements of wind-wave growth and swell decay during the Joint
North Sea Wave Project (JONSWAP). Deutches Hydrographisches Institut;
1973.
[48] Hall M, Goupee A. Validation of a lumped-mass mooring line model with
DeepCwind semisubmersible model test data. Ocean Engineering.
2015;104:590-603.
[49] API. Recommended practice for design and analysis of stationkeeping
systems for floating structures. 1995.
[50] Wang C, Jin S. Error features and their possible causes in simulated
low-level winds by WRF at a wind farm. Wind Energy. 2013;17:1315–25.
[51] Carvalho D, Rocha A, Gómez-Gesteira M, Santos CS. WRF wind
simulation and wind energy production estimates forced by different

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TSTE.2018.2834471, IEEE
Transactions on Sustainable Energy
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 11

Yichao Liu received the B.S degree in

marine technology from Dalian Maritime
University, Dalian, China, 2014. He is
currently a Ph.D. candidate in the School of
Environment, Tsinghua University.
His research interests include the
offshore wind energy and the numerical
weather prediction.

Sunwei Li received the Ph.D degree in civil

and environmental engineering department
of the University of Western Ontario,
Canada, 2012. From 2012-2014, he was a
postdoctoral researcher in Hong Kong
University of Science and Technology. He
is currently an assistant professor in the
Department of Ocean Science and
Technology, Graduate School at Shenzhen,
Tsinghua University.
His research interests include the renewable energy
technologies, wind engineering and the numerical weather
prediction.

P.W. Chan received the Master degree in

Physics and mathematics from the
University of Hong Kong, Hong Kong,
China, 1994. He is currently a senior
scientific officer of the Hong Kong
Observatory.
He has been working on aviation weather
service for nearly 20 years. He has
contributed to the developments of the various low level wind
shear and turbulence alerting algorithms of the Hong Kong
International Airport. Recently he has been involved in a wake
vortex detection programme at the airport. He has published
various papers on meteorological applications.

Daoyi Chen received the Ph.D degree in

Hydraulics and River Dynamics from the
Tsinghua University, Beijing, China, 1988.
From 2005-2011, he was a chair professor
in the University of Liverpool. He is
currently an academic leader (full professor)
in the Department of the Ocean Science
and Technology, Graduate School at
Shenzhen, and the School of Environment, Tsinghua
University.
He is one of the founders of the shallow water
hydrodynamics and has published various papers on marine
renewable and sustainable energy and the marine science and
technology.

1949-3029 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.