Multidisciplinary Design Optimization of Large Wind
Turbines—Technical, Economic, and Design Challenges
Turaj Ashuria,1,∗, Michiel B. Zaaijerb,2 , Joaquim R. R. A. Martinsc,3 , Jie Zhanga,2
b
a
Department of Mechanical Engineering, University of Texas at Dallas, Richardson, TX, USA
Department of Aerodynamics and Wind Energy, Delft University of Technology, Delft, ZH, The
Netherlands
c
Department of Aerospace Engineering, University of Michigan, Ann Arbor, MI, USA
Abstract
Wind energy has experienced a continuous cost reduction in the last decades. A popular cost reduction technique is to increase the rated power of the wind turbine by
making it larger. However, it is not clear whether further upscaling of the existing wind
turbines beyond the 5 to 7 MW range is technically feasible and economically attractive. To address this question, this study uses 5, 10, and 20 MW wind turbines that
are developed using multidisciplinary design optimization as upscaling data points.
These wind turbines are upwind, 3-bladed, pitch-regulated, variable-speed machines
with a tubular tower. Based on the design data and properties of these wind turbines,
scaling trends such as loading, mass, and cost are developed. These trends are used to
study the technical and economical aspects of upscaling and its impact on the design
and cost. The results of this research show the technical feasibility of the existing wind
turbines up to 20 MW, but the design of such an upscaled machine is cost prohibitive.
Mass increase of the rotor is identified as a main design challenge to overcome. The
results of this research support the development of alternative lightweight materials
and design concepts such as a two-bladed downwind design for upscaling to remain a
cost effective solution for future wind turbines.
Keywords: Upscaling; Levelized cost of energy; Wind turbine design; Annual energy
production; Design optimization
1. Introduction
Over the last decades, the cost of wind generated electricity has experienced a
continuous reduction thanks to research and development [1–3]. This cost reduction
Corresponding author
Email address: turaj.ashuri@utdallas.edu (Turaj Ashuri)
1
Visiting assistant professor
2
Assistant professor
3
Professor
∗
Preprint submitted to Energy Conversion and Management
June 29, 2016
has enabled wind energy to become more viable, and today wind energy is one of the
most affordable forms of renewable energy [4–7]. Upscaling has been performed as the
basis to enable this cost reduction [8, 9]. The development of larger wind turbines
is supported by several factors, such as higher energy capture per area land use, and
cost reduction per rated mega Watt (MW) capacity with fewer larger machines for the
same installed capacity. Therefore, turbine size is often considered as a merit index
for technology progress and development [10–13].
Despite significant technological improvements, the average cost of wind generated
electricity is higher than that from traditional energy resources such as coal and natural gas [14, 15], and it is subject to a larger variability [16–18]. It is not clear whether
further upscaling beyond the existing 5–7 MW range is both technically feasible and
economically attractive. Typically, analytic scaling laws, and extrapolation of existing
wind turbine data are used to develop different scaling trends. These scaling trends
are needed to investigate the impact of size on the technical design and the associated
cost [19–21].
However, analytic scaling laws are not capable of providing accurate realizations
of large scale wind turbines. This is related to the simple formulation of analytic
scalings, which makes them suitable for the conceptual design phase. As the size
increases beyond the few MW range, the pronounced interaction between disciplines
such as aerodynamics, structures, and controls, as well as the complex system dynamics of more flexible components, presents design challenges [22–25]. Therefore, it is
questionable whether analytic scaling can model the complex scaling behavior of large
wind turbines to accurately quantify the impact of size.
Extrapolation of the existing wind turbine data also introduces large uncertainties
in the design, and it is of limited use for this research. Additionally, distributed model
properties can not be extracted from existing data trends to take into account the
wide range of design solutions from every different manufacturer present in the trends.
To address these problems, this research uses 5, 10, and 20 MW wind turbines that
are developed using multidisciplinary design optimization (MDO) [26] with a similar
design concept and design assumptions for all sizes. This enables the incorporation of
all important disciplines and components for making reliable scaling trends that are
needed to accurately quantify the technical and economical aspects of upscaling.
Compared to classical upscaling methods, the same design assumptions reduce the
scattering of the data points used for scaling trends. It may also help to identify which
aspects of the chosen concept have strong or weak scaling behavior. Therefore, using
the design data of the developed 5, 10, and 20 MW turbines, loading, mass, and cost
trends are constructed and used to predict what exactly happens to the design and
its associated costs as the size increases. This enables a more accurate prediction of
the economic viability of upscaling, and the identification of the design challenges that
need to be overcome to achieve that viability.
The remainder of this paper is organized as follows. First, a brief overview of the
two classical upscaling methods is presented. Then, the development of the 5, 10, and
2
20 MW wind turbines is explained. Next, the construction of the scaling trends using
the developed wind turbines as scaling data points is discussed and trends for loads,
mass, and cost are presented. Next, the design challenges of larger wind turbines using
the developed scaling trends are identified and discussed. Finally, conclusions on the
economic viability of existing wind turbine upscaling are discussed.
2. Classical upscaling methods
To examine the effect of size on the design, two different methods are frequently
used. First, the analytic relation between a number of important parameters that
govern the design can be formulated as a function of rotor diameter (or radius) by assuming that all geometrical parameters vary linearly with size [27, 28]. This approach
is called the analytic scaling law (also known as the similarity or linear scaling rule).
Appendix A gives a detailed overview of the method.
Analytic scaling enables the realization of scaling trends based on wind turbines
that are not size optimized, and that may not meet all design constraints. Therefore,
the usage of this method is limited to the conceptual design phase [29]. This makes
the method less suited for studying detailed technical and economical characteristics,
in which the optimized end results are important.
Second, parameters of interest can be statistically correlated to rotor diameter
using the existing wind turbines’ data. In this approach, real data are viewed collectively, and scaling trends are developed by correlating these data [30]. As an example,
Figure 1 shows the correlation of the rated power output of 27 wind turbines to rotor
diameter in the range of 15 to 130 m.
In this figure, the curve fit to the data points shows almost a square relation with
size, and to study wind turbines that are larger than existing ones, the curve can be
extrapolated. However, extrapolation beyond the data range introduces uncertainty
in the results, which is considered to be the main drawback of this technique. Furthermore, the data points used to develop the scaling trends are obtained with wind
turbines of different concepts and design assumptions, and this introduces scattering
in the data points.
To overcome the drawbacks of these two methods, the present paper develops a
novel method where for several given scales of interest, optimized wind turbines are developed. Based on these optimized designs, the relation between different parameters
and rotor diameter can be extracted and used to develop more accurate trends [31, 32].
This method is explained in the next section.
3. Development of large scale wind turbines
The design of a wind turbine operating in a wind farm is a complex decision making
process [33]. To facilitate the understanding of the work presented in this research, a
brief summary of the MDO approach that serves as the basis for the study is presented.
4
http://www.thewindpower.net/turbines_databases_en.php, last accessed May 22, 2016.
3
9
8.1
Rated power output (MW)
7.2
6.3
5.4
4.5
y = 0.0009x1.80
R² = 0.88
3.6
2.7
1.8
0.9
0
0
17
34
51
68
85
102
119
136
153
170
Rotor diameter (m)
Figure 1: Rated power output of the wind turbine as a function of rotor diameter. These data are
collected by the authors from public product brochures and company web sites. Professional database
services also exist to buy these data collectively.4
An earlier study showed that the economies of scale are negligible in the rated
power range of 0.75 to 3 MW [34]. To study the challenges and trends of wind turbines
beyond the existing 5 to 7 MW range, technical and economical data of large scale wind
turbines are needed. Therefore, the realization of 5, 10, and 20 MW wind turbines is
performed using MDO to obtain the required technical and economical data. This
section explains how these wind turbines are designed.
