Ocean Engineering 236 (2021) 109441

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Dynamic analysis of two-rotor wind turbine on spar-type floating platform

Omar El Beshbichi ∗, Yihan Xing, Muk Chen Ong
Department of Mechanical and Structural Engineering and Materials Science, University of Stavanger, Stavanger, Norway

ARTICLE INFO ABSTRACT

Keywords: The dynamic response of a two-rotor wind turbine mounted on a spar-type floating platform is studied. The
Floating offshore wind turbines response is compared against the baseline OC3 single-rotor design. Structural design shows how the two-rotor
Dynamic analysis design may lead to a mass saving of about 26% with respect to an equivalent single-rotor configuration.
Multi rotor wind turbines
Simulations predict significant platform yaw response of the two-rotor floating wind turbine — about 6 deg
Spar-buoy platform
standard deviation at the rated operating wind speed. It is shown how the platform yaw response is directly
Modelica
caused by the turbulence intensity at the hub coupled with the transversal distribution of thrust loads on
the structure. A coupled control strategy for the rotor-collective blade pitch controller is proposed, in which a
simple proportional control mitigating platform yaw motion is superimposed to the baseline OC3 PI controller.
Numerical simulations show how platform yaw response is reduced by about 60%, at the cost of mean power
loss at below-rated wind speeds of about 100 kW and maximum increase of the rotor-collective blade-pitch
angles standard deviation of about 2 deg. Parametric analysis of mooring lines design shows how an equivalent
mass density of the line of at least 190 kg/m is needed to avoid vertical loads at the anchors.

1. Introduction an array of smaller turbine units advantageous (Jamieson and Branney,

2012). Multi-rotor wind turbine concepts are also interesting from an
Offshore wind energy is a steadily growing industry, reaching in economical and logistical perspective, since small blades are easier to
2019 a total worldwide offshore wind power capacity of 30 GW out manufacture, transport, and deploy with respect to state of the art blade
of a total worldwide wind power capacity of 600 GW (GWEC, 2019). sizes. Vestas Wind Systems A/S installed a multi-rotor demonstrator
Offshore wind energy is appealing since wind speed is much greater at the Technical University of Denmark, named 4R-V29, composed
offshore than inland, and since many of the drawbacks of wind turbine of four 225kW wind turbines mounted on a single structure and in
deployment derives from interaction with populated areas (Cruz and operation between 2016 and 2019. van der Laan et al. (2019) recently
Atcheson, 2016). Ideal wind energy sites are mostly locations where compared numerical results obtained from several Reynolds-Averaged
water depth far exceeds 50 m, while bottom-fixed offshore wind tur- Navier–Stokes equations (RANS) tools against field measurements of
bines are economically feasible to be deployed only in shallow water power performance and wake deficit, showing faster wake recovery
depths (Jonkman, 2007). Floating wind turbines (FOWTs), able to be and marginally higher power output at below-rated environmental
deployed in deep waters, offer a technological solution, and may thus
conditions given by the rotors aerodynamic interaction. Bastankhah
help in the reduction of the overall levelized cost of energy (LCoE)
et al. (Bastankhah and Abkar, 2019) also performed a large-eddy simu-
associated with wind energy.
lation to study the wake flow properties of a similar four-rotor concept.
Reduction of the overall LCoE may in principle be achieved also
They found out that the wake recovery is faster at short downwind
by means of downscaling wind turbines into equivalent multi-rotor
distances with respect to a single-rotor system. The EU-funded project
systems (two or more wind turbines installed on the same structure).
InnWind (Jamieson et al., 2015) proposed a 20 MW configuration com-
The development of multi-rotor wind turbines is an old idea, first
developed early in the 20th century when the lack of advanced glass posed of 45 turbines 444 KW each. Conclusions claimed a reduction of
fiber composite materials made the manufacturing of large rotors un- the LCoE against an equivalent single-rotor configuration of about 15%.
feasible (Jamieson and Branney, 2012). The rated power of a wind Kirchner-Bossi and Porté-Agel (2020) showed that the optimization of
turbine is proportional to the net area swept by the rotor blades — the layout of multi-rotor wind farms may lead to significant benefits
it thus scales with the square of the rotor radius. However, the blade also in terms of overall power density compared to a baseline single-
mass generally increases with the cube of the rotor radius, thus making rotor wind farm layout. Research of multi-rotor concepts mounted on

∗ Corresponding author.
E-mail address: [email protected] (O. El Beshbichi).

https://doi.org/10.1016/j.oceaneng.2021.109441
Received 25 November 2020; Received in revised form 30 June 2021; Accepted 2 July 2021
Available online 13 July 2021
0029-8018/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

floating platforms is to date scant. First concepts date back to the time Table 1
of the earliest works in the field, such as the work of Heronemus (1972). NREL offshore 5-MW Baseline Wind Turbine Specifications (Jonkman, 2007).

Multiple Unit Floating Offshore Windfarm (MUFOW) (Barltrop, 1993) Rotor Diameter m 126
Hub Height m 90
was a UK based project started in 1993 aiming at the investigation of
Rotor Mass kg 110 × 103
the feasibility of arrays of wind turbines mounted on a single floating Nacelle Mass kg 240 × 103
platform. However, the idea has yet to be studied thoroughly and Cut-In, Rated, Cut-Out Wind Speed m/s 3, 11.4, 25
its feasibility has yet to be analyzed in detail. The overall dynamic Cut-In, Rated Rotor Speed rpm 6.9, 12.1
response of the floating system must be carefully studied, as well as
the aerodynamic interaction of the rotors under operative and extreme Table 2
Parameters used in simple tower structural design (Eurocode 3, 2006).
environmental conditions.
In this work, the dynamic response of a two-rotor wind turbine 𝜌𝑠𝑡𝑒𝑒𝑙 kg∕m3 8500
𝑆𝑦𝑖𝑒𝑙𝑑 (JIS SS400) MPa 230
mounted on a spar floating platform is studied. The study relies upon Wire Grade, 𝑅𝑟 MPa 1960
a reduced aerodynamic model, simplified yet adequate to get the Wire Fill factor, f – 0.8
overall dynamic characteristics of the multi-rotor FOWT concept. The Wire Spinning loss factor, k – 0.9
advantageous stability and relatively simple design and manufacturing Wire Self weight N∕mm3 830 × 10−7
of spar-type platforms made it one of the most studied designs over Safety factor, 𝛾 – 1.4

