See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/323848061

Generalized Analytical Displacement Model for Wind Turbine Towers under

Aerodynamic Loading

Article in Journal of Wind Engineering and Industrial Aerodynamics · March 2018

DOI: 10.1016/j.jweia.2018.03.018

CITATIONS READS

23 7,215

4 authors, including:

Gerard Cortina Ashley Spear

University of Utah University of Utah
16 PUBLICATIONS 300 CITATIONS 50 PUBLICATIONS 1,324 CITATIONS

SEE PROFILE SEE PROFILE

Marc Calaf
University of Utah
97 PUBLICATIONS 2,380 CITATIONS

SEE PROFILE

All content following this page was uploaded by Gerard Cortina on 19 March 2018.

The user has requested enhancement of the downloaded file.

 Generalized Analytical Displacement Model for Wind Turbine Towers under
Aerodynamic Loading

J. Felicianoa , G. Cortinaa,b , A. Speara , M. Calafa

a Department of Mechanical Engineering, University of Utah, Utah, USA
b Fluid Mechanics Department, EEBE UPC – Barcelona Tech, Barcelona, Spain

Abstract
The goal of this work is two-fold: 1) to determine the angular deflection and displacement of the NREL 5 MW refer-
ence wind turbine tower under different atmospheric thermal stratifications, and under overlapping wind turbine wake
effects using a comprehensive numerical analysis; 2) to develop and verify a generalized analytical model allowing to
efficiently estimate wind turbine tower displacements under a variety of flow conditions. Large-eddy simulations are
used to generate atmospheric flows similar to those of a characteristic diurnal cycle. One hour periods are selected and
inputted into an aero-elastic simulation code which resolves the rotor-disk aerodynamic loading components, which
are then used as inputs to a 3-D finite element model of the tower to compute the angular deflection and displace-
ment. To improve efficiency in tower displacement predictions, a generalized analytical model is developed based
on a simplified 2-D cantilever beam and 1-D momentum theory. Results are compared with those obtained from the
numerical simulations. Results indicate that the tower deflection standard deviation increases with increasing atmo-
spheric turbulence and wake superposition. Also, differentiated tower displacements are experienced under different
atmospheric stratifications. Interestingly, lone standing wind turbines undergo larger tower deflections compared to
those located within very large wind farms. The wind turbine rotor thrust and the aerodynamic thrust on the tower
are found to have the greatest and least influence, respectively, on tower deflection. Displacement results from the
finite element analysis are used to verify the accuracy of the new generalized analytical model showing a very good
agreement. This new model has the advantage of estimating the tower deflection by just knowing the inflow velocity
for any wind turbine, independent of the atmospheric stratification and wind farm arrangement.
Keywords: Large-Eddy Simulations; Wind Turbine; Tower Displacement; Atmospheric Boundary
Layer Flow; Finite Element Model
2017 MSC: 01-10, 99-00

1. Introduction

Wind turbine towers are prone to angular deflection, φ, and displacement, δ, because of the loads they sustain.
These loads mainly result from the interaction of the rotor blades and tower structure with the atmospheric boundary
layer (ABL) flow, as well as from the resultant forces due to the turbines regular operation, i.e. the rotation of the
blades. As wind energy continues to globally develop [1], it is important to understand and reliably predict turbine
structural response due to these external forcing actions. In this work, a complete analysis of the towers’ loads and
deflections is developed by considering the blades’ rotational effect, different atmospheric stratifications, and the effect
of wake superposition (characteristic of very large wind farms). A detailed structural analysis is necessary for ensuring
the operational serviceability and safety of wind turbines. However, to numerically perform such detailed structural
analysis, it is required to consider full realistic atmospheric conditions, from which the structural wind loads can be
determined, together with a discretized multi-degree-of-fredom (MDOF) finite element analysis (FEA) [2, 3, 4] to
determine the tower angular deflections and displacement.

URL: http://mmm.mech.utah.edu/ (A. Spear), http://wet.mech.utah.edu/ (M. Calaf)