3.1. Aeroservoelastic simulation
MDO requires the numerical computation of the objective function and design constraints, using simulations that represent the underlying physics. Among the various
simulations that have been developed for the wind energy research community, a series of National Renewable Energy Laboratory (NREL) tools are used as presented in
Table 1.
However, these simulators are standalone codes, and the data transfer and process
flow among them have to be done by the designer manually. All these standalone codes
are coupled to obtain an integrated and automated multidisciplinary analysis that
could be linked to an optimization algorithm. Figure 2 depicts the coupled framework
to capture the aeroservoelastic behavior of the wind turbine, and to evaluate the
objective function and the design constraints. This coupling is accomplished using a
script that controls the data and process flow. More details about the development of
this integrated design tool can be found in previous works by the authors [43–46].
4
5
Figure 2: Extended design structure matrix [47] of the coupled framework to control the process and
data of the standalone codes of Table 1. The green box is the optimizer, rounded blue boxes are
the standalone simulation, and the gray parallelograms represents the data. The gray lines allow the
data flow from top to the bottom, and from left to the right on the upper triangular section, and
bottom to the top and right to the left on the lower triangular section.
Optimizer
Blade length
& hub height
Aerodynamic
& hub height
TurbSim
Structural
design
variables
Tip-speed &
deformations
System cost
Max stress
Fatigue damage
Blade length
& hub height
Aerodynamic &
structural
design variables
Blade length
& hub height
Structural
design
variables
3D correction
AeroDyn
Aerodynamic
loads
Mass models
Modal
frequency
Blade length
& hub height
3D wind
AirfilPrep
Annual
energy
production
Aerodynamic
design
variables
Aerodynamic
loads
Mass
Mass
BModes
Modal
frequency
FAST
Mass
Structural
loads
Structural
loads
WindPACT
Crunch
Stress
signals
Fatigue
Table 1: Selection of the aeroservoelastic simulation codes
Tool
Usage
TurbSim
AeroDyn
AirfoilPrep
FAST
BModes
Crunch
WindPACT
Fatigue
Simulates the 3D turbulent flow field [35]
Simulates unsteady aerodynamic loads [36]
Corrects 2D airfoil data for 3D effects [37]
Simulates the dynamic response of the wind turbine [38]
Computes modal frequencies [39]
Processes time domain output data [40]
Models mass and cost of components [41]
Computes fatigue damage [42]
3.2. Multidisciplinary design optimization formulation
The majority of large scale wind turbines designed nowadays are upwind, 3-bladed,
and pitch-regulated variable-speed turbines, and this is the type of turbine this research
focuses on. To make consistent 5, 10, and 20 MW designs and to avoid the scattering of
the data points, the 5 MW NREL wind turbine [48] is redesigned and optimized first.
In this step, no conceptual change is considered for this design, and the optimized
design has exactly the same design details, such as the airfoil type and the controller
design. Then, to provide an initial set of design data needed for the optimization to
start with, this redesigned wind turbine is upscaled to 10 and 20 MW using linear
scaling rules [49]. After this step, time domain aeroservoelastic design optimization of
the 10 and 20 MW wind turbines takes place using the same set of design assumptions
that were used to redesign the 5 MW wind turbine.
This provides the optimal preliminary data, such as rotor diameter, hub height,
rated rotational speed, and structural and aerodynamic design of the tower and rotor.
These data are then used to estimate the mass and cost of all the components other
than the rotor and tower that yield an optimal design. The FAST aeroservoelastic
code is used for the time domain simulation of the wind turbines [38].
A design optimization problem can be mathematically stated as:
Find x = {x1 , . . . , xn } that minimizes f (x) ,
(1)
where f is the objective function, and x is the n-dimensional design variable vector,
which is subject to lower and upper bounds, as follows:
xlower ≤ x ≤ xupper ,
(2)
This minimization problem is subject to equality and inequality design constraints:
hk (x) = 0 ,
6
(3)
gj {x} ≥ 0 ,
(4)
The above functions and variables are now specified for the wind turbine problem
of interest to this research.
3.2.1. Design variables
Since the tower and the rotor are the most flexible components of a wind turbine,
and their dynamic response governs the design, they are considered as the main components. There are in total 18 design variables for the rotor. These variables are the
chord and twist distribution (7 variables: 3 twists and 4 chords), structural thickness
distribution (10 variables), and rotor rotational speed (1 variable). For the tower,
there are in total 5 design variables. The tower design variables are the height (1
variable), structural thickness at the tower bottom and top (2 variables), and tower
diameter at the tower bottom and top (2 variables).
Cubic interpolation for the blade and linear interpolation for the tower are used to
find the distributed properties of the blade and tower between these sections, respectively. Upper and lower limits are considered for these variables to define the bounds
of the design space.
3.2.2. Design constraints
In addition to the upper and lower bounds of the design variables, several functional
design constraints are enforced. These design constraints are nonlinear functions of the
design variables, and in total 51 inequality constraints are enforced for the optimization
of the rotor and tower. Partial safety factors as recommended by the IEC61400 [50]
standard are also considered in these design constraints.
The blade design constraints are stresses and fatigue damage at 5 cross sections
along the blade, blade-tower clearance from 9 to 25 m/s, and the first 3 natural frequencies of the isolated blade. Fatigue damage calculation is performed using rain-flow
cycle counting based on the stress time-signals, and the Palmgren–Miner rule [51, 52].
The tower design constraints are the stresses and fatigue damage at 6 cross sections
along the tower, and the first and second natural frequencies. All the design constraints
of the blade and tower are continuous and smooth, and therefore differentiable.
3.2.3. Objective function
The objective for each design optimization is to minimize the levelized cost of
energy (LCoE) for a given wind turbine [53, 54].
The combination of various cost models [41] and the annual energy production
(AEP) enable the calculation of LCoE as:
ICC × IR + LRC + OM
LCoE =
,
(5)
AEP
where IR represents 0.1185 of interest rate [55], and AEP is defined as:
7
AEP ≈ 8760
cut-out
X
P (Vi )f (Vi ) ,
(6)
i=cut-in
where P (V ) is the wind turbine power curve, 8760 is a constant representing the
number of hours in a year, and i index shows the wind speed range from the cut-in to
cut-out. In this formulation, the wind probability distribution function f (V ) is:
" #
k−1
k
V
V
k
exp −
,
(7)
f (V ) =
c
c
c
where k is the Weibull shape factor , and c is the Weibull speed scale factor . The
optimal design of a wind turbine has strong dependency to these parameters [56, 57],
and in this research, c is 9.47, and k is 2. These are typical values as found in the
literature [58, 59]. Note that an AEP conversion loss of 5.6% is considered for the
mechanical-to-electrical losses in the drive train, similar to the DOWEC design [60].
These cost models are either dependent or independent of the design variables of
the blade and tower. Therefore, during the optimization process the value of those
dependent models is also indirectly optimized to give an integrated design with the
lowest LCoE. An example of a dependent model is the hub mass and cost that does
depend on the blade mass. The independent models do not have any size dependency,
and therefore fixed for all sizes. The safety system is an example of an independent
cost item to size. Details of these models can be found in [61–65].
The quantification of the AEP, the system masses and the costs allows for the LCoE
to be calculated and used as a multidisciplinary objective function to be minimized.
The solution of this optimization problem results in a wind turbine design that includes
rotor and tower data, cost and mass data, and the operational parameters of the wind
turbine. These data are then used to develop the scaling trends as explained in the
next section. Table 2 presents the cost share of the optimized 5, 10, and 20 MW wind
turbines.