the years, and the abundance of reference designs makes it suitable to

be used in conceptual analyses. Phase IV OC3, for instance, is widely
used as a major reference design (Jonkman, 2010, 2009a). Full-scale concept. This means that the OC3 spar design might not be the most
deployment of floating wind turbines also utilized spar-type design, as optimal concept for the multi-rotor WT. However, the OC3 spar design
in Hywind Demo (Equinor, 2020a), the first full-scale prototype of a has been studied extensively by a large number of researchers over the
FOWT deployed in Norway in 2009, as well as in Hywind Scotland last decade. Using this concept in the present paper offers the advantage
(Equinor, 2020b), the first floating wind farm situated in Scotland and of greatly aiding result assessment.
commissioned in 2017. The hub height from the sea water level (SWL) is about 90 m, while
the horizontal distance between the hubs is about 138.6 m. The center
The analysis of this work relies on an in-house tool for the simpli-
of gravity (COG) of the OC3 system is about 78 m from the SWL, while
fied fully-coupled analysis of FOWTs concepts. The predictive tool is
it is about 100.9 m for the 2WT system. Standard NREL 5-MW wind
developed in Modelica, a non-proprietary, declarative, object-oriented
turbines are used in this study, whose main specifications are listed
language developed by the non-profit Modelica Association and em-
in Table 1. The horizontal rotor spacing is assumed to be 10% of the
ployed to conveniently model multi-domain systems (The Modelica
rotor diameter, inspired by the multi-rotor turbine concept installed
Association, 2020). The multi-rotor FOWT system is modeled as a rigid
at DTU Risø by Vestas A/S (Bastankhah and Abkar, 2019). The 2WT
body. A dummy degree of freedom (DOF) describing the simplified rigid tower is composed of a main vertical tapered cylinder, a secondary
rotor dynamics is added to the system equations of motion (EQM) in vertical cylinder mounted on top of it, and two horizontal tapered
order to determine the aerodynamic state of the system. The aerody- cylindrical arms supporting the rotor nacelle assemblies (RNAs). Wires
namic loads are then modeled as concentrated thrusts acting on the are used to connect the end of the horizontal arms to the top of the
rotor hubs and as concentrated torques acting on the rotor low-speed vertical structure. The inclination angle of the wires is 30 deg. This
shafts. The aerodynamic loads are computed by considering the relative structural geometry allows for distribution of the aerodynamic loads
velocity between the hub and wind transversal to the rotor plane and on the horizontal arms as bending loads, while the wires distribute
mapping the steady-state aerodynamic coefficients of the wind turbine. the static loads given by the arms self-weight and RNAs concentrated
This method is thought of as a simplified alternative to more complex weight at the hubs as compression loads on the main structure. The
beam-element/momentum (BEM) models, and previous work (El Besh- assessment of local buckling resistance is neglected at this stage.
bichi et al., 2021) showed how results obtained are accurate in terms In the context of the present work, a simple structural dimensioning
of overall dynamic response in operative environmental conditions. of the tower is carried out with the aim of defining first-attempt global
The method, however, presents major limitations when considering inertia properties of the FOWT system. The following study is thus
more complex dynamic interactions. The rotors flexibility and dynamic simplified and not intended to focus on detailed structural design.
contributions to the overall system dynamics are neglected. The aero- Table 2 shows the parameters used for the simplified structural design,
dynamic interaction between rotors, as well as the aerodynamic effects while Table 3 summarizes the selected tower geometry in terms of
induced by skewed flows are also not considered. cross-sectional dimensions of each tower sub-domain. Data are given
The present work is structured as follows. First, the two-rotor wind in terms of inner diameter, outer diameter, and related thickness, and
turbine concept (2WT) is presented. A simple structural study is carried are given both at the base and top of the sub-structure. Intermediate
out in order to define first-attempt tower dimensions and inertial cross-sections are defined through linear tapering. The structural com-
properties. Moreover, the spar platform design is carried out by setting putation assumes the maximum aerodynamic thrust acting at the hub,
forth general hydrostatic performance considerations. Next, the dy- i.e., 808kN for the NREL 5-MW wind turbine (Jonkman, 2007). The
namic response of the 2WT system is analyzed and contrasted with the wire tension and the reaction forces acting at the horizontal arms base
response of the reference OC3 floating wind turbine. The response of are readily computed by solving the associated statically indeterminate
the 2WT system is analyzed by means of two different rotor-collective beam problem. The loaded cross-section area must be big enough
blade pitch control strategies: the baseline OC3 controller, and a cou- to prevent yield (𝑆𝑦𝑖𝑒𝑙𝑑 ). Yield assessment is performed at the most
stressed sections of the structure, i.e., at the tower base and at the
pled controller that incorporates mitigation of yaw response. Finally,
horizontal arm base. A safety factor (𝛾) of 1.4 is used to obtain a
remarks about mooring lines dimensioning applied to the 2WT system
conservative design. The yield assessment can be prescribed as follows:
are given.

4𝑟𝑜𝑢𝑡 𝑀 𝑆𝑦𝑖𝑒𝑙𝑑
2. Multi-rotor wind turbine concept ≤ (1)
𝜋(𝑟𝑜𝑢𝑡 4 − 𝑟𝑖𝑛 4 ) 𝛾
Fig. 1 shows the multi-rotor wind turbine concept defined in the where 𝑀 is the bending load acting at the base of the tower sub-
present work in relation to the standard OC3 design (Jonkman, 2007, domain, and 𝑟𝑜𝑢𝑡 and 𝑟𝑖𝑛 are the outer and inner radii at the base
2010). The concept is composed of a two-rotor wind turbine mounted cross-section, respectively. Maximum and minimum values of thickness
on a spar-buoy floating platform. As in any new technology develop- are considered, equal to 0.04 m and 0.001 m, respectively. The bending
ment, the multi-rotor FWT will eventually converge to the most optimal moment acting at the base of the horizontal arms is assumed given by

2
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 1. (a) OC3 geometry [m] (Jonkman, 2007, 2010). (b) 2WT configuration selected in the present study [m]. (* When only the floating platform is considered, the depth to
COG is 89.9 m for the OC3 platform, and 121.5 m for the 2WT configuration).

Table 3
Tower geometry selected in simple structural design, maximum loads and utilization ratio.
Vertical tower Horizontal arms Top cylinder Wire
Inner diameter (base) m 6.22 4.732 3.26 –
Outer diameter (base) m 6.28 4.79 3.29 –
Thickness (base) m 0.03 0.028 0.014 –
Inner diameter (top) m 4.75 1.79 3.26 –
Outer diameter (top) m 4.79 1.77 3.29 –
Thickness (top) m 0.0215 0.014 0.014 –
Diameter mm – – – 107
Effective load, 𝑆𝑒 (bending/axial) MPa 135 94 – 1058
𝑆𝑒 ∕𝑆𝑦𝑖𝑒𝑙𝑑 – 0.58 0.40 – 0.54

the action of the aerodynamic thrust, the RNA weight concentrated at wire diameter can be estimated as follows (Eurocode 3, 2006):
the hub, and the horizontal arm self-weight. On the other hand, the √
bending moment acting at the base of the vertical tower is assumed 4𝑇 𝛾
𝑑𝑤𝑖𝑟𝑒 ≥ (2)
chiefly related to the action of the aerodynamic thrusts. Effective loads, 𝜋𝑓 𝑘𝑅𝑟
as well as the ratio between effective and yield stress, are listed in where 𝑇 is the wire tension, 𝑅𝑟 is the wire grade, 𝑓 is the wire fill
Table 3. The bending stress obtained is equivalent to about 94 MPa at factor, and 𝑘 is the spinning loss factor. The obtained wire tension is
the horizontal arms base and about 135 MPa at the vertical tower base, about 7730kN, while the obtained wire diameter is about 107 mm. The
following an utilization ratio close to 50%. The total concentrated mass wire concentrated mass is about 4.94 tonnes.
of the tower thus defined is about 536.9 tonnes. The wire dimensioning
is carried out in accordance with the recommended guidelines of the 3. Platform design criteria
standard EN1993-1-11 (based on EN 1990) (Eurocode 3, 2006). The
wire cross-section must be big enough to prevent yield under maximum The dimensioning process of a floating platform is mainly driven
external load. A value of 𝛾 equal to 1.4 is used also in this case. The by (1) the maximization of pitch stiffness in order to reduce maximum