Preprint submitted to Journal of Wind Engineering & Industrial Aerodynamics March 19, 2018
 The Rankine-Froude momentum theory [5] is one of the most widely used approaches to determine the structural
loading. This approach assumes an unperturbed wind velocity field which is approximated by either a power-law
[4, 6, 7, 8], or a log-law [9]. In essence, aerodynamic loads are traditionally derived considering a neutrally stratified
and undisturbed wind velocity. While this approximation works well for offshore flows, it is far from realistic in
most onshore conditions. Furthermore, this approach does not take into consideration the effect of multiple turbine
wake interactions. To address this limitation, some studies have included a parameterization to account for turbulence
intensity. For example, Jonkman [10] included full-field, three component stochastic winds and varied the standard
deviation of the velocity fluctuations to determine the relationship between the turbulence and the structural fatigue.
To be able to consider more realistic flows, one must rely on high-resolution numerical simulations of the atmospheric
flow. This can be effectively done with large-eddy numerical simulations (LES). In this direction, Sim et al. [11],
for example, stated the need of efficient spatial and temporal resolution of simulated inflow wind fields to properly
represent wind turbine dynamics and derive appropriate loads. In their work, three numerical approaches to study
loads on turbines were explored: using a conventional stochastic simulation, a large-eddy simulation, and a large-
eddy simulation with fractal interpolation for a neutrally stratified boundary layer. Simulations were run at velocities
of 12, 15 and 18 ms−1 , and the largest loads were found for the 12 ms−1 case. Their findings suggested that an inflow
wind field with grid spacing approximately one-tenth that of the rotor diameter with a data frequency of 1 Hz would
be adequate for load studies. Also, interesting work from Park et al. [12] demonstrated that stochastic simulations of
turbulent inflow fields used in wind turbine load computations cannot account for the varying atmospheric stratification
and characteristics of realistic ABL flows. They stated that using LES generated wind fields for generating wind
turbine loads is beneficial since LES is much more realistic. This is because it considers important wind parameters
such as enhanced wind speed and directional wind shear. In their study, a total of 50 LES simulations were run
to account for a large variability of flow characteristics such as mean geostrophic wind, surface roughness, surface
cooling rate, initial boundary layer height, the change on the Coriolis parameter, and geostrophic departure.
On the other hand, MDOF analysis is required to accurately model the tower and rotating blades assembly to
properly obtain their aerodynamic and structural loading. For example, such analysis is used by Park et al. [13]
to investigate the effect of different ABL flow characteristics on the associated loads. In their study, wind turbine
loads are computed using the aero-elastic simulation code FAST (Fatigue, Aerodynamics, Structures and Turbulence)
developed at NREL [14]. In a similar work, Murtagh et al. [15] used the MDOF analysis to investigate the vibration
response of a wind turbine tower and rotating blades assembly to estimate tower tip displacements. Comparisons
were performed between two different study cases, one that accounts for the blade/tower interaction, and another one
that does not. Results demonstrated the importance of having in consideration the blade/tower assembly. Otherwise,
results underestimate the wind turbine response at the top of the tower, especially if the fundamental frequencies of
the tower and turbine blades are close to each other.
Because detailed structural analysis of the wind turbine assembly are computationally expensive and time con-
suming, it is important to develop simplified models that can alternatively be used for preliminary design stages. In
this regard, a simplified approach to determine the tower angular deflection often used for design optimization pur-
poses is to consider a cantilever beam under point and distributed loading [8]. Additionally, given that deflections
of the turbine tower fluctuate with aerodynamic loads, studies have also examined the turbine’s vibration behavior
[9, 10, 15, 16, 17, 18], fatigue response [7, 10, 19, 20, 21], buckling characteristics [4, 18, 22], and gyroscopic loading
on the generator and transmission assemblies [23, 24, 25].
Currently, it is clear that realistic wind turbine inflow conditions are needed to numerically determine the wind
turbine structural loading, which will directly influence the angular deflection and displacement of the tower tubular-
structure. For this purpose, this study uses LES of realistic atmospheric flows [26] to generate realistic wind fields.
In this work, we further exploit the flexibility of LES to evaluate the wind turbine behaviour under two different
scenarios: a lone standing wind turbine and a wind turbine immersed within a very large wind farm (VLWF). Similar
to the work from Jonkman et al. [10], FAST will be used to study the loads and turbine performance characteristics.
Finally a finite element analysis (FEA) is used to determine the deflection or displacement of the tower-tip. In a
second stage, this numerical analysis is used to develop and validate a new analytical model for determining tower
displacements. The advantage of this new analytical model is that it only requires the wind turbine inflow velocity,
the structural characteristics of the tower, and swept area of the rotor disk. The methodology presented here has the
potential to become a basis for the design codes, where the tower-tip displacement has to be taken into account in the
design stages. Also, it has the potential to be used for tower optimization studies as previously done [27, 28, 29].
2
 In summary, the objective of this research is to study how the aerodynamic loading affects the structural response
of a turbine tower during operational conditions under different thermal stabilities. To do so, a coupling between
LES, FAST, and ABAQUS (finite element computer code) is performed. Details can be found in Section 2 where
the numerical simulations are described and the distinct case studies are listed. Section 3 presents and discusses the
results arising from the FEA. In Section 4, a generalized deflection model is developed and validated. Finally, Section
5 provides the conclusions of this work.

2. Numerical Simulations

2.1. Large-Eddy Simulations of a Diurnal Cycle and case studies

To obtain a realistic atmospheric flow field, an LES approach is used where the energy containing turbulent eddies
are resolved and the smaller ones are parametrized. Because of the large Reynolds number characteristic of ABL
flows, viscosity effects are neglected. The effect of the unresolved scales is accounted for with the Lagrangian scale
dependent model of Bou-Zeid et al. [30] in the filtered momentum equations and with the Lagrangian scale dependent
model for scalars of Calaf et al. [31] in the advection-diffusion equation for the temperature. The thermal effects are
coupled to the Navier-Stokes equations by means of a buoyancy term, considering the Boussinesq approximation. A
vertically staggered grid is used to discretize the equations which are integrated using second order finite-differences
in the vertical direction, and a Fourier decomposition in the horizontal direction, similar to Moeng [32] and Albertson
et al. [33]. Because of the use of spectral methods in the horizontal directions, the flow is horizontally periodic
and hence in practice equivalent to a horizontal infinite domain. The equations are time integrated using a second
order Adam-Bashforth scheme. The numerical algorithm is fully parallelized using the Message-Passing Interface
(MPI). At the top of the domain, a zero-flux boundary condition is imposed for momentum and temperature. At the
bottom, the no-slip condition is applied for the vertical velocity and, because of the staggered grid, equivalent shear
stress is imposed at the first grid point for the horizontal momentum components. The shear stress at the surface is
parameterized using the traditional log-law and includes the effects of stratification [34, 35, 36]. To model a diurnal
flow variation, a time varying surface temperature is imposed. This approach has been previously used in other similar
LES studies [26, 37, 38, 39, 40, 41, 42, 43, 44, 45].
To generate an atmospheric flow under different thermal stratification conditions, the simulations were run for
two consecutive diurnal cycles, out of which characteristic time-series of 60 minutes were used for further analysis.
These periods, while short, are representative of the flow during stable and unstable stratification regimes, and avoid
the continuous non-stationarity intrinsic to any diurnal process. Additionally, because during the diurnal cycle neutral
stability conditions are ephemeral, and statistics would not be representative, an additional neutrally stratified simula-
tion was performed. This flow stability is obtained by imposing a thermal equilibrium between the surface and the air
temperature above (see Sharma et al. [46] for further details).
Further, two different wind farm arrangements have been considered for each characteristic stratification; one
with an array of six by four wind turbines (VLWF case) and another one with a lone standing wind turbine (LSWF
case). A standard turbine spacing of s x ≈ 6D and sy ≈ 4D is used in the streamwise and spanwise directions,
respectively. Here D represents the turbine rotor diameter which is fixed to 126 m according to the NREL 5 MW
wind turbine. Note that as a result of the horizontal periodicity of the numerical domain, the VLWF case represents
an infinite array of wind turbines. Thus, wakes are generated by turbines outside of the horizontal domain on the
upwind side, due to the periodic boundary conditions (such it has been previously performed in other LES studies of
VLWFs [31, 41, 42, 43, 47, 48, 49, 50, 51, 52]). For the LSWF configuration, wind turbines are spaced far apart (with
s x = sy ≈ 25D), thus eliminating interaction with any other turbine wakes. Therefore, the LSWF arrangement is in
practice equivalent to the case of a lone standing turbine. Table 1 summarizes the different study cases, and Figure
1 represents the LES domain for the VLWF. Overall, a total of 6 LES configurations are analyzed; two wind farm
configurations with three different ABL stratifications each.
The numerical domain used in the LES has a size of L x = π zi , Ly = π zi and Lz = 2 zi (see Figure 1), where
the variable zi represents the height of the boundary layer at the beginning of the diurnal cycle (zi =1000 m) and is
used as a normalization length scale. A uniform numerical grid is used to discretize the LES domain with a total of
256 × 256 × 384 grid points, providing a spatial resolution of ∆x = ∆y = 12.3 m and ∆z = 5.2 m.