Note that these cost models are not intended to yield an exact turbine pricing that
is a function of volatile market factors beyond the scope of this research. However,
they can provide an accurate projection of the cost based on the existing wind turbine
technology to be used in this optimization study. To even better reflect the existing
technology, the cost models listed in Table 2 are updated using the producer price index
(PPI5 ) to the cost of materials and labor in 2010. The PPI used by the US department
of labor to track costs of products and materials as the technology changes over time.
Therefore, they represent the existing technology for better cost estimates. These cost
models have already been detailed by Fingersh et al. [41].
A variable-speed controller is used for the partial and transition load region, and
5
http://www.bls.gov/ppi/
8
Table 2: Cost estimation of the optimized 5, 10, and 20 MW wind turbines in 1000 US Dollar
Cost
5 MW
10 MW
20 MW
3 Blades
Hub
Blade pitch system
Nose cone
Low speed shaft
Main bearing
Gearbox
Mechanical brake system
Generator
Power electronics
Yaw system
Main frame
Platform and railing
Nacelle cover
Electrical connections
Cooling and hydraulic system
Safety and condition monitoring
Tower
Marinization
Turbine capital costs (TCC)
1057.5
158.3
263.0
14.2
182.6
71.9
877.2
11.0
398.0
393.2
160.5
172.9
95.1
73.3
308.8
77.2
65.3
968.9
722.2
6072.3
2474.9
324.2
664.5
21.8
499.9
245.2
2085.0
22.2
796.1
786.4
451.2
341.8
188.1
142.1
617.7
154.5
65.3
3765.6
1842.3
15489.0
8570.7
1037.5
2143.0
36.1
1783.4
1151.5
4955.0
44.4
1592.2
1572.8
1665.5
808.1
444.7
279.6
1235.5
309.0
65.3
12497.0
5042.5
45618.0
Foundation
Port and staging
Turbine installation
Electrical connection
Permits and site assessment
Personnel access equipment
Scour protection
Decommissioning
Balance of station (BOS)
2174.7
144.9
732.8
2063.5
215.5
70.2
403.0
403.3
5834.9
4349.5
289.9
1465.8
4127.0
431.0
70.2
806.2
810.8
11539.6
8699.0
579.9
2931.5
8253.9
862.1
70.2
1612.3
2058.9
23009.0
802.5
12280.2
2047.0
29886.4
6028.8
68627.0
Levelized replacement
Operation and maintenance
Interest rate
99.0
664.5
0.1185
198.0
1316.8
0.1185
396.1
2873.7
0.1185
AEP (GW Hr)
28.397
56.273
122.806
0.0630
0.0641
0.0704
Warranty premium
Initial capital cost (ICC)
LCoE
U SD
)
( kW
hr
9
a full-span rotor-collective blade pitch controller is used for the full load region. The
controller is designed after each optimization iteration to guarantee power maximization for the below rated region, and power regulation for the above rated region [66].
Table 3 provides the gross design data and properties for the optimal 5, 10, and 20 MW
wind turbines.
Table 3: Gross design data and properties of the optimized 5, 10, and 20 MW wind turbines.
Design specification
Rotor diameter (m)
Rated tip speed (m/s)
Rated rotational speed (RPM)
Gearbox ratio (–)
Cut-in, rated and cut-out wind speed (m/s)
Maximum blade pitch rate (deg/s)
Blade-pitch angle at peak power (deg)
Maximum blade tip-deflection (m)
Hub height (m)
5 MW
10 MW
20 MW
130
88
12.9
91
3, 11.4, 25
8.0
0.0
5.6
82.4
182
82
8.5
139
3, 11.7, 25
5.6
0.0
7.9
110.4
286
98
6.5
180
3, 10.7, 25
4.8
0.0
11.8
162
4. Development of scaling trends using optimized wind turbines
To study the technical and economical feasibility of larger wind turbines up to
20 MW, it is required to analyze how size impacts the design. This is achieved by
developing the loading, mass and cost trends using properties of optimized 5, 10, and
20 MW wind turbines as data points. The optimized wind turbines have the same
design concept, which is a 3 bladed, upwind, variable-speed, pitch-regulated, geared
design with a tubular tower [67].
In the upscaling literature, scaling trends are developed using power curve fits to the
data points by having the rotor diameter as the independent parameter [27, 28, 30, 68].
That is: aRb , where R is the rotor diameter, a is the curve coefficient, and b is the
curve exponent. This research uses the same notation for consistency and ease of
comparison with the analytical scaling law, and existing data trends as presented in
the literature. However, the true independent parameters for the optimized turbines
are their rated powers, for which the rotor diameters are consecutively optimized.
4.1. Loading-diameter trends
This subsection presents the trends of loading versus diameter for the main load
carrying components of a wind turbine: blade, tower and low speed shaft. For the
loading trends, the trend exponent shows the sensitivity of the loads with respect to
size. For the blade and the tower, the loading trends are presented at the blade root
and tower base, respectively, where the moments are maximum.
10
4.1.1. Blade
The blade experiences flapwise loads that are predominantly aerodynamic, and
edgewise loads that are mostly due to gravity. Analytic scaling predicts the flapwise
bending moments to increase with R3 , and edgewise bending moment with R4 . The
prediction using the existing data trend is R2.86 for flapwise loads, and R3.25 for edgewise loads as presented in Figures 3 and 4, respectively. Loading-diameter trends using
the optimized designs are R2.62 for flapwise bending moment, and R3.41 for the edgewise bending moment. The lower exponent for the existing data trend for edgewise
loads might be due to the use of data for smaller turbines, where the aerodynamic
loads are still significant when compared to gravity loads. Note that the existing data
trends in the figures are extrapolated to allow comparison with the optimized data
trends beyond their range.
Although, the two classical upscaling methods show some level of inaccuracy compared to the optimized data trends, they all agree that the edgewise loads increase
more rapidly than the flapwise loads. This means that the gravity loads have a more
negative influence on the design compared to aerodynamic loads, and it will dominate
the blade design at larger scales [69].
140
Existing data trend
Existing data extrapolation
Optimized data trend
Blade flapwise moment (MN.m)
126
112
y = 0.0175x2.86
R² = 0.94
98
84
y = 0.0441x2.62
R² = 0.99
70
56
42
28
14
0
0
30
60
90
120
150
180
210
240
270
300
Rotor diameter (m)
Figure 3: Upscaling of blade flapwise loads from 5 to 20 MW to show the increase of aerodynamic
loads
4.1.2. Low speed shaft
For the low speed shaft both the bending and torsional moment influence the
design. Shaft bending moment is mass driven and scales with R4 using analytic scaling,
and R3.01 for the optimized wind turbines. This is shown in Figure 5. There is no
published data in the literature for the existing data trends.
Shaft torsional moment is aerodynamically driven, and it scales with R3 using
11
140
Existing data trend
Existing data extrapolation
Optimized data trend
Blade edgewise moment (MN.m)
126
112
98
y = 0.0031x3.25
R² = 0.88
84
70
y = 0.0005x3.41
R² = 0.98
56
42
28
14
0
0
30
60
90
120
150
180
210
240
270
300
Rotor diameter (m)
Figure 4: Upscaling of blade edgewise loads from 5 to 20 MW to show the increase of gravity loads
120
Low speed shaft bending moment (MN.m)
108
96
84
y = 0.004x3.01
R² = 0.99
72
60
48
36
24
12
0
120
138
156
174
192
210
228
246
264
282
300
Rotor diameter (m)
Figure 5: Upscaling of low speed shaft bending moment from 5 to 20 MW reflecting the increase of
gravity loads using the optimized wind turbine data
12
analytic scaling. The optimized data trend predicts an increase with R3.03 as depicted
in Figure 6. Also, no data trends have been found in the literature for the shaft
torsional moment.