3
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

static pitch angle, (2) the maximization of natural heave period in order considering surge–pitch coupling. The estimation is carried out
to reduce wave-induced motion, and (3) the reduction of the overall by solving the associated surge–pitch characteristic equation:
cost, chiefly driven by the platform mass and dimensions (Cruz and ( ([ ] [ ])
𝑚𝑡𝑜𝑡 𝑧𝑐𝑜𝑔 𝑚𝑡𝑜𝑡 𝐴 𝐴15
Atcheson, 2016). Fatigue criteria are also applied in later design stages −𝜔2 + 11
𝑧𝑐𝑜𝑔 𝑚𝑡𝑜𝑡 𝐼55 𝐴51 𝐴55
but may be neglected in conceptual design. The design space of a spar- ([ ] [ ]))
buoy platform is generally composed of the platform draft, the upper 0 0 𝐶𝑚,11 𝐶𝑚,15
+ + 𝜙=0 (6)
diameter 𝐷1 , and the lower diameter 𝐷2 as major design parameters. 0 𝐶55 𝐶𝑚,51 𝐶𝑚,55
Platform design parameters associated to the final system geometry where 𝜔 and 𝜙 are the eigenvalues and eigenvectors of the
adopted in this work are those depicted in Fig. 1. The distance between system, 𝑧𝑐𝑜𝑔 is the vertical location of the overall center of
the SWL and the top of the platform is equal to the value used in the gravity of the FOWT system, 𝐼55 is the overall pitch inertia, 𝐴11 ,
standard OC3 design, that is, 10 m. The distance between SWL and the 𝐴15 , 𝐴51 , and 𝐴55 are, respectively, the added mass in surge,
top of the lower spar section is also equal to the standard OC3 value, surge–pitch, pitch–surge, and pitch directions, and 𝐶𝑚,11 , 𝐶𝑚,15 ,
12 m. Spar thickness is assumed constant and equal to 0.05 m. The 𝐶𝑚,51 , and 𝐶𝑚,55 are, respectively, the mooring linear stiffness
platform design is carried out in terms of hydrostatic performance, and values in surge, surge–pitch, pitch–surge, and pitch directions.
simple computations can at this stage be employed in order to obtain The pitch period is simply computed as:
useful predictions. The following sections present a brief discussion of
2𝜋
the common criteria used in platform design. 𝑇55 = (7)
𝜔5
where 𝜔5 is the eigenvalue associated with the pitch direc-
3.1. Hydrostatic considerations
tion. The values used to estimate the added mass and mooring
lines linear stiffness are assumed in the design stage constant
The major criteria used in platform design can be described as
and equal to the value associated with the standard OC3 plat-
follows:
form (Jonkman, 2010). The mooring lines stiffness is relative to
1. The floating platform must achieve hydrostatic equilibrium. In an unstretched length of 902.2 m, a mass density of 77.71 kg/m,
order to enforce the constraint, the mass of the spar-buoy bal- a vertical static length of 250 m, and a diameter of 0.09 m. The
last is imposed by means of the following hydrostatic vertical platform draft affects the vertical static length of the mooring
equilibrium: lines, and consequently the mooring lines stiffness. Hence, a
difference is to be expected between the dynamic response of the
𝜌𝑤𝑎𝑡𝑒𝑟 𝑉 𝑤 𝑔 − 3𝐹𝑚,𝑣 − 𝑚𝑠 𝑔 − 𝑚𝑡,𝑛,𝑟 𝑔
𝑚𝑏𝑎𝑙𝑙𝑎𝑠𝑡 = (3) system and the hydrostatic results. The difference is assumed to
𝑔
be small in pitch and heave directions and therefore tolerable in
where 𝜌𝑤𝑎𝑡𝑒𝑟 is the water density, 𝑉 𝑤 is the water displaced a preliminary design context.
volume, 𝑔 is the acceleration due to gravity, 𝐹𝑚,𝑣 is the total
vertical mooring line static load, 𝑚𝑠 is the spar-buoy mass, and 3.1.1. Periods
𝑚𝑡,𝑛,𝑟 is the mass of tower, nacelles, and rotors. In the design Fig. 2 shows the variation of heave and pitch periods of the 2WT
process, the ballast center of gravity is assumed to be located at system as a function of the draft, lower diameter 𝐷2 , and higher diam-
5 m from the bottom of the spar-buoy platform. 𝐹𝑚,𝑣 is assumed eter 𝐷1 within 100 m<draft<170 m, 10 m<𝐷2 <12 m, and 𝐷1 either
equal to the vertical mooring line static load of the OC3 standard equal to 7.5 m (left-hand side plots) or equal to 10 m (right-hand side
design. plots). The red dot denotes the final spar design configuration selected
2. The maximum static pitch angle must be sufficiently small, in in the present work. As shown, the pitch period tends to increase in
order to avoid an excessive pitch dynamic response and to configurations with shorter draft and 𝐷2 lengths, while it does not vary
limit the loss of annual energy production (AEP) due to the substantially with variations of 𝐷1 . Values obtained indicate that only
skewed flow conditions (Cruz and Atcheson, 2016). According for unrealistic configurations given by extremely long draft and 𝐷2 the
to Zambrano et al. (2006), the maximum static pitch angle must pitch period becomes unacceptably short. The heave period tends to
not exceed 5 deg with ±15 deg of dynamic amplitude. The static increase in configurations with longer draft and 𝐷2 lengths and reduces
pitch angle can be estimated as follows (Pham and Shin, 2019): substantially in configurations with longer 𝐷1 lengths. Note that in the
case of 𝐷1 = 10 m, the heave period reduces to values close to 25 s for
𝐹𝑡ℎ𝑟𝑢𝑠𝑡 𝐻𝐵 feasible values of draft and 𝐷2 . Also for this reason, longer 𝐷1 lengths
𝜃5 = (4)
𝐶55 should be avoided.
where 𝜃5 is the static pitch angle, 𝐹𝑡ℎ𝑟𝑢𝑠𝑡 is the overall thrust force
acting at the hubs, 𝐻𝐵 is the vertical distance from the hubs to 3.1.2. Maximum static pitch angle
the center of buoyancy (COB) of the spar-buoy platform, and 𝐶55 Fig. 3 shows the maximum static pitch angle of the 2WT system as
is the hydrostatic restoring pitch stiffness, which can be derived a function of platform design parameters within the same value range
from metacentric height relationships (Faltinsen, 1993). used in Fig. 2. As it is clear, a greater maximum static pitch angle
3. The pitch and heave natural periods must be larger than 25–30 s is obtained for shorter lengths of draft and 𝐷2 . Longer lengths of 𝐷1
in order to avoid resonance motions with first-order wave ef- reduce to a minor extent the static pitch angle. The isoline relative to
fects (Bachynski and Moan, 2012). The heave period is estimated 4–5 deg should be considered as a threshold for acceptable platform
as follows: configurations. Note that the maximum static pitch angle is the most
√ stringent constraint in the design of multi-rotor floating wind turbines,
𝑚𝑡𝑜𝑡 + 𝐴33
𝑇33 = 2𝜋 (5) given that it eliminates most of the design space.
𝜌𝑤𝑎𝑡𝑒𝑟 𝑔𝐴𝑤
where 𝑚𝑡𝑜𝑡 is the overall FOWT mass, 𝐴33 is the added mass 3.1.3. Mass sensitivity
component in heave direction, and 𝐴𝑤 is the waterplane area. Fig. 4 shows the overall spar-buoy mass (including ballast) as a
The value used to estimate 𝐴33 is assumed in the design stage function of platform design parameters within the same value range
constant and equal to the value associated with the standard used in Fig. 2. The spar mass should always be minimized in order to
OC3 platform (Jonkman, 2010). The pitch period is estimated by reduce the construction, material, and deployment cost.

4
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 2. Pitch and Heave periods as a function of platform design parameters. Left plots are relative to an upper diameter 𝐷1 = 7.5 m. Right plots are relative to an upper diameter
𝐷1 = 10 m. The red dot denotes the design point selected in this work.

Fig. 3. Maximum static pitch angle as a function of platform design parameters (maximum thrust is assumed). Left plot is relative to an upper diameter 𝐷1 = 7.5 m. Right plot
is relative to an upper diameter 𝐷1 = 10 m. The red dot denotes the design point selected in this work.