3
Table 1: Summary of the case studies for the different wind farm configurations (VLWF or LSWF) and ABL stratifications (u=unstable, n=neutral
and s=stable).

Name of study case Wind farm Conf. Stratification No. of WTs

LSWF90u LSWF unstable 1
LSWF90n LSWF neutral 1
LSWF90s LSWF stable 1
VLWF90u VLWF unstable 24 (6×4)
VLWF90n VLWF neutral 24 (6×4)
VLWF90s VLWF stable 24 (6×4)

Figure 1: Numerical LES domain for the VLWF scenario with a total of 24 NREL 5 MW wind turbines. (a) Perspective view. (b) Top view. (c)
Front view.

2.2. Wind Turbine Parametrization

To be able to generate realistic wind farm flows, a simplified wind turbine model based on a drag force and flow
rotation effect is used (ADM-R) in the LES [53]. While the details of the flow around the turbine blades are not
captured with this method, the turbulent wakes are well reproduced, producing realistic flows and turbine-turbine
wake interactions. Additionally, to be able to capture the flow-structure interaction in detail, the LES flow field is
coupled with FAST [14, 54]. This is an aeroelastic simulation code developed by the National Renewable Energy
Laboratory capable of reproducing the dynamic response of wind turbines. FAST enables the analysis of a wide range
of wind turbine configurations given an inflow field at the front of the turbines’ rotor disk. For example, FAST allows
studying two- or three-blade horizontal-axis rotor, pitch or stall regulation, rigid or teetering hub, upwind or downwind
rotor and lattice or tubular tower. The wind turbine can be modeled on land or offshore, on fixed-bottom or floating
substructures. FAST uses wind-inflow data (here provided by the different LES study cases) and solves for the rotor-
wake effects and blade-element aerodynamic loads, including dynamic stall using a highly resolved LES independent
grid mesh. The structural-dynamic models apply the control and electrical system reactions, the aerodynamic loads
as well as the gravitational loads. It also simulates the elasticity of the rotor, drive-train, and support structure.
In this study, the NREL 5 MW wind turbine was used as the reference wind turbine model [55, 56]. This wind
turbine has a rotor diameter of 126 m and a hub height of 90 m. It uses variable-speed and collective pitch-control
configuration with a rated wind speed of 11.4 ms−1 and rated rotor speed of 12.1 rpm. The wind turbine has a cut-in
velocity of 3 ms−1 and a cut-out velocity of 25 ms−1 (see Table 2 for more details about the specifications of the NREL
5 MW wind turbine). For this work, the wind turbine is modeled using a combined modal and multi-body dynamics
formulation, and the tower was assumed to be fixed rigidly to the ground. The wind turbine is yaw-controlled (i.e. the
wind direction is averaged over 10 minutes time windows and the yaw orientation is adjusted accordingly every 10
minutes), and pitch-controlled. The pitch angles in the wind turbine can vary from 0 to 23.7 degrees, depending upon
the inflow wind speed. This specific wind turbine is implemented here because it has been used previously in a large
number of studies, its specifications are well documented, and its numerical and experimental structural behaviour is
well known. Finally to compute the displacement of the turbine tower ABAQUS is used (detailed in Section 2.3).
To feed the FAST numerical simulations with the LES instantaneous inflow velocities, a specific numerical routine
was developed, similar to the control volume approach developed by Cortina et al. [43]. This routine obtains the
4
 Table 2: Specifications for the NREL 5 MW wind turbine [55]

