A Review of Design, Analysis and

Optimization Methodologies for Floating

Wind Turbine Support Structures
Adebayo Ojoa,*, Maurizio Collua, Andrea Coraddua,b
aDepartmentof Naval Architecture Ocean and Marine Engineering, University of Strathclyde, Glasgow G4
0LZ, UK; [email protected]; [email protected]; [email protected]
b Department of Maritime & Transport Technology, Delft University of Technology, 2628 CD Delft, The

Netherland; [email protected]
*Corresponding author: [email protected]

Electronic copy available at: https://ssrn.com/abstract=3936386

A Review of Design, Analysis and Optimization Methodologies
for Floating Offshore Wind Turbine Substructures
Adebayo Ojoa,*, Maurizio Collua, Andrea Coraddua,b
a Department of Naval Architecture Ocean and Marine Engineering, University of Strathclyde, Glasgow G4 0LZ, UK;
[email protected]; [email protected]; [email protected]
b Department of Maritime & Transport Technology, Delft University of Technology, 2628 CD Delft, The Netherland;

[email protected]
*Corresponding author: [email protected]

Abstract

The development of novel energy technologies to meet net zero carbon emission is imperative in the provision of solutions
to realize an increasing worldwide demand for renewable energy. Floating Offshore Wind Turbine (FOWT) is one of the
emerging technologies to exploit the vast wind resources available in deep waters within the offshore wind sector. However,
as a result of the complexity of a FOWT system, bringing FOWT technology up to speed requires a detailed understanding
of the different disciplines within the system and the relationship between the FOWT system and the dynamics of the marine
environment; hence, the need for Multidisciplinary Design Analysis and Optimization (MDAO) of a FOWT system.
This paper reviews the MDAO of FOWT support structures proposed in the literature. This review covers an overview of
floating offshore wind turbine substructures’ concepts, parameterization and the design and analysis approaches (time and
frequency domain). It also provides a comprehensive review of MDAO frameworks for FOWT support structures.
Regarding the optimization aspect, a review of some optimization algorithms used for floating offshore wind turbine
substructure is given, i.e., from the global search heuristic and meta-heuristics algorithms to the local search gradient-based
optimization algorithms.
This work further identifies the research gaps in MDAO for FOWT support structures. The main proposed future research
areas to address these gaps are: increasing design space richness by adopting more advanced parametrization techniques to
represent the platform geometry (and other characteristics) and utilize surrogate/meta models to replace the most
computationally expensive high-fidelity models needed, for quick sensitivity studies before detailed analyses on selected
models are conducted.

Keywords: FOWT; MDAO; parameterization; optimization algorithms; platform geometry

Electronic copy available at: https://ssrn.com/abstract=3936386

2

1. Introduction

The floating offshore wind turbine (FOWT) industry is not yet in the mature stage of development and
currently, the fixed bottom foundation/platform is the dominant technology in the offshore wind turbine (OWT)
sector [1]. However, the concept of floating offshore wind turbines was proposed by Heronemus as far back as
1972 [2]. In as much as Heronemus’ vision was dated back to 1972, it was in the mid-1990s that the FOWT
started becoming a widespread concept after which several configurations of floating support platforms are
being developed for OWTs and performance of the concepts tested by numerical and experimental methods [1,
3].
Three main floater concepts from the oil and gas industries have been adopted by the offshore wind sector:
Spar, Semisubmersible, and TLP (Tension Leg Platforms). In 1998, the conceptual design for “FLOAT” – a
Spar concept floating wind turbine - was presented by Tong [4]. The objective of the FLOAT concept is to
allow the economical generation of electricity from wind power in offshore locations with water depth as deep
as 100 meters to 300 meters. For the Semisubmersible [5] presented an analytical and numerical design tools
for evaluating the performance of semi-submersible floating wind turbines. Their focus was on the
determination of an optimum hull-form for the floating structure and on developing analysis tools for the
interaction of the motion in waves of the platform, with the turbine aerodynamic performance as well as the
blade and hub loads. Development of the TLP floaters for came later as reported in [3, 6]. Withee and
Sclavounos (2004) performed fully coupled time-domain simulations of the system responses for a 1.5 MW
wind turbine mounted on a TLP floater under wind and wave forces. They presented the simulation results for
surge free decay tests carried out to estimate the damping arising from the turbine rotor and the wave and viscous
damping arising from the buoy. They found that the two damping mechanisms were comparable in magnitude.
Since the early days of OWT floaters highlighted above, extensive systems engineering analyses have been
conducted in literature, and it was not until 2017 that the first commercial floating offshore wind farm went
operational [7]. However, with the world in urgent need to reduce the carbon emission footprint, to revert the
existing trend of global warming from the emission of greenhouse gases (high temperature, shrinking ice, sea
level rise and wild fires) and the need to reduce the levelized cost of electricity generated from wind, there have
been increasing interest in the floating foundation/support wind turbine system in recent years [3]. Also, as
offshore wind turbine installation frontiers gradually move into deeper waters with abundant and high-quality
wind resources, the need for FOWT system has become imperative as the reliable fixed support/monopile
foundation offshore wind turbines become very cost prohibitive in such environmental conditions (deep water
> 60m), as mentioned by [8], [9], and [10].
The support structure/platform for a FOWT system can account for circa 29.5% of the capex cost, while the
corresponding support structure/foundation of a fixed bottom wind turbine accounts for 13.5% of capex cost of
the system [11]; hence, the need for optimization of the substructure of FOWT, to provide efficient means of
reducing the costs, is deemed more urgent than for offshore, bottom-fixed wind turbines. A FOWT is an
engineering system with a multidisciplinary set of complex subsystems, as indicated in Figure 1. These kinds
of complex systems, in other industries, have successfully been optimized adopting a MDAO approach. MDAO
is a systems engineering methodology that uses numerical optimization techniques to design and analyze
multidisciplinary engineering systems like a FOWT system [12]. This tool is perfect for both present and future
design and analysis requirements for conducting or executing the optimization of various multidisciplinary
systems.
To execute an MDAO involves overcoming design and analysis challenges amongst which are design
parameterization, modelling techniques and exploration of design space [13]. Overcoming the challenges
requires an optimization framework.

Electronic copy available at: https://ssrn.com/abstract=3936386

 3

This review focuses on the FOWT substructure as defined in the International Electrotechnical
Commission’s technical specification [14]. Chapter 1 is an introduction to the scope of this paper, while Chapter
2 details the design and analysis approaches of FOWT support structures. Section 2.1 gives an overview of the
FOWT substructures, section 2.2 details the design and analysis within the MDAO acronym introducing
parameterization concept and frequency domain dynamic model, section 2.3 reviews the MDAO approaches
for FOWT system, detailing MDAO architectures/framework, section 2.4 reviews the MDAO for FOWT
substructures with section 2.4.1 providing an overview of MDAO optimizers for FOWT system while section
2.4.2 reviews work done on MDAO for FOWT system. Chapter 3 highlights the gaps in MDAO methodologies
when applied to the offshore wind turbine sector, and Chapter 4 presents the main conclusions.