120
Low speed shaft torsional moment (MN.m)
108
96
84
y = 0.0037x3.03
R² = 0.99
72
60
48
36
24
12
0
120
138
156
174
192
210
228
246
264
282
300
Rotor diameter (m)
Figure 6: Upscaling of low speed shaft torsional moment from 5 to 20 MW reflecting the increase of
rotor torque using the optimized wind turbine data
4.1.3. Tower
The tower experiences fore-aft, side-to-side and torsional moments. The tower
fore-aft bending moment scales with R3 for the analytic scaling, with R2.33 using
extrapolation from the existing data, and with R2.92 for the optimized designs.
As Figure 7 shows, the existing data extrapolation exhibits a smaller value of the
trend exponent compared to the optimized data. This is due to the scattering of
the data points in the existing data trends (not shown due to protect intellectual
property). The quality of the correlation between two data sets is measured by the Rsquared value. R-squared is a statistical measure showing the effectiveness of the rotor
diameter in forecasting the fore-aft bending moment. When R = 1.0 the correlation is
considered 100%, and R = 0.0 when no correlation exists. Therefore, the large spread
of the data reduces R to 0.71.
Figure 8 shows the side-to-side bending moment that scales as R4.19 for the optimized design. This is considerably larger than the R3 trend in analytic scaling,
and R3.23 when using existing data trends. This shows that the side-to-side loads are
dynamically more amplified at larger scales.
This can also be recognized when looking at the side-to-side motion in Figure 9.
13
900
Existing data trend
810
Tower fore-aft moment (MN.m)
Existing data extrapolation
720
y = 1.79x2.33
R² = 0.71
Optimized data trend
630
540
y = 0.0523x2.92
R² = 0.99
450
360
270
180
90
0
0
30
60
90
120
150
180
210
240
270
300
Rotor diameter (m)
Figure 7: Upscaling of the tower fore-aft bending moment from 5 to 20 MW corresponding the increase
of rotor thrust
1000
Existing data trend
Existing data extrapolation
Optimized data trend
Tower side-side moment (MN.m)
900
800
y = 0.0348x3.23
R² = 0.74
700
600
500
y = 4E-05x4.19
R² = 0.99
400
300
200
100
0
0
30
60
90
120
150
180
210
240
270
300
Rotor diameter (m)
Figure 8: Upscaling of the tower side-to-side bending moment from 5 to 20 MW showing the increased
dynamic amplification of loads
14
As depicted in the figure, the side-to-side motion of the tower top increases more
rapidly, while the tower top fore-aft motion decreases. This behavior can act as a
design bottleneck for larger wind turbines due to the fact that the side-to-side motion
has little aerodynamic damping. Passive and active means of motion control is needed
to overcome this problem [70, 71].
0.4
0.37
Tower top displacment (m)
0.34
y = 0.7583x-0.16
R² = 0.90
0.31
0.28
y = 0.001x1.02
R² = 0.99
0.25
0.22
0.19
0.16
Tower top fore-aft trend
0.13
Tower top side-side trend
0.1
100
120
140
160
180
200
220
240
260
280
300
Rotor diameter (m)
Figure 9: Upscaling of the tower top side-to-side bending motion from 5 to 20 MW showing the
increased flexibility of the system
Figure 10 shows a good agreement for the tower torsional moment trends between
the optimized wind turbines (R3.04 ) and analytic scaling (R3 ). There is, however, a
mismatch with the existing data trend (R4.05 ). Also in this case the R-squared value
indicates a lower correlation.
As stated by Jamieson [72] the effect of turbulence in generating differential loading
on the rotor plane causes a yaw-torque that is responsible for the increase observed in
the existing data trend. However, such a mechanism was not observed in the results
of this research. Table 4 summarizes the loading-diameter trends of the tower.
Table 4: Comparison of Loading-diameter trends of the tower using different scaling methods
Scaling method
Analytic scaling
Existing data extrapolation
Optimized data trend
Fore-aft
side-to-side
torsional
R3
R2.33
R2.92
R3
R3.23
R4.19
R3
R4.05
R3.04
15
120
Existing data trend
108
Tower torsional moment (MN.m)
Existing data extrapolation
96
Optimized data trend
84
y = 0.0034x3.04
R² = 0.98
72
y = 8E-05x4.05
R² = 0.76
60
48
36
24
12
0
0
30
60
90
120
150
180
210
240
270
300
Rotor diameter (m)
Figure 10: Upscaling of the tower torsional moment from 5 to 20 MW showing mismatch of the
existing data trend with optimized and analytic scaling
4.2. Mass-diameter trends
This section presents the mass-diameter trends of the blade and tower, and their
mass share compared to the total mass of the system.
4.2.1. Blade
Figure 11 shows the mass trend of the optimized blades that scales with R2.64 . The
mass trend using analytic scaling is R3 , and R2.09 for the existing data. Compared
to the existing data trend, the optimized data indicates that upscaling with the same
materials and concepts negatively influences the design. The lower exponent for the
existing data trend may be caused by technology changes that combat the mass increase (e.g., improvements in load control from stall to collective pitch to cyclic pitch).
4.2.2. Tower
Figure 12 shows the scaling of the tower mass with R2.78 using the existing data,
and R3.22 for the optimized data. The value using the analytic scaling is R3 . The
optimized trend shows a higher mass increase compared to the two classical methods.
The reason for such a difference can best be explained when looking at the mass share
at different scales as presented in the next subsection.
4.2.3. Mass share at different scales
Figure 13 shows the mass share of the blade and tower with respect to the remaining
components of the wind turbine. Engineering models as presented in the WindPACT
16
200
Existing data extrapolation
180
Existing data trend
160
Optimized data trend
Blade mass (tonne)
140
y = 0.0571x2.64
R² = 0.99
120
100
80
y = 0.7657x2.09
R² = 0.95
60
40
20
0
0
30
60
90
120
150
180
210
240
270
300
Rotor diameter (m)
Figure 11: Scaling of a single blade mass from 5 to 20 MW showing the negative influence of upscaling
5500
Existing data trend
Existing data extrapolation
Optimized data trend
4950
4400
y = 0.0609x3.22
R² = 0.98
Tower mass (tonne)
3850
3300
2750
2200
1650
1100
y = 0.653x2.78
R² = 0.74
550
0
0
30
60
90
120
150
180
210
240
270
300
Rotor diameter (m)
Figure 12: Upscaling of the tower mass from 5 to 20 MW showing the effect of linearly varying
thickness and diameter of the optimized design
17
study are used to have an accurate estimate of the mass of components other than the
blade and tower [41]. This includes the mass of the following components: bed-plate,
gearbox, hub, low speed shaft, generator, pitch system, pitch bearing, platform and
railing, main bearing, high speed shaft and brake, nose cone, hydraulic and cooling,
nacelle cover, yaw system and foundation.
Tower
20 MW
Blades
Rest
10 MW
10%
20%
30%
34.0
10.6
55.4
0%
23.9
7.8
68.3
5 MW
21.0
8.4
70.6
40%
50%
60%
70%
80%
90%
100%
Mass share (%)
Figure 13: Mass breakdown of components from 5 to 20 MW showing how upscaling influences blade
and tower mass
As the figure shows, the mass share of the tower increases from 55.4% at 5 MW to
70.6% at 20 MW. This is consistent with the large scale exponents observed for the
tower bending and torsional moments. Therefore, tower top mass reduction and load
alleviation is an emerging need for large scale wind turbines.