5
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 4. Overall spar-buoy mass (including ballast) as a function of platform design parameters. Left plot is relative to an upper diameter 𝐷1 = 7.5 m. Right plot is relative to an
upper diameter 𝐷1 = 10 m. The red dot denotes the design point selected in this work.

3.2. Platform configuration Table 4

Geometrical–inertial–hydrostatic specifications for the platform configuration selected
in the present work and for the standard OC3 spar platform (Jonkman, 2007).
The selection process of a feasible platform configuration is based
OC3 2WT
on a trade-off among the design constraints previously mentioned.
𝐷1 m 6.5 7.6
The main design objective is to achieve acceptable hydrostatic per- 𝐷2 m 9.4 10.5
formance and to minimize the platform mass to be employed. Drafts draft m 120 140
longer than 140 m are unfeasible to deploy and are susceptible to Depth to COG m 89.92 121.5
fatigue loading, while higher 𝐷2 lengths generally lead to excessive Water Displacement m3 8 × 103 11.7 × 103
Mass (including ballast) kg 7.4 × 106 10.6 × 106
manufacturing and deployment costs (Cruz and Atcheson, 2016). From
Roll Moment of Inertia about COG kgm2 4.2 × 109 1.13 × 1010
design considerations several conclusions can be drawn. For a multi- Pitch Moment of Inertia about COG kgm2 4.2 × 109 1.13 × 1010
rotor FOWT employing multi-MW wind turbines, such as the 2WT Yaw Moment of Inertia about Centerline kgm2 1.6 × 108 1.7 × 108
concept considered in the present study, heave and pitch periods are Heave Hydrostatic restoring stiffness N/m 3.3 × 105 4.56 × 105
Roll Hydrostatic restoring stiffness Nm/rad 1.3 × 109 3.42 × 109
generally not a stringent constraint within the region of feasible designs
Pitch Hydrostatic restoring stiffness Nm/rad 1.3 × 109 3.42 × 109
in view of the greater inertia involved. For draft lengths shorter than
Mass (platform, ballast, tower, nacelle, rotor) kg 8.06 × 106 11.8 × 106
130–140 m the maximum static pitch angle exceeds the limit imposed
of 5 deg. This constraint is the most stringent of the design process.
Higher 𝐷1 lengths are slightly beneficial in terms of maximum static Table 5
pitch angle, but significantly reduce the heave period and increase the Mass saving between two OC3 units and a single 2WT concept.

platform mass of about 500 tonnes. In the present work, the platform 2-OC3 2WT Variation

configuration selected is characterized by a draft of 140 m, 𝐷1 = 7.5 m, Mass (platform, ballast, tower, nacelle, rotor) kg 16.12 × 106 11.8 × 106 −26.30%
and 𝐷2 = 10.5 m. Table 4 summarizes the geometrical, the inertial, and
the hydrostatic specifications of the configuration selected, together
with the specifications of the baseline OC3 spar-buoy platform. The place of 0.61 W/kg of the OC3-Hywind design. However, the power-
fairlead depth from SWL is assumed to be equally distanced from the to-weight ratio does not consider the overall power performance of
COG and the COB, as in the case of the baseline OC3 design. The the system but assumes rated power for each rotor as independent
fairleads depth from SWL is thus equal to 86.5 m. The ratio between systems. The aerodynamic performance of the 2WT design must be
the fairlead radius and 𝐷2 is set equal to the one used in the baseline carefully studied by considering all the major aerodynamic effects and
OC3 design. The fairlead radius from the centerline is thus equal to interactions before an accurate statement on power performance can
5.78 m. The overall concentrated mass of the 2WT system is about be drawn.
11.8 × 103 tonnes. The moments of inertia are computed by employing
a CAD model of the FOWT system. The RNA mass is assumed to be 4. Fully-coupled dynamic analysis
concentrated at the hub. The associated inertia tensor of the 2WT
system computed at the COG can be written as follows: In this work, the coupled dynamic behavior of the 2WT concept
is studied by employing an in-house code implemented by means of
⎡ 6.13 × 1010 −5.93 × 105 1.67 × 107 ⎤
the object-oriented language Modelica. The approach allows for easy
[𝐼],2𝑊 𝑇 = ⎢−5.93 × 105 5.77 × 1010 1.47 × 105 ⎥ kgm2 (8)
⎢ ⎥ implementation of arbitrary platform geometries and platform/rotor
⎣ 1.67 × 107 1.47 × 105 3.69 × 109 ⎦
configurations. Previous benchmarking work (El Beshbichi et al., 2021)
Table 5 shows the relative variation of mass between two standard based on the International Energy Agency (IEA) code-to-code compar-
OC3 wind turbine units and an equivalent 2WT concept. Even without isons for the baseline OC3 design (Jonkman et al., 2010) has also
performing structural optimization, the present 2WT design brings shown how this method gives good agreement to well established
about a significant mass saving of about 26.3% with respect to an dynamic codes in situations where rotor dynamic contribution can be
equivalent single-rotor configuration. Equivalently, the rated power- neglected. The system is assumed to be rigid. The hydrodynamic added
to-weight ratio associated to the 2WT concept is about 0.85 W/kg, in damping values employed are assumed equal to those associated with

6
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Table 6 spar-buoy platform, based on the OC3 design. 𝐻𝐹 ,1,2,3 are the horizontal
Load cases (LCs) (Bachynski and Moan, 2012). mooring loads acting at the fairleads. Even though the delta catenary
Case 1 2 3 mooring lines used in the OC3-Hywind platform are not subject to
Significant wave height, 𝐻𝑠 m 2.5 3.1 4.4 significant yaw moments, their designs allow for yaw stiffness, which
Peak wave period, 𝑇𝑝 s 9.8 10.1 10.6 may be easily increased by increasing the fairlead length. A quasi-static
Mean wind speed at hub, 𝑈 m/s 8 11.4 18
Turbulence intensity at hub, 𝐼 – 0.20 0.17 0.15
formulation of the mooring lines loads is employed in order to obtain
the loads–displacements relationship at the fairleads (Jonkman, 2007).
The effect of the mooring lines design on the yaw stiffness is neglected,
and a constant equivalent yaw stiffness of 9.8 × 107 Nm/rad is used
throughout the results to account for the effect of the delta lines. More-
over, the OC3-Hywind load–displacement relationship in the platform’s
yaw direction is found to be linear at least up to a platform’s yaw
angle of about 20 deg (Jonkman, 2010). A linear stiffness relationship
can then be used also where the platform’s yaw motion is considered
significant. The mooring lines mass density used is set in this work
equal to 200 kg/m.