Figure 1: Floating Offshore Wind Turbine System [15] (The Hywind floating wind turbine concept)

Electronic copy available at: https://ssrn.com/abstract=3936386

4

2. Design, Analysis and Optimization Approaches for Floating Offshore Wind Turbine Support
Structures

This section looks at the different design, analysis and optimization approaches for a FOWT floater/platform.
Section 2.1 details an overview of a FOWT support structure, Section 2.2 highlights the parameterization and
frequency domain dynamic model of FOWT’s substructure and its use for preliminary design as a precursor to
detailed design and analyses. Section 2.3 details the multidisciplinary design and analysis approaches for FOWT
highlighting the MDAO workflow, system scope and model fidelity and MDAO architecture /framework.
Section 2.4 provides an overview of optimizers and details the work or reviews on MDAO for FOWT system.

2.1. Floating Offshore Wind Turbine Support Structure Overview

Different designs of floating offshore platforms exist; however, based on the principle exploited to achieve
static stability, FOWT substructures can be classified under three classes, detailed in[16], [8] and highlighted
below:
 The ballast stabilized platform (Spar) – This category of platform relies mainly on heavy ballast mass
located at a deep draft, to ensure the platform’s center of mass is well below the center of buoyancy, in
order to produce a large restoring moment.
 The buoyancy stabilized platform (Semi-submersible/Barges) – This class of support structures uses the
water plane area to ensure stability of the platform. A large (second moment of) water plane area is suitable
to raise the metacenter of the platform above the center of mass to ensure platform stability.
 The mooring stabilized platform (TLP) – This category of FOWT platform uses taut vertical mooring lines
to ensure the stability of the buoyant platform.
Apart from the Spar, the Semisubmersible, and the TLP platforms mentioned above, new and unique
geometrically shaped platforms for the FOWT sector are being developed. Examples of these unique platform
designs are the IDEOL “damping pool” barge platforms, the TetraSpar floating concept, and the Hexafloat. The
Floatgen IDEOL barge concept is an altered barge design that uses a moonpool, also referred to as damping
pool system, for motion reduction [8]. The Hexafloat is a floating concept developed by Saipem. It is a
pendulum lightweight structure composed of a submersible floater made of tubular elements, a counter weight
connected to the floater with tendons, simple mooring lines with drag anchors, and a lazy wave dynamic cable
[17]. The TetraSpar floating concept was developed by Stiesdal Offshore Technologies A/S. This concept aim
to provide a low-cost FOWT platform that can be easily installed in any conditions; hence, contributing to low
cost of electricity in comparison to bottom-fixed OWT [18]. Depending on the site conditions, the TetraSpar
can be configured as a semisubmersible, as a Spar (pendulum configuration), or as a TLP.
For a detailed mathematical model of the inclining and restoring moment physics, see Borg et al., [19].
The three stability classifications described are also illustrated in Figure 2 a, b, c and d below, representing
the buoyancy/waterplane stabilized ballast stabilized, mooring line stiffness stabilized and buoyancy/waterplane
stabilized platforms respectively.

Electronic copy available at: https://ssrn.com/abstract=3936386

 5

Figure 2: Floating platform classification of OWT, left to right: Semisubmersible, Spar, TLP and Barge [20]

In their chapter, Collu et al [16] discussed the classifications based on existing codes and standards
verification societies i.e., Bureau Veritas [21], American Bureau for Shipping [22] and Det Norske Veritas [23].
The Bureau Veritas [21] adopts the classification criterion based on the ballast floating platform, buoyancy
floating platform, and tensioned stability platform classes. American Bureau for Shipping [22] adopts the
classification criterion based on the structural elements of the different floating platforms/support structure,
without defining the stabilising mechanism.
For the Det Norske Veritas offshore standard [23], the criteria are based on whether a structure is restrained
(displaced in the order of centimetres) of compliant (displaced in the order of meters or more).

2.2. Parameterization and Frequency Domain Dynamic Model

Efficient evaluation of a large number of FOWT designs require adequate parameterization to explore the
design space. The design parameterization should ideally cut across more floating platform classes and different
geometrical variables for optimization purposes. A detailed parametric study of a FOWT system is presented
in [24], where the optimization leads to the definition of the Pareto fronts for mean square acceleration of the
turbine against multiple cost drivers of the offshore structure (simply put – a tradeoff between performance and
cost). The cost drivers include displacement of the structure and total mooring line tension.
Another work on FOWT parameterization and optimization can be traced to Sclavounos et al [13]. In this
work, they presented a coupled dynamic analysis of floating wind turbines incorporating a parametric design
study of floating wind turbine concepts and mooring system. They presented a Pareto optimal design that has a
favorable combination of nacelle acceleration, mooring system tension, and displacement of the floating
substructure supporting a 5 MW wind turbine.
Their results show that, for a fully coupled dynamic analysis conducted including the wind turbine, the
floating substructure and the mooring system, considering both wind and sea state environmental conditions,
the Pareto optimal structures are generally either a narrow deep drafted spar or a shallow barge ballasted with
concrete.

Electronic copy available at: https://ssrn.com/abstract=3936386

6

As highlighted in the work of Zhang et al [25], successful hydrodynamic optimization of ship hull depends
on the geometric variation of hull planer forms. The parametric design of hull forms involves; specifying form
parameters, design of a set of longitudinal curves, parametric modelling of sections which forms body parts and
generating hull forms [25]. This curve parameterization technique has been successfully used in the design of
ship hulls, and can be implemented in the design of FOWT platforms. Another parameterization and
optimization work, carried out for oil & gas offshore structures, with an optimization methodology based on
linear analysis of wave-body interaction, has been done by Birk et al [26], and it was started over a decade
earlier as hydrodynamic shape optimization of large offshore structures by the same authors [27]. In [26], they
integrated a reliable hydrodynamic analysis run with the WAMIT program with a newly developed parametric
hull design methodology, which enables the automated generation of hull shapes without the need for user
interaction. The optimization algorithm is used to optimize the shapes with the hull responses.
For optimization purposes that entails large design space exploration, time domain dynamics evaluation
becomes computationally expensive and time consuming. The way around the computationally expensive time
domain dynamics evaluation issue is to conduct the dynamic analysis in the frequency domain. In spite of the
frequency domain analysis advantage of being computationally less expensive, it has its own limitations
amongst which are:
1. Frequency domain analysis is not suitable for non-linear dynamic systems. It only applies to linear
systems such that the system’s behavior is linearly related to its displacement, velocity and acceleration
[28].
2. Frequency domain analysis does not take into consideration the impulse response function – irradiated
waves that keep exciting the body due to memory of past motion of the body even when the body has
suddenly stopped [28]. This memory effect is effectively covered in the time domain analyses using the
Cummings equation [28].
These limitations are not deterrent to the use of frequency domain analysis technique to solve optimization
problems in comparison to computationally expensive time domain techniques. As highlighted in Section 2.4.2,
which reviews the multi-disciplinary design analysis and optimization of a floating offshore wind turbine
system, most of the analysis conducted for the research work reviewed are conducted using the frequency
domain analysis technique. However, verification of the optimal design can be done with the more accurate
non-linear time domain analysis technique, when the design space has been narrowed down.
Table 1 shows a list of reviews on optimization of FOWT system and the analysis domain used to solve the
optimization problem or form the objective functions of the optimization. The majority of the works shown in
Table 1 adopt the frequency domain approach.