Power Rating 5 MW Rotor mass 110,000 kg

Control Variable Speed, Collective Pitch Nacelle Mass 240,000 kg
Rotor orientation Configuration Upwind, 3 blades Blade mass 17,740 kg
Drive-train High speed, multi-stage gearbox Rotor diameter 126 m
Cut-In, Rated, Cut-out Wind Speed 3 ms−1 , 11.4 ms−1 , 25 ms−1 Rated tip speed 80 ms−1
Cut-In and Rated rotor speeds 6.9 rpm, 12.1 rpm Hub-height 90 m

different LES velocity components, u(y0 , z0 ), v(y0 , z0 ) and w(y0 , z0 ), in a y0 − z0 plane in front of the wind turbine and
normal to the axis of the rotor disk (see shaded grey area marked in Figure 2); where u(y0 , z0 ), v(y0 , z0 ) and w(y0 , z0 ),
are the corresponding streamwise, spanwise, and vertical velocity components relative to the inflow velocity plane.
Because wind turbines are changing their yaw-orientation every 10 minutes, the inflow velocity plane is also changing
coordinates correspondingly, always maintaining the perpendicularity to the rotor disk, and hence, aligned with the
mean wind vector. As depicted in the schematic of Figure 2a, this plane is located D/2 upstream of the wind turbine.
The D/2 distance has been selected because it is close to the wind turbine but outside of the wind turbine induction
zone [57]. Consequently, the flow field is initially computed within the LES coordinate frame (yLES , zLES ) and then
transformed into the new plane coordinate frame (y0 , z0 ). Notice from Figure 2b that the size of the velocity plane is
slightly larger than the size of the turbine rotor, being equal to L py0 = L pz0 = 156m. A similar approach was also used
in works from Sim et al. [11] and Park et al. [12, 13]. Finally, Figure 2c and 2d depict two different orientations for
the wind turbine yaw alignment. In the first one (Figure 2c), the velocity plane is represented aligned with the LES
coordinate frame, and in the second one (Figure 2d) the velocity plane is represented with a different alignment with
respect to the LES coordinate frame. For each study case, 60-minute time-series of velocity components are obtained
with a frequency of 10 Hz, which is equivalent to the LES time integration of 0.1 seconds. While the rotor loads for
the NREL 5 MW are computed using FAST, the aerodynamic tower loads are obtained directly from the LES using
the well known drag force of a cylinder (more details about this approach can be found in Cortina et al. [45]).
To study the angular deflection and displacement of the wind turbine tower structure, LES and FAST were coupled
with the finite element method (FEM) solver, ABAQUS [58]. Results of the top displacement of the tower were used
to determine deflection angles and to evaluate differences among the different study cases previously stated in Table
1. The algorithm describing the coupling between the LES, FAST and ABAQUS (FEM) is featured in Figure 3. First,

Figure 2: Schematic representation of the NREL 5 MW wind turbine and location of the inflow velocity plane used as input data for FAST numerical
simulations; (a) lateral view, (b) frontal view, (c) isometric and top-view of a turbine yaw-aligned with the LES coordinate frame and (d) same
representation as (c) but with the turbine misaligned with the LES coordinate frame.

5
the inflow wind velocity is obtained from LES (step 1) and fed into FAST to obtain the loads from the rotor disk (step
3). Then the finite element model is generated (step 4), and the aerodynamic loads of the tower (step 2) and the rotor
loads (step 3) are applied to the model, to simulate structural response of the tower in a recursive way (steps 6 and 7).
For every ‘Load Case’ (step 5), the positions of the nodes at the top of the tower are stored (step 8) to be analyzed
later in step 9.

Figure 3: Flow chart illustrating the coupling among LES, FAST and ABAQUS (FEM).

2.3. Finite Element Analysis

2.3.1. Assumptions
The FEM implemented in this study includes assumptions on the following topics: 1) the aerodynamic loading
scheme, 2) the foundation type, and 3) the geometric/construction simplifications.
First, there are different methods to apply aerodynamic loads. Some studies [9, 4, 18, 7] only applied 2 loading
components, the predominant wind load in the streamwise direction, F x , and the bending moment associated to
the rotation of the rotor, My . Another approach is the one used by Uys et al. [8] and Lee and Bang [4], which
estimated wind load distribution up to the height of the tower based on a prescribed velocity profile. Uys et al. [8]
applied three different line-loads [Nm−1 ] along the tower, and Lee and Bang [4] applied a series of point loads [N], i.e.
discrete loading, corresponding to the incremental positions where the wind velocity was approximated. Alternatively,
Lavassas et al. [7] applied wind load on the tower as a pressure [Nm−2 ] that varied with height and position about its
circumference. In this study, all 6 loading components are incorporated, e.g. forces in the streamwise, F x , spanwise,
Fy , and vertical, Fz , directions and bending moments, M x , My and Mz , (see Section 2.4 and Figure 5 for more details
about the different loading components). These loads are the result of the interaction between the ABL flow and
the wind turbine rotor disk. Additionally, a distributed surface traction was implemented to account for the direct
interaction between the wind and the tower (see Section 2.4 for more details).
Second, there are two approaches to model the soil-foundation-tower interaction. One approach is to explicitly
account for the soil-structure interaction (Bazeos et al. [9], Jonkman [10] and Fadaeinedjad et al. [17]). While this
approach is the most accurate one, specific soil properties are required, and results are highly dependent on the soil
properties and the stiffness of the tower. Because the purpose of this work is to estimate the effect of the different
6
ABL conditions on the deflection of the tower, a simplified approach is used, which assumes a rigid foundation and
an encastre condition at the base of the tower model. Studies by Lee and Bang [4], Guo et al. [18], Lavassas et al.
[7], and Bazeos et al. [22] did also utilize this approach.
Finally, the third assumption is to consider geometric simplifications. Here, structural details such as flanges,
tower sections, door frames have not been structurally resolved in the finite element model. Following the approach
used by Jonkman [10], the density of the steel AISI 1020 was increased by 8.28%, from 7,850 to 8,500 kg-m−3 , to
account for the geometrical simplifications, such the bolts, welds, and flanges that are not accounted for in the tower
thickness data (see Jonkman [10] for more details). Table 3 lists the details about the FEM geometry specifications
and material specifications.