4.4. Aerodynamic modeling

In this work, the aerodynamic loads are computed by mapping the

steady-state thrust and torque aerodynamic coefficients of each rotor.
This approach is different to standard codes where a full BEM method
Fig. 5. Mooring lines schematization (top-view) (El Beshbichi et al., 2021). is employed, and to simplified codes where the aerodynamic thrust is
simply computed as function of the wind speed (Karimirad and Moan,
2012). Integrated loads are used, and hence the distribution of the local
aerodynamic loads and the associated local moments on the blades is
the baseline OC3 design. Albeit yaw mooring stiffness is associated with
neglected (Karimirad and Moan, 2012). A simplified rigid rotor EQM is
the mooring lines design, in this work its value is assumed constant
considered to emulate the rotors aerodynamic response. Aerodynamic
and equal to the one characterizing the baseline OC3 design. The rotor
concentrated thrusts are applied at the hubs, while aerodynamic con-
inertial effects, including the contribution of gyroscopic effects on the
centrated torques are applied at the equivalent low-speed rotor shafts.
system dynamics, are not considered.
The relative velocity between hub and wind speed transversal to the
rotor plane is used. The thrust loads are computed as (El Beshbichi
4.1. Environmental conditions
et al., 2021):
1
Three specific load cases (LCs) are considered in this work. Each 𝐹 = 𝜌 𝐶 (𝜆, 𝛽)𝐴𝑈𝑟𝑒𝑙 2 (9)
case is characterized by directionally congruent turbulent wind and 2 𝑎𝑖𝑟 𝑡
irregular waves based on standard JONSWAP spectra. Table 6 sum- where 𝜌𝑎𝑖𝑟 is the air density, 𝐶𝑡 is the steady-state thrust coefficient, 𝜆
marizes the characteristics of the LCs selected. The cases considered is the tip speed ratio, 𝛽 is the rotor-collective blade pitch angle, 𝐴 is
are relative to different environmental severity regions. The first case the rotor plane area, and 𝑈𝑟𝑒𝑙 is the relative speed between local wind
(LC1) is relative to a below-rated operational wind speed, the second and hub. The torque loads are computed as:
case (LC2) is relative to the rated operational wind speed, and the last 1
𝑇 = 𝜌 𝑅𝐶 (𝜆, 𝛽)𝐴𝑈𝑟𝑒𝑙 2 (10)
case (LC3) is relative to above-rated operational wind speed. Wave 2 𝑎𝑖𝑟 𝑞
energy content increases accordingly. The turbulence intensity at the where 𝑅 is the rotor radius, and 𝐶𝑞 is the steady-state torque coefficient.
hub follows the Kaimal spectrum with IEC Class B normal turbulence The three-dimensional wind velocity profiles at the hubs are computed
model (NTM), based on the standard IEC 61400-1 (IEC, 2005). in Jonkman (2009b) and imported in the code. A 15X15 grid is
used, wide enough to encompass the rotors space, and a time step of
4.2. Hydrodynamic modeling 0.05 s is used to generate the wind profiles. Wind time histories of
4000 s are computed for each environmental condition. This method,
The hydrodynamic loads are computed from linear (Airy) wave although generating accurate overall dynamic predictions in opera-
theory. Airy theory can be applied if the water depth is sufficiently tional environmental conditions, is not able to detect more complex
deep (DNV-GL, 2010). The numerical-panel code Sesam-Wadam (DNV- aerodynamic effects. For instance, the wind profile time realizations
GL) (DNV-GL, 2020) is used to solve the frequency-domain linear used for the computation of the concentrated aerodynamic loads are
hydrodynamic problem relative to the selected platform geometry. associated only with the hub locations. This assumption neglects the
Time realizations of irregular wave loads are preventively computed spatial turbulence variation on the rotors swept area. The effect of
by means of Inverse Fourier Transformations (IFT). The additional the horizontal arms on the local wind induction factors is also not
load contribution associated with hydrodynamic viscous drag is not considered. Moreover, skewed blade aerodynamics is not considered,
considered in this work (𝐶𝑑 = 0). However, it should be noted that vis- as well as the aerodynamic interaction between the rotors. Albeit liter-
cous drag loads may be significant in extreme environmental situations ature on similar multi-rotor FOWTs is scant, research works in different
where much larger waves and current effects are considered (Zheng fields concerning similar applications, such as the study of thrust deficit
et al., 2020). induced by the aerodynamic interaction among rotors of Unmanned
Aerial Vehicles (UAVs) (Zhou et al., 2017; Alvarez and Ning, 2017),
4.3. Mooring lines modeling suggest that the effect of the rotor aerodynamic interaction on the
overall system response may not be significant enough to compromise
Mooring lines design is assumed in this work equal to the standard the general dynamic behavior obtained in the present study. At any
OC3 design, that is, three catenary mooring lines mounted at 120 rate, the significance of these effects on multi-rotor FOWT system
deg from one another (Jonkman, 2010). Fig. 5 shows a top-view dynamics can be quantified by employing corrected BEM aerodynamic
schematization of the standard mooring system employed in the 2WT capabilities in Modelica, which is a task currently under development.

7
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 6. Wind profiles at the OC3, 2WT left turbine, and 2WT right turbine hub locations relative to the same realization (LC1 — Kaimal turbulence spectrum IEC Class B NTM).

The tower base local reference frame is assumed placed at the same where 𝐾𝑃 ,𝑞6 is the constant gain to be determined from tuning analysis.
inertial location for all configurations. Fig. 6 shows the wind profiles Fig. 7 shows the block diagram representation of the coupled control
relative to the same realization and attributed to the hub locations for strategy. Yaw motion is positive if counterclockwise. When the left
the OC3 wind turbine, the 2WT left wind turbine (2WTL), and the 2WT rotor is considered, the yaw P control contribution is active only if yaw
right wind turbine (2WTR). The time histories are relative to a mean motion is negative. On the other hand, the right rotor yaw P control
wind speed of 8 m/s (LC1). contribution is active only if the yaw motion is positive. The absolute
value of the yaw control contribution is used to compute the final rotor-
4.5. Control modeling collective blade pitch angle induced by yaw dynamics. The standard
OC3 variable-speed generator-torque control, as well as saturations of
As it will be presented in detail in Section 5.4, the platform yaw the rotor-collective blade pitch angles and pitch rates are also enforced
motion response of the 2WT is particularly critical. For this reason, in both cases (Jonkman, 2007). While the PI control on the generator
the quantification of the feasibility of mitigating the platform yaw speed is only active if the generator rotational speed reaches rated
motion by means of an optimized rotor-collective blade pitch control values, the P control on yaw motion is always activated.
strategy is proposed. The pitch actuator dynamics is assumed fast Table 7 summarizes the control gains used in this work. As it will be
enough to be neglected for rigid dynamics analysis. That is, there is described in detail in Section 5.2, a constant gain for the yaw control
no delay between the reference pitch angle and the actual pitch angle. equal to 1.5 is selected. Fig. 8 shows a time history of about 500 s
A second-order low-pass filter can generally be used to represent pitch of the rotor-collective blade pitch angle dynamics and thrust forces
actuator dynamics. Common values for cut-off frequency and damping for both wind turbines of the 2WT concept with coupled control. The
ratio are about 1 Hz and 0.7, respectively — fast compared to the control dynamic response is associated to above-rated environmental
system dynamics (Dunne and Pao, 2016). Two rotor-collective blade conditions (LC3 - Table 6). The plot also shows the associated yaw
pitch control strategies are separately applied to the 2WT concept. The response. The effect of the linear coupling between the baseline OC3
controls employed are the following: PI control and the yaw P control is clearly noticeable, as well as the
associated influence on the rotor thrusts.
• OC3 baseline control. The standard OC3 PI rotor-collective blade
pitch control on the generator speed as described in Jonkman 5. Results
(2007) is used to independently control both rotors.
• Coupled control. The baseline OC3 baseline PI rotor-collective The integration method dassl is employed to solve the equations of
blade pitch control on the generator speed is linearly coupled with motion of the system, with a tolerance equal to 1 × 10−6 and a time
a proportional rotor-collective blade pitch control on the 2WT step equal to 0.1 s. A simulation time equal to 4000 s is carried out.
platform yaw motion. The first 400 s are discarded in order to let the initial transients of
The coupled control strategy proposed is designed to induce a re- the system die out. The effective time series used to compute results
duction of aerodynamic thrust at the hub whose surge motion brought are thus about 1-h long. Three systems are tested under the same
environmental conditions: the baseline OC3 design, the 2WT concept
about by the yaw dynamics is positive. The rotor-collective blade pitch
with OC3 baseline control, and the same 2WT system with coupled
angle, 𝛽𝑖 , where 𝑖 denotes either the left (L) or the right (R) wind
control as described in Section 4.5. Every system employs mooring lines
turbine, can be computed as:
with a mass density of 200 kg/m. The dynamic response results are
( ) 𝑡( ) given in terms of platform motion q, upstream fairlead tension 𝑇2 , and
𝛽𝑖 (𝑡) = 𝐾𝑃 𝛽𝑖 (𝑡) 𝜔𝑖,𝑔𝑒𝑛 (𝑡) − 𝜔𝑔𝑒𝑛,𝑟𝑒𝑓 + 𝐾𝐼 𝛽𝑖 (𝑡) 𝜔𝑖,𝑔𝑒𝑛 (𝑡) − 𝜔𝑔𝑒𝑛,𝑟𝑒𝑓 𝑑𝑡
∫0 electric power production 𝑃𝑒 , and are expressed in terms of overall
+ 𝐾𝑞6,𝑖 𝑞6 (𝑡)|𝑞6 (𝑡)| (11) mean values, 𝜇, and standard deviations, 𝜎.