Electronic copy available at: https://ssrn.com/abstract=3936386

 7

Table 1: Analysis domain overview for optimization of FOWT system

Work Analysis Reference

Type/Domain
Practical application of global optimization to the Frequency domain [29]
design of offshore structures
WINDOPT - An optimization tool for floating Frequency domain [30]
support structures for deep water wind turbines
Evolving Offshore Wind: A genetic algorithm- Frequency domain [31]
based support structure optimization framework for
floating wind turbines
A multi-objective design optimization for floating Frequency domain [32]
offshore wind turbine support structures
Integrated design optimization of spar floating wind Frequency domain [33]
turbines
Platform Optimization and Cost Analysis in a Frequency domain [34]
Floating Offshore Wind Farm
Optimization of floating wind turbine support Reduced order time [35]
structures using frequency domain analysis and domain and Frequency
analytical gradients domain
Development of a framework for wind turbine Frequency domain [36]
design and optimization

2.3. Multidisciplinary Design Analysis and Optimization Approaches for Floating Offshore Wind Turbine
System

MDAO is an engineering/research field that studies the use of numerical optimization techniques to design
engineering systems that involves multiple disciplines / subsystems or components [37]. It is a systematic design
and analysis process that deals with the interfacing between different components and disciplines within a
system. This review will look at how MDAO is applied to the FOWT substructure system (platform and
mooring/station keeping).
MDAO was initially developed in the aerospace industry as a result of strong influences between different
disciplines (aerodynamics and structural dynamics) that affect the performance of the aircraft [38]. MDAO went
to be further successfully applied in other industries amongst which are automotive, civil and naval engineering
[12]. The IES (International Energy Agency) Wind Task 37 [39] identified three important dimensions of an
MDAO simulation set-up or workflow amongst which are: model fidelity, size and scope of simulation, and
MDAO architecture.
Earlier examples of applications of MDAO to wind energy systems are conducted by [40-42]. As reported
by [38], each of these optimization studies result shows a system-wide reduction in the cost of energy from 2%
to 15% based on the sub-system optimization.
Crawford et al [40] incorporates the National Renewable Energy Laboratory (NREL) aeroelastic design
codes with a cost-scaling model based on linear, quadratic and cubic function of the rotor diameter from the
CAD geometry to influence the cost changes in the respective subsystems. The MDAO approach is the
sequential optimization of the turbine using the NREL aeroelastic codes, CAD software interface and the custom
cost of energy algorithm.

Electronic copy available at: https://ssrn.com/abstract=3936386

8

The MDAO approach for Bottasso et al [41] is on the design of a wind turbine blade focusing on the structural
and aerodynamic trade-offs in blade design taking into consideration, the total aero-servo-elastic effects on the
blade structure and the noise constraints. Dykes et al [38] describes work done by Bottasso et al[41] to use a
sequential MDAO approach that involves a comprehensive aero-servo-elastic, non-linear, finite-element-
method-based, multibody dynamics solver at a first level and a second level using a finite-element, cross-
sectional model of the blade to perform a section-wise load calculations to determine the blade weight and a
third level using macro parameters to optimize the overall objective of the annual energy production (AEP) to
weight ratio minimization.
The authors of [42] applied a multi-level MDAO approach to the system design utilizing two disciplines
(maximizing AEP and minimizing blade root moment) under a system level analysis and optimization. Their
work borrowed the NREL aeroelastic design codes and cost models and in addition to the distinct multi-level
approach, their work incorporated the use of the Kriging-based metamodels to replace higher fidelity models in
order to save computation time required for the optimization process.
An extensive review of approaches in the design optimization of wind turbine support structures and the
challenges associated with it is presented in [43]. In this work, they reviewed the different techniques of
optimizing wind turbine structures amongst which are optimization of wind turbine structures using static
analysis, optimization of wind turbine structures using frequency domain and time domain analyses. Further to
this, [43] reviewed Windopt - a well-known optimization tool used with the spar-type FOWTs. Windopt allows
for the designing of the spar buoy, mooring system and the power cable, using sequential quadratic
programming and a combination of commercial analysis tools. However, its limitation is that the wind turbine
rotor is only represented as a state-dependent drag coefficient/force acting in a single node at the top of the
tower. Also in their work, [43] made recommendations to the field of structural optimization and amongst their
recommendations are the use of gradient-based and gradient-free optimization which are largely in use today.
Other structural optimization recommendations made in their work are: modeling with a hierarchy of fidelities,
reduction of load cases and interfaces for efficient integrated design and exploration of probabilistic design.
In [38], Dykes et al researched MDAO works relevant to both wind turbines and wind farms. From their
work, they laid the foundation for MDAO workflow WISDEM (Wind Plant Integrated System Design and
Engineering Model). In their report, [38] observed that most researches are conducted on singular components
or disciplines and concluded there are large opportunities for MDAO research and development in the wind
energy sector (offshore/onshore).
Ashuri et al [44] conducted a research on design optimization, capable of simultaneous designs of wind
turbine blade and tower subject to constraints on fatigue, stresses, deflections and frequencies with the Levelised
Cost of Energy (LCOE) as the objective function. From their experiment, the results show an improvement in
the quality of the design process with a realistic assessment of the LCOE and constraints, while preserving the
coupling of the components and disciplines by using the power of numerical optimization. Since then,
researchers like [45], [31] and [33] have been able to demonstrate the effectiveness of using numerical
optimization algorithms in MDAO for the design of FOWT substructures, and these research works are
highlighted with more details in section 2.4.2.

2.3.1. MDAO Workflow

MDAO comprises of a workflow with a set of computational tools (analysis block) that represents different
components and disciplines coupled together to simulate an entire system [46]. With this technique, drivers can
be included to control how and when each tool can be executed. The functionality of the workflow is defined
as a use case which describes any domain problem that can be solved by MDAO i.e., optimization of the
objective function. A simplified diagram of an MDAO is shown in Figure 3

Electronic copy available at: https://ssrn.com/abstract=3936386

 9

Figure 3: Simplified diagram of an MDAO workflow comprising an analysis block of two modules and a driver

The driver (numerical method governing the use case) integrating the modules in an MDAO workflow can
have different uses amongst which are, performing uncertainty quantification (UQ), running design of
experiments (DOE) or implementing optimization algorithms [46]. Optimization algorithms helps in finding
the optimal system design that maximizes system’s performance by exploring the design space smartly. More
on optimization algorithm is discussed in Section 2.4.
MDAO workflow consists of system scope, model fidelity and architecture/framework as highlighted in
sections 2.3.1.1 and 2.3.1.2.

2.3.1.1. System Scope and Model Fidelity

The scope of the system is clearly defined before instantiating the MDAO workflow because, not all
components or disciplines influence one another with the same intensity. Moreno et al [46] highlighted two
examples of use cases with different system scope in the field of wind energy. The examples are the optimization
of the layout of an offshore wind farm and the sensitivity analysis of LCOE with respect to foundation type.
For the optimization of the layout of an offshore wind farm, the workflow will have to include the calculation
of wake losses and cable lengths; however, for the latter, there is no need to re-analyze the performance or cost
of the electrical connection system as the interaction between them are negligible. The scope of an MDAO
example is shown in Figure 4.

Figure 4: Two workflows with different system scope. Dashed arrows include components/disciplines 1 and 2 while straight arrows
include 1, 2 and 3.