4.3. Cost-diameter trends
Wind turbine design is a complex multidisciplinary task, and a technical improvement in one discipline requires to study its effect on other disciplines as well. As an
example, increasing the blade length to have higher annual energy production (aerodynamic improvement) increases the loads at the blade root in turn (negative structural
effect). Therefore, if the proposed improvement does not overcome all the negative
effects, then it will end up with a negative overall impact.
Considering these facts, a multidisciplinary merit index is needed to evaluate the
overall impact of any single disciplinary improvement. The LCoE is a multidisciplinary
merit index that couples different disciplines and allows the system evaluation of any
design improvement. Therefore, this paper uses LCoE to investigate the overall impact
18
of upscaling on the design. Unfortunately, no comparison could be made with the
analytic scaling, and existing data. For the analytic scaling no formulation can be
made to correlate the size to cost. For the existing data, manufacturers treat cost
information confidential due to the market competitiveness.
Cost models presented in the WindPACT study are used in this study to have an
estimate of the cost of components other than the blade and tower [41]. These include
the cost models for hub, pitch mechanism and bearings, nose cone, low speed shaft
and bearings, gearbox, mechanical brake, high speed shaft and coupling, generator,
power electronics, yaw drive and bearing, main frame, electrical connections, hydraulic
system, cooling system, nacelle cover, control equipment, safety system, condition
monitoring, tower, and marinization (extra cost to protect against marine environment
like salty water).
Figure 14 shows the cost breakdown of blade and tower compared to the total cost
of the machine. For the blades, the cost share increases from 7.7% at 5 MW to 10.7%
at 20 MW, and for the tower, the cost share increases from 7.0% at 5 MW to 15.6% at
20 MW. The exponent of a power curve fit to these cost elements is given in Table 5.
As presented in the table, the tower experiences the highest cost impact followed by
the yaw system, low speed shaft and blade. This means that upscaling influences more
negatively the tower cost compared to other components. Note that in this table some
cost elements have the same trend exponent, because they are a function of the rated
power, and therefore they have the same trend exponent.
Table 5: Exponent of the curve fit for 5 to 20 MW wind turbines, the higher the exponent the higher
the negative impact of upscaling
Cost component
Tower
Yaw system
Main shaft
Blades
Pitch system
Warranty
Marinization
Main bearing
Hub
Decommissioning
Gearbox
Bedplate
Railing and platform
Operation and maintenance
Foundation system
Trend exponent
3.22
2.97
2.89
2.66
2.66
2.55
2.55
2.44
2.39
2.22
2.19
1.95
1.95
1.85
1.75
Cost component
Electrical interface
Installation
Scour protection
Generator
Power electronics
Electrical connection
Engineering
Port and staging
Levelized replacement
Hydraulic and cooling
HSS and brake
Nacelle cover
Hub cone
Access equipments
Safety and control
Trend exponent
1.75
1.75
1.75
1.75
1.75
1.75
1.75
1.75
1.75
1.75
1.75
1.69
1.18
Fixed
Fixed
The influence of upscaling on the Turbine Capital Cost (TCC) and Balance of Sta19
tion (BOS) is depicted in Figure 15. The TCC consists of the following cost elements:
blades, hub, pitch mechanism and bearings, nose cone, low speed shaft and bearings, gearbox, mechanical brake, high speed shaft and coupling, generator, power electronics, yaw drive and bearing, main frame, electrical connections, hydraulic system,
cooling system, nacelle cover, control equipment, safety system, condition monitoring,
tower, and marinization. The BOS includes the following cost elements: monopile,
port and staging equipment, turbine installation, electrical interface and connections,
permits, engineering, site assessment, personnel access equipment, scour protection,
transportation, offshore warranty premium, and decommissioning.
TCC and BOS are almost the same for the 5 MW design, but during upscaling the
TCC increases more than the BOS. This requires a careful consideration of the TCC
during upscaling to cut the costs down.
20
Foundation
system and
accessories
24%
Cabling and
interconnecion
17%
Permits, warranty and
decommisioning
10%
Port, staging and
installation
6%
Other
22%
Tower and internals
8%
Nacelle assembly
17%
O&M and
replacements
6%
Blades
8%
Hub and pitch system
4%
(a) 5 MW
Foundation
system and
accessories
23%
Cabling and
interconnecion
15%
Permits, warranty and
decommisioning
10%
Port,
staging and
installation
6%
O&M and
replacements
5%
Other
21%
Tower and
internals
13%
Nacelle
assembly
16%
Blades
8%
Hub and pitch system
4%
(b) 10 MW
Foundation
Cabling and
system and
accessories interconnecion
12%
20%
Tower and
internals
18%
Permits, warranty and
decommisioning
11%
Other
19%
Nacelle
assembly
15%
Port, staging and
installation
4%
O&M and
replacements
4%
Blades
11%
Hub and pitch system
5%
(c) 20 MW
Figure 14: Cost share for the optimized 5, 10, and 20 MW wind turbines shows the dramatic increase
of tower cost with upscaling.
21
50
45.6
5 MW
10 MW
20 MW
Cost (Million US Dollar)
40
30
23.0
20
15.5
11.5
10
6.1
5.8
0
TCC
BOS
Figure 15: TCC and BOS for the optimized 5 to 20 MW wind turbines show a rapid increase in TCC
The advantage of upscaling is the increase of the AEP, and its disadvantage is the
increase in the Initial Capital Cost (ICC) that is built up from TCC and BOS. This
is presented in Figure 16, where a trade-off is made between the ICC and AEP. The
1.29 trend exponent of the AEP as an advantage of upscaling does not balance out the
1.57 trend exponent of the ICC. In turn, this negatively influences the LCoE shown
in Figure 17, with a trend exponent increase of 0.14.
5. Challenges of larger wind turbines
In this research, the influence of upscaling on the design and economical characteristics of large wind turbines was studied. This has been done using loading, cost
and mass-diameter trends. These trends were a function of rotor diameter to reflect
properly the size dependency. Blade, main shaft and tower were investigated carefully,
because they are the main load carrying, and the most flexible components of a wind
turbine.
Loading trends show that the edgewise moment of the blade increases more rapidly
than the flapwise moment. This indicates that gravity driven loads are more important
to consider than the aerodynamic driven loads as the size increases. In the case of
the tower, the side-to-side bending moment exhibits a higher sensitivity to upscaling
compared to the fore-aft moment. This is an important design consideration, since
the side-to-side motion has little aerodynamic damping, and any misalignment, e.g.,
due to wind veer or lateral waves, could cause substantial structural damage.
Referring to the mass-diameter scaling trends, the design suffers from the excessive
mass growth of the rotor-nacelle assembly. The mass of the rotor causes progressively
22
80
ICC trend
72
120
AEP trend
64
100
80
48
40
60
y = 3E+07x1.29
R² = 0.96
32
y = 1E+07x1.57
R² = 0.97
AEP (GWh)
ICC (Milion US Dollar)
56
40
24
16
20
8
0
0
130
184
286
Rotor diameter (m)
Figure 16: ICC and AEP trade-off for upscaled wind turbines from 5 to 20 MW shows how upscaling
impacts the costs.
0.0714
y = 0.031x0.1431
R² = 0.9073
LCOE (USD/kWh)
0.07
0.0686
0.0672
0.0658
0.0644
0.063
0.0616
120
140
160
180
200
220
240
260
280
300
Rotor diameter (m)
Figure 17: Trend of LCoE for wind turbines ranging from 5 to 20 MW, showing the overall negative
impact of upscaling
23
the mass increase of its supporting components: main bearing, yaw system, and tower.