where 𝐾𝑃 (𝛽) and 𝐾𝐼 (𝛽) are, respectively, the proportional and integral 5.1. System natural periods
gain-scheduling laws for the baseline OC3-Hywind PI control on the
generator speed, 𝜔𝑔𝑒𝑛 (𝑡) is the generator speed, 𝜔𝑔𝑒𝑛,𝑟𝑒𝑓 is the reference Table 8 summarizes the damped natural frequencies of the 2WT
(rated) generator speed, 𝑞6 (𝑡) is the platform yaw motion in radians, system as computed from free decay tests in Modelica. The results
and 𝐾𝑞6 (𝑞6 ) is the proportional gain-scheduling law for the P control obtained are compared against the damped natural frequencies of the
on the platform yaw motion, which can be expressed as: OC3 system when the mooring mass density is equivalent to either the
{ one used in the 2WT system (200 kg/m) or to the standard value found
⎧ 𝐾𝑞6,𝑖 = 0, if 𝑞6 > 0
⎪if 𝑖 = 𝐿 in the literature (77.7 kg/m) (Jonkman, 2007). The yaw stiffness of the
⎪ 𝐾𝑞6,𝑖 = 𝐾𝑃 ,𝑞6 , if 𝑞6 < 0
⎨ { (12) OC3 system is assumed constant in the two cases. The assumption is
⎪if 𝑖 = 𝑅 𝐾𝑞6,𝑖 = 𝐾𝑃 ,𝑞6 , if 𝑞6 > 0 reflected in a constant OC3 yaw period for different mooring charac-
⎪ 𝐾𝑞6,𝑖 = 0, if 𝑞6 < 0 teristics. Higher inertia involved in the 2WT system lead to significantly
⎩

8
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 7. Block diagram of the 2WT coupled control concept for left wind turbine (right wind turbine condition: 𝑞6 > 0).

Table 7
Control systems specifications (Jonkman, 2010).
Proportional Gain at Minimum Blade-Pitch Setting (Generator speed control) – 0.00627
Integral Gain at Minimum Blade-Pitch Setting (Generator speed control) – 0.00089
Generator Torque at Rated Speed Nm 43093
Proportional Gain (2WT yaw motion control) – 1.5

Fig. 8. (a) Rotor-collective blade pitch angles and platform yaw motion. (b) Aerodynamic thrusts at hub. Coupled control schedule for 2WT at above-rated environmental condition
(LC3).

Table 8 5.2. Effect of yaw control proportional gain tuning (coupled control)
2WT and OC3 damped natural periods, obtained from free decay tests in Modelica
(note that mooring line mass density and yaw stiffness used with 2WT system are 200
kg/m and 9.8 × 107 Nm/rad, respectively Jonkman, 2007). Generally, the tuning process of PI/PID wind turbine controllers is
2WT OC3 (200 kg/m) OC3 (77.7 kg/m) first obtained using methods such as pole-placement or Ziegler–Nichols
Surge s 132.3 103.1 123.45 and then refined by the employment of fully-coupled aeroelastic simu-
Heave s 32.1 30.9 31.8
lations to obtain an optimized tuning in terms of loads reduction and
Pitch s 29.7 28.2 28.5
Yaw s 33.6 8 8.19 motion regulation (Ziegler and Nichols, 1993; Mirzaei et al., 2016). On
this line, Hansen et al. (2005) determined the gains of a standard rotor-
collective blade pitch PI controller by employing a minimization of the
blade root flapwise bending moments. Tibaldi et al. (2012) performed
higher surge and yaw periods with respect to the OC3 system. As a fine-tuning of two PI controllers respectively associated with below-
expected from preliminary hydrostatic considerations, 2WT heave and rated and above-rated environmental conditions by minimizing a cost
pitch periods are for this configuration sufficiently longer than the limit function based on fatigue loads, ultimate loads, annual energy produc-
imposed to avoid first-order wave excitations. tion, and blade pitch actuator duty cycle. Control gain tuning is hence

9
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 9. Standard deviation of (a) platform yaw motion, (b) electric power output, (c) platform surge motion, (d) platform pitch motion, (e) left rotor-collective blade pitch angle,
and (f) right rotor-collective blade pitch angle of the 2WT system in relation to the yaw control proportional gain and under different operating wind speed (Table 6).

a trade-off analysis, necessary in order to identify the best tuning setup platform yaw motion is linearly superimposed to the OC3 baseline
under contrasting objectives. rotor-collective blade pitch PI control, as presented in Section 4.5.
The main goal of the current analysis is to quantify the capability The tuning process is only focused on the additional proportional gain
of a coupled control strategy to mitigate the platform yaw motion of 𝐾𝑝 , while the baseline PI controller retains the original OC3 gain-
the 2WT system. For this reason, a simple proportional control of the scheduling (Jonkman, 2007). The study of more advanced controllers,

10
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 10. Mean of (a) platform yaw motion, (b) electric power output, (c) platform surge motion, (d) platform pitch motion, (e) left rotor-collective blade pitch angle, and (f) right
rotor-collective blade pitch angle of the 2WT system in relation to the yaw control proportional gain and under different operating wind speed (Table 6).

as well as the study of more thorough tuning strategies being able to (2) the maximization of the mean electric power output, and (3)
further optimize the system response, are left as questions for further the minimization of the rotor-collective blade pitch angles standard
research. deviation, which can be correlated with the aerodynamic thrust, torque,
The tuning is performed heuristically, and the main drivers are the and the associated blade root loads standard deviation. Fig. 9 illustrates
following: (1) the minimization of platform yaw standard deviation, the standard deviations of platform yaw, surge, and pitch motions,

11
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 11. Mean surge, pitch, electric power production, and upstream fairlead tension (all configurations employ an equivalent mooring line mass density of 200 kg/m).