Electronic copy available at: https://ssrn.com/abstract=3936386

10

Model fidelity is very important in MDAO as it represents the degree to which a model or simulation
reproduces the state and behavior of a real-world object which helps to define the objective function within the
optimization problem. Different model fidelities or level of accuracy and sophistication of the integrated models
are available for the different disciplines in a FOWT system. Examples are: spreadsheet model, a simple beam
model or a full finite element (FEM) model with a higher precision or a computationally expensive
computational fluid dynamics (CFD) model.
In Systems engineering, model fidelity ranges from low fidelity (LF) model to the high-fidelity (HF) model
while the middle model between the low and high-fidelity models can be classed as a multi-fidelity surrogate
model. In [47], Shi et al shows that to take advantage of HF and LF models, multi-fidelity surrogate models
integrating information from both HF and LF models can be used, and are increasingly gaining popularity.
Examples of multi-fidelity surrogate models are Kriging based, radial basis function (RBF) and support vector
regression CO_SVR surrogate models [47]. These surrogate models used with optimization algorithms provides
competitive accuracy as HF models. Example of the multi-fidelity surrogate model where the Kriging based
example has been employed is highlighted in [32] discussed in Section 2.4.2.

2.3.1.2. MDAO Architecture/Framework

MDAO architecture/framework defines how the different models are coupled and how the overall
optimization problem is solved. Martins et al [37] highlights MDAO architecture as either monolithic or
distributed. In a monolithic architecture approach, the MDAO problem is solved as a single optimization
problem. A distributed approach solves the MDO problem using a set of optimization problems or subproblems.
MDAO architectures from the Monolithic approach are the simultaneous analysis and design (SAND),
multidisciplinary feasible (MDF) and individual discipline feasible (IDF) architectures. The difference between
these three architectures depends on the equality constraint group eliminated from the optimization problem. In
the SAND approach, the consistency constraint is eliminated from the optimization problem while for the IDF
approach, the disciplinary analysis constraint is eliminated from the optimization problem. MDF approach is
the most commonly used of the monolithic approach and both disciplinary analysis constraint and consistency
constraint are eliminated from the optimization problem. For further reading, a comprehensive details of other
monolithic and distributed MDAO architectures is presented in [37]. MDAO architecture can be executed by
developing powerful scripts to execute design and optimize a problem of interest or use commercial MDAO
packaged to provide solution to the problem of interest. The development of commercial MDAO frameworks
dates back to the late 1990s with iSIGHT [48]. Since the development of iSIGHT, several other commercial
frameworks have been developed amongst which are: Phoenix Integration’s ModelCenter/CenterLink, Esteco’s
model FRONTIER, TechnoSoft’s AML suite, Noesis Solutions’ Optimus and Vanderplaats’ VisualDOC [48].
Since the development of the highlighted frame works, MDAO framework has evolved. One of the recent
evolutions of optimization framework is the open-source, freely available OpenMDAO, with the capability of
gradient-based and metaheuristic optimization algorithm. OpenMDAO is an open-source multidisciplinary
design, analysis and optimization software for the exploration and exploitation of coupled multidisciplinary
system to determine the system’s global optimum design.
To develop MDAO architecture with powerful scripts will require an automated framework. Example of an
automated framework developed for wind turbine design optimization is highlighted in Leimeister et al [36].
There are two parts to the framework which are automation and automation plus optimization. The first part of
the framework (automated simulation) comprises of the modeling environment, simulation tool and the
programming framework. The holistic framework integrates an optimizer to the automated simulation
framework (automation plus optimization). An example of a holistic frame work with optimization
functionalities used for a FOWT optimization in [36] is highlighted in Figure 5. This type of framework with a
LCOE objective function for optimization can be found in [32].

Electronic copy available at: https://ssrn.com/abstract=3936386

 11

Modeling
Environment Programming Framework

Optimization
Simulation Tool Algorithm

Figure 5: Framework with Optimization functionalities for FOWT [36]

2.4. MDAO for FOWT Substructures

This section provides an overview of MDAO algorithms and optimizers and a background of the MDAO
work conducted for the FOWT substructure.

2.4.1. Overview of MDAO optimizers for FOWT substructures

The main objective of FOWT stakeholders is to minimize the cost of energy of wind turbines and increase
its reliability to compete and surpass fossil-fuel sources of energy. In [11] it is stated that the floating platform
accounts for about 29.5% of the total capital expenditure (CAPEX) of a FOWT system [11], and the fixed
bottom platform accounts for 13.5% of the total CAPEX [11] for a fixed offshore wind turbine system; hence,
a clever way of designing a floating substructure to minimize the cost will contribute to the reduction of CAPEX
for a FOWT system and subsequently, a reduction in the LCOE of the FOWT system. This clever approach to
design requires the need of optimization algorithms for selecting optimal solutions.
With multidisciplinary optimization algorithms, designers can identify the Pareto front / trade-off curve that
reveals the weaknesses, anomalies and rewards of a certain target [49], like minimizing the LCOE or improving
the performance metrics, such as the root mean square (RMS) of the nacelle acceleration. Optimization
algorithms are mainly categorized into two groups: Gradient Based (GB) optimization algorithm and Gradient
Free (GF) optimization algorithm.
GB methods are iterative methods that use gradient information of the objective function during iterations
[50]. They are efficient for finding local minima for high dimensional, non-linearly constrained convex
problems.
GF, also called Metaheuristic optimization algorithms, are usually characterized by a superior search
efficiency and robustness unlike GB that has the tendency of being stuck in local minima for optimization
problem with a multimodal objective function [33]. GF have been introduced to solve complex nonlinear
optimization problems that GB optimization methods cannot deal with [51].Once the optimization problem has
been defined, optimizers to execute the optimization algorithms must be selected to solve the optimization task.
A table of available optimizers is highlighted in Table 2. The optimizers are classed into Quasi-Newton method,
Sequential Quadratic Programming (SQP), Evolutionary Algorithm (EA), Particle Swarm Optimization (PSO)

Electronic copy available at: https://ssrn.com/abstract=3936386

12

and other types and grouped into the GB and GF optimization algorithms. Also highlighted in Table 2 are
optimizers with the capability of handling Multi-Objective (MO) functions.

Table 2: Overview of applicable optimizers

Class Optimizer GB GF MO Reference

Newton Conjugate Gradient (Newton-CG)  [52]
Quasi-Newton

Powell  [53]
Truncated Newton (TNC)  [36, 54]
Broyden-Fletcher-Goldfarb-Shanno (BFGS)  [36, 54]
Limited-memory BFGS with Box constraints (L-BFGS-B)  [36, 54]
Feasible SQP (FSQP)  [36, 54, 55]
SQP

Preconditioned SQP (PSQP)  [36, 54, 55]

Sequential Least Squares Quadratic Programming (SLSQP)  [36, 54]
Genetic Algorithm (GA)   [54, 56]
Non-dominated Sorting GA II (NSGAII)   [36, 54, 55, 57]
Non-dominated Sorting GA III (NSGAIII)   [36, 54, 55, 57]
Steady-state Epsilon-MO EA (EpsMOEA)   [36, 55, 57]
MO EA based on Decomposition (MOEAD)   [36, 55, 57]
  [36, 55, 57]
EA

Generalized Differential Evolution 3 (GDE3)