The major reasons for the excessive mass increase of the blade are: maintaining the
required blade-tower clearance to prevent hitting the tower in the upwind design configuration, and the required strength against the edgewise loads that are gravity driven.
Therefore, additional material in the construction of the blade is needed, which cascades a mass increase of other supporting components of the blade such as main bearing and tower. Any future design must therefore resolve the conflict between higher
stiffness needed to have the desired blade-tower clearance, and the low mass needed
to reduce edgewise loads of the blade. In such a situation, design options would be
the application of lightweight materials such as carbon fiber reinforced composites to
reduce edgewise blade loads, downwind concepts to alleviate the blade tower clearance,
and a flexible two-bladed rotor to reduce the tower top mass and loads.
The mass and cost of the tower are also influenced negatively by upscaling. The
two main reasons are higher loads, and as mentioned above, the mass of the tower top
components. To overcome these problems, load alleviation and mass reduction are
considered essential for larger towers. Additionally, active and passive load alleviation
techniques are needed to control the motion of the tower side-to-side loads, and to add
damping to decay misaligned wind-wave loading in offshore applications.
From a cost point of view, the tower has the highest trend exponent, followed by
the yaw system, low speed shaft and the blades. Comparison between the TCC and
BOS shows a higher increase of the TCC during upscaling. Furthermore, the trade off
between the ICC and annual energy production shows that the extra captured energy
by larger wind turbines is not enough to balance out the ICC. This in turn causes the
increase of the LCoE.
These results show that upscaling without changing the chosen concept does not
lead to a reduction in the costs. It is therefore suggested that a conceptual change
in the design of large wind turbines is needed to compensate for the negative effect
of upscaling on the design and costs. This conceptual change should target the load
and mass reduction of large scale wind turbines to reduce the costs. Two-bladed
downwind turbines may reduce the disadvantages of upscaling that are observed for
the three-bladed upwind turbine of this study, since the downwind rotor reduces the
effect of tower-clearance on rotor mass [73, 74]. This, in turn, mitigates the loads.
Furthermore, a two-bladed design is potentially lighter compared to a three-bladed
design and this reduces the overall system mass. The lower aerodynamic efficiency of
a two-bladed design (1-3 % lower [75]) reduces the AEP, but this may pay off by having
a lighter design and loads. Additionally, in a two-bladed design the transportation and
installation costs are lower [63].
Several other alternative design concepts can also be beneficial for large scale designs, such as: segmented, smart, and folding wind turbine blades for mass reduction
and easier transportation [76–78], new control algorithms for load alleviation and better power regulation [79–82], and novel grid integration algorithms for better and
higher penetration [83–85].
24
6. Conclusion
Referring to the results of this research, it can be concluded that upscaling without
changing the concept, materials, and technology is not viable. The relatively small
increase in costs for the balance of station still gives an incentive for upscaling, but
this is counteracted by a large increase in the costs of the rotor-nacelle assembly.
Upscaling only can remain an effective cost reduction strategy, if it is associated
with some conceptual changes. It also needs to be performed in smaller scaling steps
with the introduction of advanced technologies to minimize the negative overall impact
of every step. Therefore, upscaling needs to be done in incremental steps to allow
the mitigation of negative impacts by advancing the current technology and design
methods. Otherwise, upscaling using existing materials and design concepts will result
in massive wind turbines that are economically inferior to the existing state of the art.
The scaling observed in data of existing wind turbines is generally more benign than
the scaling observed in this study. This is associated with the technology improvements
in existing turbines, which have gone hand-in-hand with the increase in turbine size.
It would be worthwhile to investigate whether the application of such technology
improvements to smaller scale turbines might also increase their economical value.
7. Acknowledgments
This research was part of the UpWind project supported by the European Union
sixth framework program, grant number 019945 (2006–2011). The financial support
is greatly acknowledged.
Appendix A
The analytic scaling laws presented earlier can be used to obtain an initial design for
the 10 and 20 MW wind turbines needed for optimization. These laws not only can
describe the dependency of turbine properties to rotor diameter, but they can also
be employed to obtain the design properties of an unknown turbine with respect to a
known turbine. This is explained by the following equation:
0.5Cpx · ρx · π · rx2 · Vx3
Px
,
=
Pa
0.5Cpa · ρa · π · ra2 · Va3
(8)
where subscript a and x refer to a known and unknown turbine, using the following
parameters:
P
Cp
ρ
r
V
Power output
Power coefficient
Air density
Rotor radius
Hub height wind speed
25
This research assumes the power coefficient and air density of the unknown wind
turbine to be the same as the known turbine. Knowing the dependency of the hub
height wind speed to the rotor radius as V ≈ rα (with α as the wind shear), and
replacing that in equation 8 yields:
Px
rx2 · rx3α
rx2+3α
= 2 3a = 2+3α ,
(9)
Pa
ra · ra
ra
or:
r
2+3α Px
,
(10)
rx = ra ·
Pa
The only unknown variable in equation 10 is the rotor radius of the unknown
turbine, because the designer has already selected the power output of interest to
design the unknown wind turbine for. In this formulation, the ratio of the rotor radius
between the unknown and the known turbine is defined as the scaling ratio (SR). The
SR can be used to obtain any design parameters of interest for the unknown turbine
based on the known turbine as:
Qx = Qa · (SR)SF ,
(11)
where, Q is a design parameter for both the known and unknown turbine, and SF is
the scaling factor based on the analytic scaling.
8. References
[1] Klaassen G, Miketa A, Larsen K, Sundqvist T. The impact of R&D on innovation for
wind energy in Denmark, Germany and the United Kingdom. Ecological Economics
2005;54(2):227–40.
[2] Blanco MI. The economics of wind energy. Renewable and Sustainable Energy Reviews
2009;13(6):1372–82.
[3] BoroumandJazi G, Rismanchi B, Saidur R. Technical characteristic analysis of wind
energy conversion systems for sustainable development. Energy Conversion and Management 2013;69:87–94.
[4] McGinn D, Green D, Taylor R, et al. . Global status report. Tech. Rep.; Renewables;
2013.
[5] Hepbasli A. A key review on exergetic analysis and assessment of renewable energy resources for a sustainable future. Renewable and Sustainable Energy Reviews
2008;12(3):593–661.
[6] McKenna R, Hollnaicher S, Fichtner W. Cost-potential curves for onshore wind energy:
A high-resolution analysis for Germany. Applied Energy 2014;115:103–15.
[7] Cheng M, Zhu Y. The state of the art of wind energy conversion systems and technologies: A review. Energy Conversion and Management 2014;88:332–47.
26
[8] Lantz E, Wiser R, Hand M. The past and future cost of wind energy. Tech. Rep.;
National Renewable Energy Laboratory, Golden, CO, Report No. NREL/TP-6A2053510; 2012.
[9] Islam M, Mekhilef S, Saidur R. Progress and recent trends of wind energy technology.
Renewable and Sustainable Energy Reviews 2013;21:456–68.
[10] Kaldellis JK, Zafirakis D. The wind energy (r) evolution: A short review of a long
history. Renewable Energy 2011;36(7):1887–901.
[11] Joselin Herbert G, Iniyan S, Sreevalsan E, Rajapandian S. A review of wind energy
technologies. Renewable and sustainable energy Reviews 2007;11(6):1117–45.
[12] Bulder B, Hendriks H, van Langen P, Lindenburg C, Snel H, Bauer P, et al. The icorass
feasibility study, final report. Tech. Rep.; Energy Center Nederland, ECN-E–07-010;
2007.