electric power output, and rotor-collective blade pitch angles of the motion is minimized at the cost of an electric power output loss of 100
2WT system in relation to 𝐾𝑝 for below-rated, rated, and above-rated kW for below-rated environmental conditions and the increase of rotor-
environmental conditions. The platform yaw motion standard deviation collective blade pitch angle standard deviation in the range of 0.5–2
reduces significantly for every environmental condition, reaching a deg.
minimum region at 𝐾𝑝 ≃ 2. For above-rated environmental conditions,
the yaw motion standard deviation increases significantly at high 𝐾𝑝 5.3. Dynamic response
values due to stalled-induced thrust fluctuations. Platform surge and
pitch standard deviations are not notably affected by 𝐾𝑝 . It can be Fig. 11 shows the overall mean values for surge and pitch motion,
noted how the rotor-collective blade pitch angles standard deviation electric power production, and upstream fairlead tension under differ-
is zero for 𝐾𝑝 = 0 at below-rated environmental conditions, given by ent load cases for the OC3 system, the 2WT system configured with
the lack of the PI controller contribution to the system response. The baseline OC3 control, and the 2WT system configured with the coupled
rotor-collective blade pitch angles standard deviation greatly increases control. Platform sway, heave, and roll motions are found to be small
with greater values of 𝐾𝑝 , showing values higher than 2 deg for 𝐾𝑝 ≥ for each system and thus are not depicted in the results. Changing the
2. Fig. 10 illustrates the mean values for the same response parameters rotor-collective blade pitch control strategy from the baseline OC3 to
used in Fig. 9. Platform surge and pitch mean motion reduces at high the coupled control strategy does not influence significantly the mean
𝐾𝑝 values for below-rated environmental conditions, caused by the response of the 2WT system. As it is clear from the figure, the mean
increased rotor-collective blade pitch mean angle and the associated surge values are significantly lower in the OC3 system with respect to
reduction of aerodynamic thrust. The aerodynamic torque is reduced in the 2WT concept, at each load case about twice the values obtained in
the same manner, leading to a reduction of the overall electric power the system. This is clearly associated with the doubling of aerodynamic
output. For 𝐾𝑝 = 4 the electric power output loss is about 288kW. As thrust force in the 2WT concept. The maximum surge response in the
listed in Table 7, the constant yaw control proportional gain selected 2WT concept is about 30 m and is obtained at the rated operating wind
for the present work is 1.5. Under this tuning schedule, platform yaw speed. The mean pitch values obtained in the OC3 system are higher

12
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 12. Standard deviations for surge, pitch, electric power production, and upstream fairlead tension (all configurations employ an equivalent mooring line mass density of
200 kg/m).

Fig. 13. (a) Time histories for platform yaw motion at the rated operating wind speed (LC2). (b) Standard deviations for platform yaw motion (all configurations employ an
equivalent mooring line mass density of 200 kg/m).

with respect to the 2WT concept. The difference is of about 1 deg at still available to further optimize the platform design. The mean electric
the rated operating wind speed. As the pitch angle is one of the floating power production is broken down into single rotor performance. Since
platform chief design drivers, it may be concluded that some margin is the mean values are computed from single 1-h realizations of turbulent

13
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Fig. 14. Standard deviation of platform yaw motion as function of turbulence intensity
(11.4 m/s - Kaimal IEC 61400-1 turbulence spectrum).

responses, the mean electric power production for all rotors at rated
and above-rated operating wind speed is lower than the rated power
of 5 MW (Clifton and Wagner, 2014). This is due to fluctuations of the
generator speed in the below-rated region. Moreover, the aerodynamic
model employed in this work does not allow for assessment of the con-
tribution of complex aerodynamic effects on the system performance
— especially with respect to aerodynamic efficiency and consequently
electric power production. Consequently, the mean electric production
obtained at below rated speed for the 2WT system with baseline OC3
control is similar to the one obtained in the OC3 system. The interaction
effect may be significant and is therefore left at this stage as a question
for further research.
Fig. 12 shows the standard deviations for platform surge and pitch
motion, electric power production, and upstream fairlead tension for
the same load cases and system configurations given in Fig. 11. The
pattern obtained is similar to the one characterizing the mean values.
Standard deviations obtained for the 2WT system under different con-
trol strategies are found to be similar. The electric power production
standard deviations associated with the 2WT rotors are found to differ
by about 100 kW, particularly at below rated wind speed. The differ-
Fig. 15. Effect of mooring line equivalent mass density on the upstream mooring line
ence is reasonably associated with the different wind velocity profiles seabed length (a) and on the platform surge motion (b). Coupled control schedule for
at the hub locations. Tension standard deviation is important when 2WT at the rated operating wind speed (LC2).
evaluating the probability of line slack and fatigue life. In order to avoid
slack conditions, the tension standard deviation must be sufficiently
smaller than its associated mean value (Bachynski and Moan, 2012). rated operating wind speed, while the associated standard deviation
The standard deviation of the upstream fairlead tension for the 2WT
is about 6.5 deg. The coupled control strategy proposed in this work,
system at the rated operating wind speed is significantly higher than in
the OC3 system. However, the tension standard deviation is relatively albeit simply implementing a proportional control on the platform yaw
small if compared with its associated mean value (ratio 𝜎∕𝜇 ≈ 0.07), motion, reduces the overall yaw response of about 60%. The maximum
implying that the current mooring layout may be sufficient to withstand yaw angle at the rated operating wind speed reduces to about 7.7
survival environmental conditions (Cheng et al., 2017). deg, while the associated standard deviation reduces to about 2.5 deg.
The thrust discrepancy at the hubs increases with turbulence intensity,
5.4. Yaw response
thus increasing the standard deviation of platform yaw motion. Fig. 14
Fig. 13 shows the time histories for the platform yaw motion at shows the standard deviation for platform yaw motion as a function of
the rated operating wind speed (LC2), and the standard deviations for the turbulence intensity I, computed at the rated operating wind speed
the environmental conditions used in the analysis (see Table 6). The (11.4 m/s). Platform yaw motion is nonlinearly correlated with the
transversal distribution of thrust loads given by the two-rotor configu- wind turbulence level. The significant reduction of the yaw standard
ration significantly affects yaw motion. The response is characterized
deviation in the case the system operates with a coupled control
by long periods. This can be correlated with the concentration of wind
turbulence energy in the low-frequency region (Li et al., 2019). The strategy endures at varying turbulence intensities. It is also clear how
maximum yaw angle in the 2WT concept operating with the baseline the standard deviation for platform yaw motion is zero when the wind
OC3 control strategy is about 21.6 deg and it is obtained at the profile is not turbulent (𝜎(𝑞6 )|𝐼=0 = 0).