Strength Pareto EA 2 (SPEA2)   [36, 55, 57]
Indicator-Based EA (IBEA)   [36, 55, 57]
Parallel Eas (PEAS)   [36, 55, 57]
Pareto Envelope-based Selection Algorithm (PESA2)   [36, 55, 57]
Covariance Matrix Adaptation Evolution Strategy (CMAES)   [36, 54-57]
Augmented Lagrangian PSO (ALPSO)  [36, 54]
PSO

Our multi-objective PSO (OMOPSO)   [55, 57]

Speed-constrained multi-objective PSO (SMPSO)   [55, 57]
Non-linear Optimization Mesh Adaptive Direct (NOMAD)   [58]
Sparse Nonlinear OPTimizer (SNOPT)  [36, 54]
CONstrained function Minimization (CONMIN)  [36, 54]
Others

Interior Point OPTimizer (IPOPT)  [36, 54]

Nelder-Mead  [36, 54]
Constrained Optimization BY Linear Approximation (COBYLA)  [36, 54]
Simulated Annealing (SA)  [54, 56, 59]

2.4.2. Review of MDAO work for FOWT system

Modelling FOWT systems involves complex integration/coupling of multidisciplinary systems together.
MDAO and parameterization of a system go hand in hand as the parametric scheme describes the design space
of the system for exploration.
Some examples of parametric studies conducted on floaters are reviewed here. A precursor to the
parametrization of floating offshore wind turbine support structure is the parametric design model of oil & gas
support structures, optimized to reduce the downtime through improved seakeeping by Birk, Clauss and Lee
and detailed in [29]. In this work, they automated the hull design stage by introducing parametric shape

Electronic copy available at: https://ssrn.com/abstract=3936386

 13

generation, numeric hydrodynamics analysis assessment tools and non-linear programming algorithms for
process control. Their investigation compares the performance of three different optimization algorithms (SQP,
GA and SA) within a shape optimization framework and found that the GF optimizers (SA and GA) require
more computation time and do not always produce better results than the classical deterministic SQP method.
However, both sets of algorithms show significant improvement of seakeeping qualities.
A parametric optimization of a semi-submersible platform with heave plates was conducted by Aubault et al
[60]. Their work was conducted on Minifloat, a novel concept of semi-submersible platform developed to
enable hydrocarbon production from marginal fields in deep and ultra-deep water. In their work, they developed
a simplified hydrodynamic model to capture the parametric sensitivity of the platform responses to primary
design parameters as the hydrodynamic responses of the platform are driven by its mass properties and
geometric parameters, including that of the heave plates. Also, the use of GA to optimise the responses of the
platform was discussed in this work, and an optimized design solution was found for the simple Minifloat
platform with no substructure accessories. Results with static constraints show a linear relationship between the
payload and the platform displacement. However, the need of a sizeable draft is determined by hydrodynamic
considerations, the GA optimization process for the Minifloat resulted in a shallow operating draft.
For FOWT, Bachynski conducted a parametric work related to TLP as part of her thesis [61]. Here,
hydrodynamic loading of first, second and third order is considered with the combination of the controller and
controller faults in extreme sea states.
A couple of MDAO studies in the offshore wind turbine industry are detailed in this section. In the work of
Fylling et al [30], a GB optimization approach (SLSQP) for a spar floater, including the mooring lines and the
power cables, was presented. The objective function modelled the cost of the system, and the design variables
represented the geometric properties of the spar and mooring system. The constraints considered are the nacelle
acceleration, tower inclination and maximum tensions in mooring lines. The results indicate that response can
be optimized by modifying the cylindrical shape of the spar.
In the work of Hall et al [31], the authors conducted a study on the hull shape and mooring line optimization
of FOWT across different platform categories using a GA and a frequency domain model derived from FAST
software with a linear representation of the hydrodynamic viscous damping and no representation of the wind
turbine control. The GA is applied for single and multi-objective optimization, and the results indicate an un-
conventional design that shows the necessity for cost function refinement.
Karimi et al [32] improved Hall et al’s work using a new optimization algorithm and a linearized dynamic
model, which improved the optimal solutions. The authors incorporated a fully coupled frequency domain
dynamic model and a design parameterization scheme to evaluate the system motions and forces in turbulent
winds and irregular wave scenarios. They also selected the Kriging-Bat optimization algorithm (a surrogate-
based evolutionary algorithm) to represent the design exploration and exploitation of optimal designs across
three stability classes of platform (MIT/NREL TLP, OC3-Hywind Spar, and OC4-DeepCwind semi-
submersible platform). This optimization aimed to explore the cost implications of platform stability, expressed
through the nacelle acceleration objective function, across the three FOWT platform stability classes. An
improved correlation between cost and substructure design was obtained in this study in comparison to the work
of [31].
In [33], Hegseth et al developed a linearized aero-hydro-servo-elastic model to optimize the platform, tower,
mooring, and blade-pitch controller of a 10MW spar floating wind turbine. In this work, optimal design
solutions are found using GB optimization algorithm, considering fatigue and extreme response constraints,
taken into account as objective function – a weighted combination of system cost and power quality. The
geometric shape of the platform below the waterline is an hourglass shape that maximizes the distance between
the centre of buoyancy and center of gravity, to increase the restoring moment and natural frequency in pitch.
The large bottom diameter of the platform increases the added mass in heave, which helps to place the natural

Electronic copy available at: https://ssrn.com/abstract=3936386

14

frequency outside the wave frequency to avoid resonance. The optimization results show that local minima
occur in both the soft-stiff and stiff-stiff range of the first tower bending mode. Also, the result shows a stiff-
stiff tower design is required to reach a solution that satisfies the fatigue constraints.
The work of Ghigo et al [34] is based on the use of an in-house hydrostatic tool used to estimate the main
hydrostatic parameters of five different floating substructures. Some of the hydrostatic parameters estimated by
the in house tools are the metacentric height and hydrostatic stiffness in heave, roll and pitch. Furthermore, by
application of a generic thrust force at the centre of the rotor, the maximum inclination angle in pitch can be
estimated. [34] verified the validity of results from their in-house tools by comparing with results obtained from
Ansys Aqwa. The inhouse tool was further enhanced introducing a GA-based optimization framework order to
identify the best concept in terms of reducing the LCOE while satisfying all design requirements and the
constraints imposed by the standards. This work yielded a new floating platform concept, a derivative of the
Hexafloat with all lateral brackets removed from the Hexafloat in order to reduce weight and cost of the new
structure.
The authors of [35] developed an optimization framework for floating wind turbine support structure (spar-
buoy floater), including the mooring system. The framework builds on frequency domain modelling, and the
analysis capabilities are extended to provide analytical design sensitivities for the considered design
requirements. This capability allows quick optimization using SQP optimization algorithm [35].
Recently, Leimeister et al. [36] developed a holistic and highly flexible framework for automated simulation
and optimization of wind turbine systems, including all components within the system and their fully coupled
aero-hydro-servo-elastic behaviour. The framework consists of a modelling environment using the MoWiT
software, the simulation engine (Dymola) and a GF MO genetic optimization algorithm. This holistic
framework provides suitable applications in the areas of design optimization of floating wind turbine support
structures, optimization of wind turbine performance (power output) and loading (thrust force), tuning of wind
turbine controller for load reduction and other optimization tasks within a wind farm.
A recent investigation of estimating a platform’s hydrodynamic response by surrogacy approach is
conducted by [62]. Their work demonstrates the feasibility and performance of a surrogate model to determine
the hydrodynamic response of an axis-symmetric spar-buoy type of platform. To conduct their analyses, [62]
used a family of meta-model choice listed in Figure 6 (ANN) and the sub family of the ANN meta-model choice
used is the Extreme Learning Machines (ELMs), developed with dataset of simulations from state-of-the art
potential flow based computational code. The authors found that based on the result of a state-of-the- art
potential flow code on a limited set of geometries, the ELM based surrogate model developed to approximate
the Response Amplitude Operators of the axis-symmetric spar-buoy type of FOWT can predict the RAOs of
any FOWT geometry to an average Mean Absolute Percentage Error (MAPE) OF 2% across all degrees of
freedom. This demonstrates the feasibility of replacing computationally expensive and accurate time domain
solvers with fast and reasonably accurate surrogate model. The categorization of MDAO work done from
literature on FOWT platform is presented in

Table 3.