Table 3: Specifications for the NREL 5 MW Wind Turbine tower [55]

FEM geometry specifications FEM material specifications
Tower height (ht ) 87.6 m Material Custom AISI 1020
Taper linear Material mode Isotropic linear elastic
Base diameter (Dtb ) 6m Young’s modulus (E) 210 GPa
Top diameter (Dtt ) 3.87 m Shear modulus (G) 80.8 GPa
Base Thickness (twb ) 0.027 m Poisson’s ratio (ν) 0.29
Top Thickness (twt ) 0.019 m Effective density of steel (ρe ) 8,500 kg m−3

2.3.2. Mesh Convergence for the FEM

An important step in this work was to ensure mesh convergence. For this reason, the average load components
(see Figures 6 and 7) and surface traction for the LSWFu case were used to evaluate the tower-tip angular deflection
for each mesh size. Table 4 lists the different grid sizes and corresponding number of elements selected for the
convergence analysis, from the coarsest mesh, M1, to the finest, M8. Figure 4a illustrates the angular deflection
results for each mesh size. Figure 4b shows the error in angular deflection for each model relative to the finest mesh,
M8. Based on the results shown in Figure 4, it was determined that an acceptable mesh convergence is achieved with
M6, which has 5,400 elements of size 0.5 m. The same mesh refinement level is used for all subsequent analyses.

Table 4: Characteristics for the different mesh sizes. The mesh type selected for this study is the 3-dimensional Linear Hex. From left to right, the
table lists the mesh number used just for reference, the size of the grid elements, the number of elements and the aspect ratio. The aspect ratio of
the mesh elements refers to the ratio between its largest and smallest dimension. For example, the elements of mesh M8 have (L:B) average ratios
of (5.78:1), and the worst one being (6.58:1).
Mesh No. Size No. Elements Aspect Ratio
(m) Average Worst
M1 16.0 48 678.80 789.50
M2 8.0 88 365.30 430.70
M3 4.0 184 173.20 206.00
M4 2.0 450 88.20 105.30
M5 1.0 1,260 48.83 52.64
M6 0.5 5,400 23.43 26.32
M7 0.25 21,960 11.61 13.16
M8 0.125 88,560 5.78 6.58

2.4. Loads Classification: Forces, Bending Moments and Distributed Surface Traction
In this section, the aerodynamic loads resultant from the interaction between the ABL flow and the wind turbine
rotor are classified based on the tower-top coordinate system or reference frame (see Figure 5a). This coordinate
system is located at the top-center of the tower and it is denoted with the sub-index ‘t’ (xt , yt , zt ). Therefore, the main
forces are classified as the streamwise, F x,t , spanwise, Fy,t , and vertical component, Fz,t ; and the bending moments
about each axis are M x,t , My,t and Mz,t . Notice that these forces and bending moments applied to the top of the tower
7
 (a) (b)
0.4 40

30
0.35
φ [deg]

εφ [%]
20

0.3
10

0.25 0
10 1 10 2 10 3 10 4 10 5 M1 M2 M3 M4 M5 M6 M7 M8
N umber of elements M esh N umber

Figure 4: Sub-figure (a) illustrates the mesh convergence analysis for the FEM, based on the angular deflection of the wind turbine tower, the
horizontal and vertical red lines depict the location where convergence is accomplished. In sub-figure (b) the relative error is computed with respect
to mesh M8.

account for the self weight of the blade and nacelle assembly, the wind thrust acting on the blades and nacelle, and
other loads that result from the wind turbine rotation. Moreover, the nature of these forces is complex, and they can
be characterized as quasi-steady, stochastic, and cyclic [59]. In this regard, some examples include: the aerodynamic
loads from a uniform and steady wind speed which can generate quasi-steady loads; the aerodynamic loads from a
non-uniform and unsteady wind speed which can generate stochastic loads; and mass forces that result from the rotor
and blades weight during their rotation which can cause periodic, non-stationary loads. Also, the rotor can be exposed
to non-periodic and random loads caused by wind gusts (extreme turbulent events).
In addition to the forces and bending moments applied to the top of the tower, loading conditions are applied along
the tower height to account for the non-uniformly distributed wind load acting on the tower, w x (z), (see Figure 5b),
which causes a non-uniformly distributed thrust force on the tower, FT , obtained as described in Cortina et al. [45].
To accurately account for the thrust force in the FEM, a surface traction [58], σ(z), has been imposed along the height

Figure 5: Schematic representation of the NREL 5 MW wind turbine; (a) illustrates the coordinate system used for the loads analysis, named
tower-top coordinate system. This coordinate system is fixed at the top of the tower, xt pointing in the normal downwind direction, yt pointing to
the left when looking the wind turbine from the upwind direction and zt axis points vertically upward, opposite to the gravity. Sub-figures (b), (c)
and (d) illustrate a lateral view of a schematic representation of the tower deflection for different stages, under a given aerodynamic load, w x (z), (b)
wind turbine without deflection, (c) wind turbine deflecting, and (d) wind turbine deflected.

8
of the tower using the following expression;
FT
, σ (z) = (1)
dA
where dA is the differential cross-sectional area of the tower, computed as dA = Dt dz. The parameter dz represents
the differential height and Dt represents the tower diameter. Also, following works from Lee and Bang [4], an encastre
condition was used to secure the base of the tower.

3. Results & Discussion

Next, a statistical analysis of the aerodynamic loads is presented in Section 3.1. Section 3.2 illustrates the tower
displacement and angular deflection results from FEA. In Section 3.3, the FEA results are compared with the tower-
displacements obtained by an idealized analytical model.

3.1. Statistical Analysis of the Rotor Loads

The loads at the tower-top coordinate reference frame are featured in Figures 6 and 7. Box plots represent the loads
for the 60-minute study periods in different wind farm configurations (VLWF, LSWF) and atmospheric stratifications
(u, s, n). The 25th and 75th percentiles are illustrated by the lower and top limit of the box diagram, respectively. The
distance between the top and bottom of the boxes is defined as the interquartile range and the line in the middle of
the box is the median of the data. For those cases where the median is not centered, a skewed distribution exists. The
whiskers represent the furthest observations within the data. The outliers are marked with a plus red marker (+) which
displays a value that is more than 1.5 times the interquartile range from the top or bottom of the box plot. Finally, the
circles represent the average value.