14
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

5.5. Mooring configuration Acknowledgment

A simple parametric mooring line design of the 2WT system is This PhD project is financed by the Equinor Akademia Program at
performed in terms of the effect of the mooring lines equivalent mass the University of Stavanger, Norway.
density on the system response computed at rated operating wind speed
(LC2). The associated findings are illustrated in Fig. 15. The mooring References
lines diameter is equal to 0.09 m, and the unstretched mooring line
length is equal to 902.2 m. For an equivalent mass density of 170 kg/m Alvarez, E., Ning, A., 2017. Modeling multirotor aerodynamic interactions through the
the peak minimum seabed length is negative, i.e., no portion of the vortex particle method. AIAA 2019–2827.
Bachynski, E.E., Moan, T., 2012. Design considerations for tension leg platform wind
mooring line rests on the seabed and the anchor tension includes a
turbines. Mar. Struct. 29 (1), 89–114.
nonzero vertical component. In order to ensure excess mooring line Barltrop, N., 1993. Multiple unit floating offshore wind farm (MUFOW). Wind
length, an equivalent mass density higher than 190 kg/m is thus nec- Engineering 17 (4), 183–188.
essary. Peak platform surge motion decreases with increasing mooring Bastankhah, M., Abkar, M., 2019. Multirotor wind turbine wakes. Phys. Fluids 31 (8).
line weight. Cheng, Z., Madsen, H.A., Gao, Z., Moan, T., 2017. Effect of the number of blades on
the dynamics of floating straight-bladed vertical axis wind turbines. Renew. Energy
101, 1285–1298.
6. Conclusions
Clifton, A., Wagner, R., 2014. Accounting for the effect of turbulence on wind turbine
power curves. J. Phys. Conf. Ser. 524.
The dynamic analysis of a two-rotor wind turbine mounted on a Cruz, J., Atcheson, M., 2016. Floating Offshore Wind Energy - the Next Generation of
spar-type floating platform has been performed. A simple structural Wind Energy. Springer.
analysis showed how a mass saving of about 26% may be achieved by DNV-GL, 2010. Offshore standard DNV-OS-j101 - design of offshore wind turbine
structures.
employing the two-rotor configuration instead of an equivalent single-
DNV-GL, 2020. WADAM Official website.
rotor configuration. The numerical simulations showed an increased Dunne, F., Pao, L.Y., 2016. Optimal blade pitch control with realistic preview wind
low-frequency yaw response of the two-rotor system compared with measurements. Wind Energy 19 (12), 2153–2169.
the response of a single-rotor configuration, of about 6 deg standard El Beshbichi, O., Xing, Y., Ong, M.C., 2021. An object-oriented method for fully coupled
deviation at the rated operating wind speed. The yaw excitation is di- analysis of floating offshore wind turbines through mapping of aerodynamic
coefficients. Mar. Struct. 78.
rectly induced by the turbulence intensity at the hub and the transversal
Equinor, 2020a. Hywind demo. https://www.equinor.com/en/what-we-do/floating-
distribution of thrust loads on the structure. A rotor-collective blade wind/hywind-demo.html.
pitch angle coupled control has been proposed for the mitigation of yaw Equinor, 2020b. Hywind scotland. https://www.equinor.com/en/what-we-do/floating-
response. The numerical simulations showed a reduction of the yaw wind/hywind-scotland.html.
response of about 60% at the cost of a reduction of mean power output Eurocode 3, 2006. Design of steel structures - part 1-11: Design of structures with
tension components (EN 1993-1-11).
at below-rated wind speed of about 100 kW. In addition, parametric
Faltinsen, O.M., 1993. Sea Loads on Ships and Offshore Structures. In: Cambridge Ocean
analysis showed that an equivalent mass density of the mooring lines Technology Series.
of at least 190 kg/m is necessary in the 2WT in order to avoid vertical GWEC, 2019. Global Wind Report.
loads at the anchors. Stiffer mooring lines configurations in yaw direc- Hansen, M., Anca, D., Larsen, T., Stig, O., Sorensen, P., Fuglsang, P., 2005. Control
tion would also reduce platform yaw response. The analysis is based on design for a pitch-regulated, variable speed wind turbine. In: Risø-R-1500(EN).
in-house code developed in Modelica. The aerodynamic model is based Heronemus, W.E., 1972. Pollution-Free Energy from the Offshore Winds. Marine
on mapping the steady-state aerodynamic coefficients characteristic Technology Society.
IEC, 2005. International Electrotechincal Commission, Wind Turbines: Part 1: Design
of the wind turbines employed. The approach is therefore not able
Requirements. Tech. rep. IEC61400-1:2005(E).
to assess complex aerodynamic situations which may be significant.
Jamieson, P., Branney, M., 2012. Multi-rotors; a solution to 20 MW and beyond? Energy
Skewed effects on the aerodynamic efficiency of the blades, especially Procedia 24, 52–59.
related to the significant platform yaw angles, cannot be considered to Jamieson, P., Branney, M., Hart, K., Chaviaropoulos, P.K., Sieros, G., Voutsinas, S.,
date. Moreover, the aerodynamic interaction between the rotors is not Chasapogiannis, P., Prospathopoulos, J.M., 2015. Innovative turbine concepts -
assessed. multi-rotor system. In: INNWIND-EU Deliverable 1.33.
Subjects for future work include the study of the aforementioned Jonkman, J.M., 2007. Dynamics modeling and loads analysis of an offshore floating
aerodynamic effects, the study of more advanced strategies for multi- wind turbine. In: NREL/TP-500–41958.
Jonkman, J.M., 2009a. Dynamics of offshore floating wind turbines-model development
rotor floating wind turbines control, as well as aeroelastic effects and
and verification. Wind Energy 12 (5), 459–492.
tower structural dynamics interactions. Moreover, the aerodynamic Jonkman, B., 2009b. Turbsim user’s guide: version 1.50. In: NREL/TP-500–46198.
interaction between the rotors should be investigated and included in Jonkman, J.M., 2010. Definition of the floating system for phase IV of OC3. In:
the dynamic model. Finally, an optimal station-keeping configuration NREL/TP-500–47535.
able to maximize yaw stiffness, an optimal platform configuration for Jonkman, J.M., Larsen, T., Hansen, A., Nygaard, T., Maus, K., M.Karimirad, Gao, Z.,
multi-rotor applications, as well as fault effects of rotors and control Moan, T., Fylling, I., Nichols, J., Kohlmeier, M., Vergara, J., Merino, D., Shi, W.,
system, should be investigated. Park, H., 2010. Offshore code comparison collaboration within IEA wind task 23:
phase IV results regarding floating wind turbine modeling. In: European Wind
Energy Conference. EWEC.
CRediT authorship contribution statement
Karimirad, M., Moan, T., 2012. A simplified method for coupled analysis of floating
offshore wind turbines. Mar. Struct. 27 (1), 45–63.
Omar El Beshbichi: Conceptualization, Methodology, Software, Kirchner-Bossi, N., Porté-Agel, F., 2020. Multi-rotor wind farm layout optimization. J.
Validation, Formal analysis, Investigation, Writing - original draft, Phys. Conf. Ser. 1618.
Writing - review & editing, Visualization. Yihan Xing: Conceptualiza- van der Laan, M.P., Garcí a, N., Angelou, N., Pirrung, G., Ott, S., Sørensen, K., Xavier, J.,
tion, Methodology, Software, Formal analysis, Investigation, Writing - Neto, V., Larsen, G., Mikkelsen, T., Kelly, M., Sjöholm, M., Andersen, S., 2019.
review & editing, Supervision. Muk Chen Ong: Conceptualization, In- Power curve and wake analyses of the vestas multi-rotor demonstrator. Wind
vestigation, Writing - review & editing, Resources, Supervision, Project Energy Science 4, 251–271.
Li, L., Liu, Y., Gao, Y., 2019. Dynamic and structural performances of offshore floating
administration, Funding acquisition.
wind turbines in turbulent wind flow. Ocean Eng. 179, 92–103.
Mirzaei, M., Tibaldi, C., Hansen, M.H., 2016. PI Controller design of a wind tur-
Declaration of competing interest
bine: evaluation of the pole-placement method and tuning using constrained
optimization. J. Phys. Conf. Ser. 753.
The authors declare that they have no known competing finan- Pham, T.D., Shin, H., 2019. A new conceptual design and dynamic analysis of a
cial interests or personal relationships that could have appeared to spar-type offshore wind turbine combined with a moonpool. Energies 12 (19).
influence the work reported in this paper. The Modelica Association, 2020. Modelica official website - modelica language.

15
O. El Beshbichi et al. Ocean Engineering 236 (2021) 109441

Tibaldi, C., Hansen, M.H., Henriksen, L.C., 2012. Optimal tuning for a classical wind Zheng, Z., Chen, J., Liang, H., Zhao, Y., Shao, Y., 2020. Hydrodynamic responses of
turbine controller. J. Phys. Conf. Ser. 555. a 6 MW spar-type floating offshore wind turbine in regular waves and uniform
Zambrano, T., MacCready, T., Kiceniuk, T., Roddier, D.G., Cermelli, C.A., 2006. current. Fluids 5 (4–187).
Dynamic modeling of deepwater offshore wind turbine structures in gulf of mexico Zhou, W., Ning, Z., Li, H., Hu, H., 2017. An experimental investigation on rotor-to-rotor
storm conditions. In: Volume 1: Offshore Technology; Offshore Wind Energy; Ocean interactions of small UAV. AIAA 2017–3744.
Research Technology; LNG Specialty Symposium, Vol. 1. pp. 629–634. Ziegler, J.G., Nichols, N.B., 1993. Optimum settings for automatic controllers. J. Dyn.
Syst. Meas. Control 115 (2B), 220–222.

16