Electronic copy available at: https://ssrn.com/abstract=3936386

 15

Table 3: MDAO work on FOWT substructures

Architecture Type Algorithm Platform Reference

MDF SNOPT using SQP Spar [33]
Gradient
Based

MDF SQP Spar [35]

MDF SQP Spar [30]
MDF GA Spar; Semisub; TLP [31]
Gradient Free

MDF Bat (BA) Spar; Semisub; TLP [32]

MDF GA Spar [36]
MDF GA New concept [34]

Electronic copy available at: https://ssrn.com/abstract=3936386

16

3. Research gaps and future areas of research proposed

As highlighted in the introductory section of this review, the offshore wind turbine sector is still at an infancy
stage, with most of the design and optimization methodologies transferred from the oil & gas sector (fixed and
floating structure). The reliance on design and optimization methodologies from the oil and gas sectors prompts
the need to identify gaps needed for development within the FOWT sector as the design requirements for an oil
and gas structure is different from a FOWT structure. From the review conducted on MDAO, a number of gaps
in the FOWT sector craving for more research are detailed below.

3.1. Surrogacy and MDAO

A surrogate is a mathematical approximation method used to predict the behavior of a system using a set of
sampling points, generally acquired from numerical simulations [63]. Surrogate models/metamodels are models
that mimic or clone the behavior of the engineering system or the asset under investigation as closely as possible
while being computationally less expensive to evaluate in comparison to the simulation model. The concept of
surrogacy in any multidisciplinary system is fundamental. The surrogate model provides a more realistic
representative model of the system than a low fidelity model while also avoiding the high computational
expense associated with high fidelity models, as discussed in section 2.3.1.1. Different surrogate/meta
modelling techniques of choice for multidisciplinary design analysis and optimization study are presented in
Figure 6. Detailed review of these surrogate modelling techniques are made in [64] and [65].

Electronic copy available at: https://ssrn.com/abstract=3936386

 17

Figure 6: Surrogate / Meta Modelling as part of system optimization (left), Surrogate / Meta modelling techniques (right). [66]

Optimization technique within an MDAO framework can have a combination of metamodel choices and
optimization algorithms for effective system optimization. In Karimi et al’s work on multidisciplinary design
optimization of floating offshore wind turbine support structures for levelized cost of energy, the Kriging BAT
(K-BA) optimisation algorithm was used to increase the efficiency of the BA algorithm to find the global
optimal solutions [45]. Just like the K-BA, these surrogate modelling techniques highlighted in Figure 6 can be
combined with optimization algorithms for FOWT substructure optimization
As highlighted in the works of [45] and [63], surrogacy (in this case Kriging-Surrogate model) helps to
increase the efficiency of the BA algorithm to find global optimal solutions. Results of the work done by [63]
shows that in terms of search capability, efficiency and robustness, the new K-BA could demonstrate superior
capability and suitability to other well-known global optimization (GO) optimization algorithms. This is an area
of research to be explored as it has the potential to make feasibility studies of projects to be conducted faster.
Figure 6 also mentions the design of experiments (DOE), a technique for the optimal placing of test points
within the design space to estimate the actual system model using one of the surrogate techniques [63]. Some
of the widely used DOE techniques shown in Figure 6 are Fractional Factorial, Central Composite Design
(CCD), Box-Behnken, and Latin Hypercube Sampling (LHS).

3.2. Larger design space exploration

Design space exploration provides the ability to explore design alternatives prior to implementation [67].
Design space exploration is important to perform optimization, eliminate inferior designs and select a set of
final design candidates for further study or validation. Large design space exploration and exploitation can be
tailored to optimize the FOWT support structures.
In the works of [32] and [31], the design space explored for optimization purposes spans across three stability
classes of platforms with the main parameterization variables of diameter and draft. This design space can be

Electronic copy available at: https://ssrn.com/abstract=3936386

18

made more expansive by including the mooring line design variables and constraints to increase the design
space. A more expansive design space exploration and exploitation has the capability of providing more
information with regards to the understanding and optimization of FOWT systems. At the moment, design space
exploration of FOWT substructural system is mainly confined to the stability of the FOWT substructure. In
simpler cases, the design space may be characterized as single body substructure (Spar) or multi-body
substructure (Semi-Submersible, TLP).
To expand the scope of study conducted in [32] and [31] for offshore wind turbine platform is to perturb /
alter the shape/geometry of the platform. Instead of focusing on the diameter and draft variables for
characterizing the design space as highlighted in [32] and [31], perturbation of the geometry expands the design
space and enhances the selection of optimal and richer designs. Expanding a design space is achievable by
increasing the variables in the parameterization scheme. Increasing the number of combinations of substructural
parameters or the use of robust parametric schemes to describe the design space increases the chances of
identifying an optimally designed system. The search for the optimal system is conducted using an optimization
search algorithm and in cases where the search is exhaustive, surrogate-based optimization algorithms as
discussed in section 3.1 - Surrogacy and MDAO can be used to identify the optimal design.
Another way of creating a large design space is to deviate from the traditional design in terms of geometric
shapes and size as highlighted in Section 3.2.1 and furthermore, selection of optimal design in a large design
space is quite imperative as discussed in Section 3.2.2.

3.2.1. Deviation from the traditional geometric shapes of FOWT substructure

Amongst the traditional geometrical/structural designs transferred from the oil and gas sector is the FOWT
platforms/substructure, which are based on the stability classes highlighted in Section 2.1. In this infancy stage
of FOWT systems, there is need for deviation from the traditional shapes of floating substructure/ platforms for
design and optimization purposes. From this review, a research gap in platform’s geometric shapes design for
optimization purpose is identified with a need to develop a novel design framework that allows the exploration
and analysis of unconventional floating support structural geometries optimized for FOWT requirements i.e.,
minimal requirements of effective hydrodynamic stability in deep waters coupled with the provision of a low
levelized cost of energy (LCOE) from the FOWT system.
A design and optimization framework developed in the work of [36] shows that the OC3 floating spar-buoy
wind turbine system is heavily over-dimensioned as unnecessarily high safety factors are applied which
inherently makes the design more costly. [36] designed a FOWT system which is still safely operating but close
to the operational limits while constraining the outer floater dimensions to less than what obtains in the OC3
floater design; hence, a potential cost reduction.
As highlighted in section 2.2, design curve parameterization technique used for the design of ship hulls in
[25, 26, 29] can methodically be applied to the design of FOWT system to optimize floater design and generate
design with optimal shapes satisfying the design requirements. A good representation of the different optimal
shape is shown in Figure 7 and Figure 8. Figure 7 shows early shape design of semisubmersible and a new
optimized semisubmersible shape while
Figure 8 shows different design shape configuration for spar platform. This process of parameterization of
the polynomial curves to automatically generate shapes for platform is discussed in the work of [27, 29];
although, for platforms used in the oil and gas sector. This concept of shape generation and subsequent
optimization can be used to increase the design space and design, analyze and select optimal platforms in for a
floating wind turbine.
This shows that a semisubmersible does not have to be same as earlier generations and a spar does not
necessarily have to be in the conventional cylindrical shape as we know it.