[13] Purvins A, Zubaryeva A, Llorente M, Tzimas E, Mercier A. Challenges and options
for a large wind power uptake by the european electricity system. Applied Energy
2011;88(5):1461–9.
[14] Conti JJ, Holtberg PD, et. al. . Annual energy outlook 2014. Tech. Rep.; Energy
information association, Report No, DOE/EIA-0383; 2014.
[15] Berry D. Innovation and the price of wind energy in the US.
2009;37(11):4493–9.
Energy Policy
[16] Afanasyeva S, Saari J, Kalkofen M, Partanen J, Pyrhönen O. Technical, economic and
uncertainty modelling of a wind farm project. Energy Conversion and Management
2016;107:22–33.
[17] Ashuri T, Zhang T, Qian D, Rotea M. Uncertainty quantification of the levelized cost
of energy for the 20mw research wind turbine model. In: AIAA Science and Technology
Forum and Exposition (SciTech), Wind Energy Symposium, San Diego, California.
American Institute of Aeronautics and Astronautics.; 2016, p. 1–11.
[18] Dı́az G, Gómez-Aleixandre J, Coto J. Dynamic evaluation of the levelized cost of wind
power generation. Energy Conversion and Management 2015;101:721–9.
[19] Capponi PAC, Ashuri T, van Bussel GJW, Kallesøe B. A non-linear upscaling approach
for wind turbine blades based on stresses. In: European Wind Energy Conference and
Exhibition, Brussels, Belgium. The European wind Energy Association; 2011, p. 1–8.
[20] Sieros G, Chaviaropoulos P, Sørensen JD, Bulder B, Jamieson P. Upscaling wind
turbines: theoretical and practical aspects and their impact on the cost of energy.
Wind energy 2012;15(1):3–17.
[21] Hendriks B. Upscaling: Consequences for concepts and design. In: The European wind
Energy Association, Brussels, Belgium. 2008, p. 1–20. Workshops Track, Session Code:
CP2.
27
[22] Veers P, Ashwill TD, Sutherland HJ, Laird DL, Lobitz DW, Griffin DA, et al. Trends
in the design, manufacture and evaluation of wind turbine blades. Wind Energy
2003;6(3):245–59.
[23] Griffith DT, Ashwill TD. The sandia 100-meter all-glass baseline wind turbine blade:
Snl100-00. Tech. Rep.; Sandia National Laboratories, SAND2011-3779, Albuquerque;
2011.
[24] Ashuri T, Zaaijer M. The future of offshore wind turbines. In: 3rd PhD seminar on
wind energy in Europe, Pamplona, Spain. European Academy of Wind Energy; 2007,
p. 1–7.
[25] Höyland J. Challenges for large wind turbine blades. PhD thesis, Norwegian University
of Science and Technology; 2010.
[26] Martins JRRA, Lambe AB. Multidisciplinary design optimization: A survey of architectures. AIAA Journal 2013;51:2049–75. doi:10.2514/1.J051895.
[27] Chaviaropoulos P. Similarity rules for wind turbine upscaling. Internal report; UPWIND; 2007.
[28] Sieros G, Chaviaropoulos P. Aspects of upscaling beyond similarity. In: The science of
making torque from wind, Crete, Greece. 2010, p. 1–10.
[29] Nijssen R, Zaaijer M, Bierbooms W, Van Kuik G, Van Delft D, van Holten T. The application of scaling rules in up-scaling and marinisation of a wind turbine. In: European
Wind Energy Conference and Exhibition, Copenhagen, Denmark. 2001, p. 619–22.
[30] Jamieson P. Innovation in wind turbine design. John Wiley and Sons; 2011.
[31] Dykes K, Damiani R, Felker F, Graf P, Hand M, Meadows R, et al. Systems engineering
applications to wind energy research, design, and development. Tech. Rep.; US Dept.
Energy, Washington, DC, USA, Tech. Rep. NREL/PO-5000-54717; 2012.
[32] Dykes K, Ning A, King R, Graf P, Scott G, Veers P. Sensitivity analysis of wind plant
performance to key turbine design parameters: A systems engineering approach. In:
32nd ASME Wind Energy Symposium, National Harbor, Maryland. 2014, p. 1–26.
[33] Lee AH, Hung MC, Kang HY, Pearn W. A wind turbine evaluation model under
a multi-criteria decision making environment. Energy Conversion and Management
2012;64:289–300.
[34] Lee J, Kim DR, Lee KS. Optimum hub height of a wind turbine for maximizing annual
net profit. Energy Conversion and Management 2015;100:90–6.
[35] Jonkman B, Buhl M. TurbSim user guide. Tech. Rep.; National Renewable Energy
Laboratory, Rept. NREL/TP-500-41136, Golden, Colorado; 2007.
28
[36] Laino D, Hansen A. User’s Guide to the Wind Turbine Dynamics Aerodynamics Computer Software AeroDyn. Tech. Rep.; Windward Engineering LLC, Prepared for the
National Renewable Energy Laboratory under Subcontract No. TCX-9-29209-01, Salt
Lake City, UT; 2002.
[37] Viterna LA, Janetzke DC. Theoretical and experimental power from large horizontalaxis wind turbines. NASA Technical Report 1982;82.
[38] Jonkman JM, Buhl Jr ML. Fast user guide. Tech. Rep.; National Renewable Energy
Laboratory, Golden, CO, Technical Report No. NREL/EL-500-38230; 2005.
[39] Bir G. BModes user guide. Tech. Rep.; National Renewable Energy Laboratory, Golden,
Colorado; 2007.
[40] Buhl M. Crunch User Guide. Tech. Rep.; National Renewable Energy Laboratory,
NREL/EL-500-30122, Golden, CO; 2003.
[41] Fingersh L, Hand M, Laxson A. Wind turbine design cost and scaling model. Tech.
Rep.; National Renewable Energy Laboratory, NREL/TP-500-40566, Golden, Colorado;
2006.
[42] Frendahl M, Rychlik I. Rainflow analysis: Markov method. International journal of
fatigue 1993;15(4):265–72.
[43] Ashuri T, Zaaijer MB, Martins JRRA, van Bussel GJW, van Kuik GAM. Multidisciplinary design optimization of offshore wind turbines for minimum levelized cost of
energy. Renewable Energy 2014;68:893–905. doi:10.1016/j.renene.2014.02.045.
[44] Ashuri T. Beyond classical upscaling: Integrated aeroservoelastic design and optimization of large offshore wind turbines. Ph.D. thesis; Delft University of Technology, the
Netherlands; 2012.
[45] Ashuri T, Zaaijer MB, van Bussel GJW, van Kuik GAM. An analytical model to extract
wind turbine blade structural properties for optimization and up-scaling studies. In:
The Science of Making Torque from Wind, Crete, Greece. Journal of Physics, Conference
series; 2010, p. 1–7.
[46] Ashuri T, Martins JRRA, Zaaijer MB, van Kuik GA, van Bussel GJ. Aeroservoelastic design definition of a 20 mw common research wind turbine model. Wind energy
2016;(DOI: 10.1002/we.1970).
[47] Lambe AB, Martins JRRA. Extensions to the design structure matrix for the description of multidisciplinary design, analysis, and optimization processes. Structural and
Multidisciplinary Optimization 2012;46(2):273–84.
[48] Jonkman J, Butterfield S, Musial W, Scott G. Definition of a 5-MW reference wind
turbine for offshore system development. Tech. Rep.; National Renewable Energy Laboratory, NREL/TP-500-38060, Golden, CO; 2009.