(a) (b) (c)

500 40 -3400

400 -3420
20

300 -3440
Fx,t [kN]

Fy,t [kN]

Fz,t [kN]

0
200 -3460

-20
100 -3480

0 -40 -3500
V

V
V

V
LS

LS

LS
LS

LS

LS
V

V
LS

LS

LS
LW

LW

LW
LW

LW

LW
LW

LW

LW
W

W
W

W
W

W
Fn

Fn

Fn
Fn

Fn

Fn
Fu

Fu

Fu
Fs

Fs

Fs
Fs

Fs

Fs
Fu

Fu

Fu

Figure 6: Force components at the top-tower coordinate axis obtained from FAST for the (a) streamwise, (b) spanwise and (c) vertical direction.

In Figure 6c, it can be observed that the vertical component, Fz,t , is the largest force experienced by the tower.
This force is mainly the result of the rotor and nacelle weight, and fluctuations on this force are dependent on the
acceleration and deceleration of the blades. Among the different study cases, the vertical component for the LSWFs
case exhibits the largest average force and smallest fluctuations with a null interquartile range. This case is character-
ized by having a nocturnal low-level jet (LLJ) that intersects the rotor disk [60, 40, 44] with very steady winds and an
associated low turbulence intensity which favors a constant wind turbine rotational speed. In contrast, the spanwise
force, Fy,t , is the smallest force, with the corresponding values fluctuating between positive and negative, and of the
order of 20 kN. This force results from the mass imbalance when the wind turbine is rotating, and their fluctuations
are again dependent on the blades acceleration and deceleration. While spanwise and vertical forces do not exhibit
large differences between the different study cases, the force acting in the streamwise direction, F x,t , shows important
differences. Overall, the values oscillate by a factor 10, and the LSWF cases produce the largest values due to starker
inflow velocities. Again, the thrust force for the LSWFs case creates the largest load with smallest fluctuations, and
the LSWFu and LSWFn show similar characteristics. The smallest force is observed for the VLWFs case, which is
9
characterized by smaller velocities, due to the increased drag by the presence of the wind farm in combination with
low atmospheric turbulence. For a more detailed analysis about the influence of atmospheric stability on wind turbine
loads for the NREL 5 MW see the detailed work from Sathe et al. [61].
A different behaviour can be observed for the bending moments at the top-tower coordinate frame (Figure 7).
In this case, M x,t is the dominant component corresponding to the clockwise rotation of the blades about the x-
axis. For the LSWF configurations, the M x,t magnitudes are greater than for the VLWF configurations. Again, the
stable distributions exhibit a smaller standard deviation than the unstable ones (which reflects the variation on the
rotational speed of the wind turbine) with neutral conditions showing a behavior similar to the unstable cases. Also,
as previously seen in Figure 6, the LSWF cases produce the largest bending moments. While M x,t and My,t are more
or less consistent among the different study cases, special attention should be placed on the LSWFs case My,t average
value. Here the value is positive, but negative for the other cases, with oscillations from positive to negative values.
The nature of this bending moment is the result of the balance between the weight of the rotor (producing negative
bending moments) and the effects of a non-uniform vertical velocity profile (producing positive bending moments
if the velocity at top-half rotor disk is larger than at the bottom-half of the rotor). Therefore, positive My,t values
indicate that the velocity intersecting the top-half rotor disk creates a larger moment than the weight of the rotor,
and otherwise for negative values of My,t . Thus, the differentiated behavior of the LSWFs case is the result of the
large difference between the velocity at the top- and bottom-rotor disk. This effect is associated with the presence
of a LLJ, characteristic of stable stratified ABL flows. The LLJ is a fast moving ribbon of air in the low levels of
the atmosphere which, in this case, intersects with the top half of the rotor disk. This phenomena has also been also
previously reported by Fitch et al. [60] and Sharma et al. [40] in other related wind energy studies.

(a) (b) (c)

2500
2000 2000
2000
1000 1000
My,t [kN-m]

Mz,t [kN-m]
Mx,t [kN-m]

1500
0 0
1000
-1000 -1000
500
-2000 -2000
0
V

V
V

V
LS

LS
LS

LS
V

V
LS

LS
V

LS
LS

LS

LW

LW
LW

LW

LW
LW

LW
LW
LW

W
W

W
W

W
W
W

Fn

Fn
Fn

Fn
Fu

Fu
Fs

Fs
Fn

Fs

Fs
Fn
Fu

Fu

Fu
Fs

Fs
Fu

Figure 7: Bending moments at the top-tower coordinate axis obtained from FAST, (a) about the x-axis, (b) about the y-axis and (c) about the z-axis.

From both Figures 6 and 7, larger fluctuations are observed for the VLWFu case due to the coupling of the
background ABL turbulence, characteristic of the unstable ABL stratification, and wind farm induced turbulence.
The lowest loading is observed for the VLWFs due to reduced inflow velocities. On the contrary, the LSWFs case
exhibits the largest average values and smallest fluctuations.