Electronic copy available at: https://ssrn.com/abstract=3936386

 19

Figure 7: Deviation model from conventional semisubmersible design (Optimised model vs earlier generation models) [27, 29]

Figure 8: Deviation from conventional spar design with automated shape generation using polynomial curves [29]

3.2.2. Selection of optimal design

An expansive search of a larger design space provides more opportunity of identifying the best/optimal
design based on either adequate parametric combination or designated objective functions and the use of
efficient surrogate-based optimization algorithm to efficiently search/explore the design space. On thorough
exploration of the design space, the optimization solution is situated in the Pareto optimal set. This is a feasible
point in the design space where there is no other parametric variable combination that improves an objective
function without worsening another objective function. The identified solution/solutions are further subjected
to detailed design process and validated with a higher fidelity tool.

Electronic copy available at: https://ssrn.com/abstract=3936386

20

4. Conclusion

Presently, the concept of multidisciplinary design analysis and optimization is mostly used in the aerospace
and automotive sector. The use of MDAO is at its infancy stage with regards to the floating offshore wind
turbine sector. This paper discusses the multidisciplinary design analysis and optimization of a floating offshore
wind turbine system with a focus on the substructure. The aim of this paper is to review the available MDAO
work on FOWT substructures in literature and identify research gaps to improve the optimization framework.
Firstly, the review of the available MDAO work for FOWT substructure looks at the design, analysis and
optimization approaches for FOWT support structures. This starts with an overview of FOWT substructures,
parameterization technique, and frequency domain analysis of the model, MDAO approaches for FOWT
system, optimization algorithms, and a critical review of MDAO on FOWT substructures. The review also
highlights the available optimizers and their classification and group in Table 2. This review shows that the use
of MDAO can yield an efficient approach to design, analyze and select the optimal design of a FOWT
substructure which can substantially contribute to the reduction of the CAPEX for a FOWT substructure.
Secondly, the review identifies gaps in multi-disciplinary design analysis and optimization of FOWT
substructure amongst which is increasing the design space of the substructure by deviating from traditional
geometric shapes utilizing curve parameterization techniques like Bspline and Free-form deformation used in
the design and optimization of ship hulls to create richer design space with fewer parameters in comparison to
standard CAD design. The expanded design space results in an effective exploratory process and selection of
optimal design. In addition, another gap is the selection of a suitable optimization technique (gradient based or
gradient-free) for an optimization process for optimal design selection. More recently, the development of a
surrogate model coupled with meta-heuristic optimization algorithms is finding its way to the FOWT sector
from other industries like the aerospace and automotive. This process can be further developed to suit the design
and optimization of FOWT substructures.
.
This review concludes that if the gaps highlighted are applied to the design of a FOWT system that meets all
standard design requirements, there will be significant reduction in the capital cost required for design and build
of a FOWT system which leads to a reduction in the LCOE without compromising design requirements. In
addition to this, futuristics platform design that meets stability design requirements with a reduced CAPEX is
achievable and can be produced at the front-end engineering design (FEED) stage of a project leading to detailed
engineering design.

Electronic copy available at: https://ssrn.com/abstract=3936386

 21

References
1. Zheng, X.Y. and Y.J.A.S. Lei, Stochastic response analysis for a floating offshore wind turbine integrated with a steel fish
farming cage. 2018. 8(8): p. 1229.
2. Heronemus, W.E. and S. Marine Technology, Pollution-free energy from the offshore winds. 1972, [Washington, D.C.]: Marine
Technology Society.
3. Wang, C.M., et al., Research on floating wind turbines: a literature survey. The IES Journal Part A: Civil & Structural
Engineering, 2010. 3(4): p. 267-277.
4. Tong, K.C., Technical and economic aspects of a floating offshore wind farm. Journal of Wind Engineering and Industrial
Aerodynamics, 1998. 74-76: p. 399-410.
5. Henderson, A.R. and M.H. Patel. Floating offshore wind energy. in BWEA Conference. 1998.
6. Withee, J.E., Fully coupled dynamic analysis of a floating wind turbine system. 2004, Massachusetts Inst of Tech Cambridge.
7. WindEurope, "Offshore Wind in Europe, Key trends and statistics 2019," 2020. 2019.
8. Leimeister, M., A. Kolios, and M. Collu. Critical review of floating support structures for offshore wind farm deployment. in
Journal of Physics: Conference Series. 2018.
9. Lefebvre, S. and M. Collu, Preliminary design of a floating support structure for a 5MW offshore wind turbine. Ocean
Engineering, 2012. 40: p. 15-26.
10. Dan Kyle Spearman, S.S.e.a., Floating Wind Joint Industry Project - Phase II. 2020.
11. Ioannou, A., et al., A preliminary parametric techno-economic study of offshore wind floater concepts. Ocean Engineering,
2020. 197: p. 106937.
12. Perez-Moreno, S.S., et al., Roadmap to the multidisciplinary design analysis and optimisation of wind energy systems. Journal
of Physics: Conference Series, 2016. 753: p. 062011.
13. Sclavounos, P.D., C. Tracy, and S. Lee. Floating Offshore Wind Turbines: Responses in a Seastate Pareto Optimal Designs and
Economic Assessment. 2008.
14. IEC-61400-3-2, Wind energy generation systems – Part 3-2: Design Requirements for Floating Offshore Wind Turbines. . 2019.
15. Crome, T., Hywind Demo - Technip Experience with Offshore Wind. 2010.
16. Collu, M. and M. Borg, Design of floating offshore wind turbines, in Offshore wind farms. 2016, Elsevier. p. 359-385.
17. Ribuot, J. HEXAFLOAT Innovative Competitive Offshore Energy Production. 2019 [cited 2021 7th March 2021]; SAIPEM
MCEDD Deepwater Development]. Available from: https://mcedd.com/wp-content/uploads/2019/04/MCEDD-2019-
Presentation-SAIPEM-18-March.pdf.
18. Stiesdal. [cited 2021 9th March 2021]; Available from: https://www.stiesdal.com/offshore-windpower/.
19. Borg, M. and M. Collu, A comparison between the dynamics of horizontal and vertical axis offshore floating wind turbines.
Philos Trans A Math Phys Eng Sci, 2015. 373(2035).
20. DNVGL-ST-0119, Floating Wind Turbine Structures. 2018.
21. Veritas, B., Classification and Certification of Floating Offshore Wind Turbines. 2010.
22. ABS, Guide for Building and Classing Floating Offshore Wind Turbine Installations. American Bureau of Shipping. 2015,
Updated March 2018.
23. DNV, Design of Floating Wind Turbine Structures. 2013.
24. Tracy, C.C.H., Parametric design of floating wind turbines. 2007, Massachusetts Institute of Technology.
25. Zhang, P., D.-x. Zhu, and W.-h. Leng, Parametric approach to design of hull forms. Journal of Hydrodynamics, Ser. B, 2008.
20(6): p. 804-810.