29
[49] Ashuri T, Zaaijer MB. Review of design concepts, methods and considerations of offshore wind turbines. In: European Offshore Wind Conference and Exhibition, Berlin,
Germany. The European Wind Energy Association; 2007, p. 1–10.
[50] IEC61400 . Wind turbines, part 3: Design requirements for offshore wind turbines.
2009.
[51] Miner MA. Cumulative damage in fatigue. Journal of applied mechanics 1945;12(3):159–
64.
[52] Haghi R, Ashuri T, van der Valk PL, Molenaar DP. Integrated multidisciplinary constrained optimization of offshore support structures. Journal of Physics: Conference
Series 2014;555(1):1–10.
[53] Chehouri A, Younes R, Ilinca A, Perron J. Review of performance optimization techniques applied to wind turbines. Applied Energy 2015;142:361–88.
[54] Mirghaed MR, Roshandel R. Site specific optimization of wind turbines energy cost:
Iterative approach. Energy Conversion and Management 2013;73:167–75.
[55] Baringo L, Conejo A. Wind power investment within a market environment. Applied
Energy 2011;88(9):3239–47.
[56] Mohammadi K, Alavi O, Mostafaeipour A, Goudarzi N, Jalilvand M. Assessing different
parameters estimation methods of weibull distribution to compute wind power density.
Energy Conversion and Management 2016;108:322–35.
[57] Mohammadi K, Mostafaeipour A. Economic feasibility of developing wind turbines in
aligoodarz, iran. Energy Conversion and Management 2013;76:645–53.
[58] Chang TP. Performance comparison of six numerical methods in estimating weibull
parameters for wind energy application. Applied Energy 2011;88(1):272–82.
[59] Morgan EC, Lackner M, Vogel RM, Baise LG. Probability distributions for offshore
wind speeds. Energy Conversion and Management 2011;52(1):15–26.
[60] Kooijman H, Lindenburg C, Winkelaar D, van der Hooft E. DOWEC 6 MW pre design,
Aero-elastic modelling of the DOWEC 6 MW pre-design in PHATAS. Tech. Rep.; ECNCX-01-135, Energy Research Center of the Netherlands, Petten; 2003.
[61] Poore R, Lettenmaier T. Alternative design study report: WindPACT Advanced wind
turbine drive train designs study. Tech. Rep.; National Renewable Energy Laboratory,
NREL/SR-500-33196, Golden, CO.; 2003.
[62] Griffin D. WindPACT Turbine Design Scaling Studies Technical Area: Composite
Blades for 80-to 120-Meter Rotor. Tech. Rep.; National Renewable Energy Laboratory,
NREL/SR-500-29492, Golden, CO; 2001.
30
[63] Smith K. WindPACT Turbine Design Scaling Studies Technical Area 2: Turbine, Rotor,
and Blade Logistics. Tech. Rep.; National Renewable Energy Laboratory, NREL/SR500-29439; 2001.
[64] Bywaters G, John V, Lynch J, Mattila P, Norton G, Stowell J, et al. Northern power
systems windpact drive train alternative design study report; period of performance:
April 12, 2001 to January 31, 2005. Tech. Rep.; National Renewable Energy Laboratory,
NREL/SR-500-35524, Golden, CO.; 2004.
[65] Shafer D, Strawmyer K, Conley R, Guidinger J, Wilkie D, Zellman T, et al. Windpact
turbine design scaling studies: Technical area 4–balance-of-station cost. Tech. Rep.;
National Renewable Energy Laboratory, NREL/SR-500-29950, Golden, CO.; 2001.
[66] Ashuri T, Zaaijer MB, van Bussel GJW, van Kuik GAM. Controller design automation
for aeroservoelastic design optimization of wind turbines. In: The science of making
torque from wind, Crete, Greece. 2010, p. 1–7.
[67] Muskulus M, Schafhirt S. Design optimization of wind turbine support structures-a
review. Journal of Ocean and Wind Energy 2014;1(1):12–22.
[68] Molly J. Maximum economic size of wind energy converters. In: The European wind
Energy Association. 1989, p. 1013–6.
[69] Ashuri T, Zaaijer MB. Size effect on wind turbine blade’s design driver. In: European Wind Energy Conference and Exhibition, Brussels, Belgium. The European wind
Energy Association; 2008, p. 1–6.
[70] Lackner MA, Rotea MA. Passive structural control of offshore wind turbines. Wind
energy 2011;14(3):373–88.
[71] Van Wingerden JW, Hulskamp A, Barlas T, Houtzager I, Bersee H, Van Kuik G, et al.
Two-degree-of-freedom active vibration control of a prototyped smart rotor. Control
Systems Technology, IEEE Transactions on 2011;19(2):284–96.
[72] Jamieson P. Loading and cost trends using certification calculation. Internal report;
UPWIND; 2007.
[73] Reiso M, Muskulus M. The simultaneous effect of a fairing tower and increased blade
flexibility on a downwind mounted rotor. Journal of Renewable and Sustainable Energy
2013;5(3):033–106.
[74] van Solingen E, van Wingerden JW, Beerens J. Integrated yaw design of a downwind
two-bladed wind turbine. In: American Control Conference (ACC), 2015. IEEE; 2015,
p. 3702–7.
[75] Paul AJ. A comparative analysis of the two-bladed and the three-bladed wind turbine
for offshore wind farms. Master’s thesis; Delft University of Technology, the Netherlands;
2010.
31
[76] Lu H, Zeng P, Lei L, Yang Y, Xu Y, Qian L. A smart segmented blade system for reducing weight of the wind turbine rotor. Energy Conversion and Management 2014;88:535–
44.
[77] Xie W, Zeng P, Lei L. A novel folding blade of wind turbine rotor for effective power
control. Energy Conversion and Management 2015;101:52–65.
[78] Van Wingerden J, Hulskamp A, Barlas T, Marrant B, Van Kuik G, Molenaar D, et al.
On the proof of concept of a smart wind turbine rotor blade for load alleviation. Wind
Energy 2008;11(3):265–80.
[79] Viveiros C, Melicio R, Igreja J, Mendes V. Performance assessment of a wind energy
conversion system using a hierarchical controller structure. Energy Conversion and
Management 2015;93:40–8.
[80] Nasiri M, Milimonfared J, Fathi S. Modeling, analysis and comparison of tsr and otc
methods for mppt and power smoothing in permanent magnet synchronous generatorbased wind turbines. Energy Conversion and Management 2014;86:892–900.
[81] Ashuri T, Rotea M, Xiao Y, Li Y, Ponnurangam CV. Wind turbine performance decline
and its mitigation via extremum seeking controls. In: AIAA Science and Technology
Forum and Exposition (SciTech), Wind Energy Symposium, San Diego, California.
American Institute of Aeronautics and Astronautics; 2016, p. 1–11.
[82] Lu Q, Bowyer R, Jones BL. Analysis and design of coleman transform-based individual
pitch controllers for wind-turbine load reduction. Wind Energy 2015;18(8):1451–68.
[83] Tang Y, Bai Y, Huang C, Du B. Linear active disturbance rejection-based load frequency control concerning high penetration of wind energy. Energy Conversion and
Management 2015;95:259–71.
[84] Wang S, Chen N, Yu D, Foley A, Zhu L, Li K, et al. Flexible fault ride through
strategy for wind farm clusters in power systems with high wind power penetration.
Energy Conversion and Management 2015;93:239–48.
[85] Kesraoui M, Chaib A, Meziane A, Boulezaz A. Using a DFIG based wind turbine for
grid current harmonics filtering. Energy Conversion and Management 2014;78:968–75.
32