3.2. Tower Displacement and Angular Deflection from FEA Results

While both angular deflection and tower displacement results are presented, discussion is primarily focused on the
tower displacement. First, the FEA results are provided and compared to an existing analytical solution, where the
wind turbine tower is modeled as a cantilever beam.
Figure 8a illustrates the angular deflection (left y-axis) and displacement (right y-axis) of the tower as a function
of the study case. By comparing Figure 8a with previous Figures 6 and 7, there is a distinguishable relationship
between the wind turbine loading and the tower’s structural response. Namely, the tower displacement and angular
deflection illustrates the same behavior as the rotor thrust force (see Figure 6a) and the bending moment about the
x-axis (see Figure 7a). This interdependence strongly correlates deflection to the incoming wind conditions and mass
imbalance due to the rotational speed of the wind turbine. Following the aerodynamic loading, the smallest average

10
tower deflections were observed in the VLWF cases with values ranging from 0 to 40 cm. By contrast, LSWF cases
produced the largest deflections, from 20 to 70 cm.
(a) (b) (d)
70 10 12

11.5 LSW F u
0.4 8
60 V LW F u

U∞ [m/s]
6 11 LSW F s
50 V LW F s
4 10.5
0.3 LSW F n
10 V LW F n
40 2
10 20 30 40 50 60
φ [deg]

δ [cm] 9.5 F itting

(c)

λ
0.2 30
10 9

20 9
8.5
ω [rpm]

0.1
8
8
10
7
7.5
0 0 6
10 20 30 40 50 60 7
10 20 30 40 50 60
V
V

LS
LS

LS

LW
LW
LW

δ [cm]
W
W

δ [cm]
Fn
Fn
Fu

Fs

Fs
Fu

Figure 8: Sub-figure (a) represents the angular deflection in the left y-axis, φ, and the tower displacement, δ, in the right y-axis, as a function of
the study case, represented in the x-axis. Sub-figure (b) illustrates the tower displacement as a function of the average wind turbine inflow velocity
and sub-figure (c) with respect to the rotation speed of the wind turbine. Finally, sub-figure (d) represents the tip-speed ratio, λ, as a function of the
displacement.

The displacement for the LSWFs case exhibits the smallest standard deviation, 1.2 cm, and the largest average dis-
placement, 57.8 cm; these results correspond with the strong steady winds characteristics of this stratification regime.
The results for the LSWFu and LSWFn cases illustrate similar average deflection values near 40 cm. The differences
in average and standard deviation between the LSWFu, LSWFn, and LSWFs cases suggest that the fatigue lifespan
of a tower could be largely affected during unstable and neutral stratified regimes [20]. Regarding the wind farm
configuration, similar mixing conditions yielded similar average displacement and standard deviations; the VLWFu
case experienced smaller deflection, around 25 cm, which is a 45% reduction with respect to the LSWFu case. Also,
the VLWFu case illustrates one of the largest standard deviations, 10.2 cm, corresponding to the high turbulent levels
characteristic of unstable conditions in large wind farm scenarios [62]. The VLWFs case yields the smallest aver-
age displacement, 17.1 cm. Notably, the average displacement of the VLWFu, LSWFu, and LSWFs cases exceed
their neutral condition counterpart, which suggests that using aerodynamic forces based on neutral conditions would
underestimate the actual forces experienced by the wind turbine under unstable or stable stratified flows.
Because of the apparent relationship between the tower displacement and the streamwise force and bending mo-
ment along the x-axis, it seems that the displacement of the tower is, therefore, dependent on the inflow wind velocity,
U0 , and rotational speed of the blades, ω. Figures 8b and 8c illustrate these relationships and indicate a prominent de-
pendency. Here, only the average displacement value has been included in the figures, depicting each study case with a
different marker. The dashed lines illustrate a linear relationship between U∞ and ω with δ, with associated coefficients
of determination of R2 = 0.9928 and 0.9992, respectively. Finally, combining U∞ and ω, Figure 8d demonstrates the
relationship between the tip-speed ratio, λ, with respect to the tower deflection, where λ = (ω R)/U∞ with R equal
to the radius of the wind turbine defined in Table 2. In this case, since the relationship between the angular velocity
and the inflow velocity is not linear, a quadratic polynomial was fitted to the data. It is important to highlight that
as the tip-speed ratio diminishes, the maximum tower displacement increases. This relationship indicates that the
inflow wind velocity is the variable dominating the tower deflection when compared to the angular rotation of the
wind turbine rotor. This finding is used later in Section 4 to develop a new generalized analytical model to estimate
tower deflections.

3.3. A Simplified Model to Determine the Tower Displacement

3.3.1. Description of the 2-D Cantilever Beam Analysis
Given that deflections are strictly related to the incoming wind conditions and following the verification method of
Lee and Bang [4], the tower is simplified as a 2-D cantilever beam [63] with a constant effective diameter and effective
11
thickness throughout its entire length. Therefore, the maximum tower displacement can be found by superposition
of the displacements caused by the three primary load components, as illustrated in Figure 9: a concentrated force
at the free end (Sub-figure 9a), a couple moment at the free end (Sub-figure 9b), and a uniformly distributed load
(Sub-figure 9c). Using Expression 2 for the maximum deflection, δAnalytical , estimates are determined for each of the
six study cases:

F x,t h3t My,t h2t ω x h4t

δAnalytical = δF + δ M + δω = + + , (2)
3EI 2EI 8EI
where δAnalytical is the total deflection due to contributions from the axial force, δF , the bending moment about the
y-axis, δ M , and the wind force acting on the tower, δω . F x,t , is the primary stream-wise force component, My , t, is the
bending moment about the y-coordinate axis and ω x is the uniformly distributed aerodynamic wind load. The tower
height is represented by ht , the elastic modulus by E, and I is the moment of inertia of a uniform, hollow cylinder.
This simplified approach assumes that all the tower properties are constant along its height. For this reason, it needs
to be defined an effective tower diameter, De , which is computed as the average value between the top diameter Dtt
and the base tower diameter Dtb , De = 0.5(Dtb + Dtt ) − te . The same applies to the effective thickness, te . Table 5
summarizes the required parameters to compute the tower deflection using Expression 2.