Electronic copy available at: https://ssrn.com/abstract=3936386

22

26. Birk, L. and G.n.F. Clauss. Optimization of Offshore Structures Based on Linear Analysis of Wave-Body Interaction. in ASME
2008 27th International Conference on Offshore Mechanics and Arctic Engineering. 2008.
27. Clauss, G.F. and L. Birk, Hydrodynamic shape optimization of large offshore structures. Applied Ocean Research, 1996. 18(4):
p. 157-171.
28. Journée, J.M.J., et al., Offshore Hydromechanics. 2000: TU Delft.
29. Birk, L., G.n.F. Clauss, and J.Y. Lee. Practical application of global optimization to the design of offshore structures. in
International Conference on Offshore Mechanics and Arctic Engineering. 2004.
30. Fylling, I. and P.A. Berthelsen. WINDOPT: An Optimization Tool for Floating Support Structures for Deep Water Wind
Turbines. in Volume 5: Ocean Space Utilization; Ocean Renewable Energy. 2011.
31. Hall, M., B. Buckham, and C. Crawford. Evolving offshore wind: A genetic algorithm-based support structure optimization
framework for floating wind turbines. in OCEANS 2013 MTS/IEEE Bergen: The Challenges of the Northern Dimension. 2013.
32. Karimi, M., et al., A multi-objective design optimization approach for floating offshore wind turbine support structures. Journal
of Ocean Engineering and Marine Energy, 2017. 3(1): p. 69-87.
33. Hegseth, J.M., E.E. Bachynski, and J.R.J.M.S. Martins, Integrated design optimization of spar floating wind turbines. 2020. 72:
p. 102771.
34. Ghigo, A., et al., Platform Optimization and Cost Analysis in a Floating Offshore Wind Farm. 2020. 8(11): p. 835.
35. Dou, S., et al., Optimization of floating wind turbine support structures using frequency-domain analysis and analytical
gradients. Journal of Physics: Conference Series, 2020. 1618: p. 042028.
36. Leimeister, M., A. Kolios, and M.J.M. Collu, Development of a framework for wind turbine design and optimization. 2021.
2(1): p. 105-128.
37. Martins, J.R. and A.B.J.A.j. Lambe, Multidisciplinary design optimization: a survey of architectures. 2013. 51(9): p. 2049-2075.
38. Dykes, K., et al. Applications of Systems Engineering to the Research, Design, and Development of Wind Energy Systems. 2011.
39. Bortolotti, P., Tarres, Helena C, Dykes, Katherine L, Merz, Karl, Sethuraman, Latha, Verelst, David, & Zahle, Frederik., IEA
Wind TCP Task 37: Systems Engineering in Wind Energy - WP2.1 Reference Wind Turbines. United States. 2019.
40. Crawford, C. and R. Haimes. Synthesizing an MDO Architecture in CAD. in 42nd AIAA Aerospace Sciences Meeting and
Exhibit. 2004.
41. Bottasso, C.L., et al., Spatial estimation of wind states from the aeroelastic response of a wind turbine. 2010: p. 28-30.
42. He, J., et al., Multidisciplinary Design Optimisation of a Ship Hull Using Metamodels. 2011. 58(3): p. 156-166.
43. Muskulus, M., S.J.J.o.O. Schafhirt, and W. Energy, Design optimization of wind turbine support structures-a review. 2014. 1(1):
p. 12-22.
44. Ashuri, T., et al., Multidisciplinary design optimization of offshore wind turbines for minimum levelized cost of energy. 2014.
68: p. 893-905.
45. Karimi, M., Frequency domain modeling and multidisciplinary design optimization of floating offshore wind turbines. 2018.
46. Moreno, S.S.P. A guideline for selecting MDAO workflows with an application in offshore wind energy. 2019.
47. Shi, M., et al., A support vector regression-based multi-fidelity surrogate model. 2019.
48. Gray, J.S., et al., OpenMDAO: An open-source framework for multidisciplinary design, analysis, and optimization. 2019. 59(4):
p. 1075-1104.
49. Chehouri, A., et al., Wind turbine design : mult‰ objective optimization. 2016.
50. Yang, X.-S., 3 - Optimization algorithms, in Introduction to Algorithms for Data Mining and Machine Learning, X.-S. Yang,
Editor. 2019, Academic Press. p. 45-65.
51. Saad, A.H., Z. Dong, and M. Karimi, A comparative study on recently-introduced nature-based global optimization methods in
complex mechanical system design. Algorithms, 2017. 10(4).
52. Buckley, A.G.J.M.P., A combined conjugate-gradient quasi-Newton minimization algorithm. 1978. 15(1): p. 200-210.
53. Xian, X., et al. A powell optimization approach for example-based skinning in a production animation environment. in Computer
Animation and Social Agents. 2006. Citeseer.
54. Izzo, D.B., F, Welcome to PyGMO. . 2015.
55. openmdao.org., Optimizer. 2016.
56. Siarry, P., Metaheuristics. 2016: Springer.
57. Hadka, D., Platypus Documentation. 2015.
58. Digabel, S.L., Algorithm 909: NOMAD: Nonlinear Optimization with the MADS Algorithm. 2011. 37(4 %J ACM Trans. Math.
Softw.): p. Article 44.
59. Janga Reddy, M. and D.N. Kumar, Evolutionary algorithms, swarm intelligence methods, and their applications in water
resources engineering: a state-of-the-art review. 2020. 3: p. 135-188.
60. Aubault, A., C. Cermelli, and D. Roddier, Parametric Optimization of a Semi-Submersible Platform With Heave Plates. 2007.
61. Bachynski, E.E., Design and dynamic analysis of tension leg platform wind turbines. 2014.
62. Coraddu, A., et al. Floating Spar-Type Offshore Wind Turbine Hydrodynamic Response Characterisation: a Computational
Cost Aware Approach. in Global Oceans 2020: Singapore – U.S. Gulf Coast. 2020.
63. Saad, A., et al., A new Kriging–Bat Algorithm for solving computationally expensive black-box global optimization problems.
Engineering Optimization, 2019. 51(2): p. 265-285.
64. Younis, A. and Z. Dong, Trends, features, and tests of common and recently introduced global optimization methods.
Engineering Optimization, 2010. 42(8): p. 691-718.

Electronic copy available at: https://ssrn.com/abstract=3936386

 23

65. Jin, Y.J.S. and E. Computation, Surrogate-assisted evolutionary computation: Recent advances and future challenges. 2011.
1(2): p. 61-70.
66. LIFES50Plus. <LIFES50Plus-FOWT Optimisation Framework.pdf>. [cited 2020 October 3]; Available from:
https://lifes50plus.eu/wp-content/uploads/2015/11/GA_640741_LIFES50_D4.3-web.pdf.
67. Kang, E., E. Jackson, and W. Schulte. An approach for effective design space exploration. in Monterey Workshop. 2010.
Springer.

Electronic copy available at: https://ssrn.com/abstract=3936386