Tower height (ht ) 87.6 m

Young’s modulus (E) 210 GPa
Base diameter (Dtb ) 6m
Base Thickness (twb ) 0.027 m
Top diameter (Dtt ) 3.87 m
Top Thickness (twt ) 0.019 m
Effective diameter (De ) 4.9235 m
Effective thickness (te ) 0.023 m Figure 9: Schematic representation of the tower idealized as a 2-
D cantilever beam for: (a) a concentrated load at the free end,
Table 5: Parameters used to compute the tower deflection using F x,t , (b) a couple moment at the free end, My,t , and (c) uniformly
the analytical model approach. distributed load, ω x , along the tower.

3.3.2. Performance of the 2-D Cantilever Beam Analysis

Results obtained from considering the tower as a 2-D cantilever beam are presented in Figure 10. The vertical
axis is represented in logarithmic scale to sense the order of magnitude between each load component’s contribution
to the overall displacement. For specific values, see Table 6 where the tower displacement associated to the different
contributions are listed.
It is important to highlight that this simplified model under-predicts the tower deflections for all study cases.
However, the trends match those from the detailed FEA. Consistent errors of around 25% and 21% exist for the
isolated and farm configuration cases (see Table 6). The LSWFs case still produced the greatest tower displacement
while the VLWFs produced the smallest. Therefore, we can conclude that the FEM analysis qualitatively corroborates
the values from the analytical model, and the under-prediction is possibly related to the fact that the mass of the
nacelle and rotor contribute in increasing the displacement of the tower. This last argument is stated since larger
absolute errors observed in the LSWF cases, compared to the VLWF cases (see Table 6). In other words, the LSWF
cases experience faster rotational speeds due to larger ABL inflow wind velocities. This results in an increase of the
vertical force component at the top of the tower (see previous Figure 6c). This fact correlates well with the larger
absolute errors observed in the LSWF cases, compared to the VLWF cases (see Table 6).
The tower displacement due to the wind load intersecting the tubular-tower, δω , is three orders of magnitude
smaller than the one caused by the rotor thrust force, δF . Also, the displacement caused by the moment at the free
end is approximately two orders of magnitude smaller than that caused by δF . Therefore, the tower displacement is
mainly caused by the thrust force and, as a result, the bending moment and the wind load acting on the tower can be

12
 δABAQUS δF δM δω
2
10

1
10
δ [cm]

0
10

-1
10

-2
10
LSWFu VLWFu LSWFs VLWFs LSWFn VLWFn

Figure 10: Bar representation of the tower average maximum displacement computed by ABAQUS, δABAQUS , and computed using Equation 2.
The maximum displacement for the simplified 2-D cantilever beam is represented as a function of the different contributions: thrust force (δF ),
bending moment (δ M ) and wind force at the tower (δω ). It is important to notice that the y-axis is represented in logarithmic scale.

neglected on a first order magnitude approach. This important information is used to develop a generalized analytical
model.

Table 6: Values for the tower displacement associated to the different contributions of the analytical model, δAnalytic = δF + δ M + δω , and
computed by ABAQUS, δABAQUS . The last column corresponds to the relative error between the simplified analysis and ABAQUS results, εδ =
|(δAnalytic − δABAQUS )|/δABAQUS . Units are in centimeters.

Study case δF δM δω δAnalytic δABAQUS εδ [%]

LSWFu 32.43 -0.84 0.11 31.69 41.95 24.46
VLWFu 19.32 -1.34 0.06 18.03 22.94 21,40
LSWFs 39.38 1.10 0.04 40.51 56.14 27,84
VLWFs 13.95 -0.97 0.03 13.02 16.60 21,56
LSWFn 29.46 -0.12 0.07 29.40 39.68 25,90
VLWFn 16.47 -1.27 0.04 15.24 19.26 20,87

4. Generalized Analytical Model (GAM)

In this section, the 2-D cantilever beam approach (Equation 2) is generalized to predict tower-tip displacements
for any wind turbine, and the predictions are compared to results from both the 3-D FEA and the previous analytical
expression. For this purpose, the two previous findings are used: (i) the tower deflection is mainly due to the inflow
wind velocity, (ii) the bending moment about the y-axis and the wind load along the tower are negligible. Using the
second assumption, Equation 2 can be reduced to,

F x,t h3t
.
δGAM = δF = (3)
3EI
From one dimensional momentum theory or Actuator Disk Theory [64, 65], the wind turbine thrust force can be
written as
1 2
Fx = ρ A U 0 CT , (4)
2
where U 0 is the mean incoming velocity at hub height and at D/2 upstream of the wind turbine (the D/2 distance has
been selected because it is close to the wind turbine and still outside of the induction zone [57]), A is the blades swept

13
area, ρ is the air density, and CT is the thrust coefficient. Substituting Equation 4 into Equation 3, one obtains that
2
1 ρ A U 0 h3t
δGAM = CT . (5)
6 EI
This last expression is referred as the generalized analytic model, δGAM , and it is dependent on the structural design of
the tower (E, I, ht ), the swept area of the wind turbine (A), the wind turbine inflow velocity (U0 ), the air density (ρ),
and the wind turbine thrust coefficient (CT ). Alternatively, if the induction factor (a) is known, the thrust coefficient
can be replaced by the following expression CT = 4a(1 − a) (see Manwell et al. [66] for more details), tending to
2
2 ρ A U h h3t
δGAM = a(1 − a). (6)
3 EI
Figure 11 illustrates, in dashed lines, the behaviour of the tower displacement using the generalized analytical model
(Equation 6), and based on the structural properties of the NREL 5-MW tower, for different induction factors as a
function of the mean incoming velocity at the hub height. The red dashed line depicts the deflection for an ideal
scenario, where a=1/2, and the blue line depicts the optimal scenario at the Betz Limit, when a=1/3. The different
markers depict the displacement obtained from Equation 2 and using the loads obtained from FAST. The error bars
display the error found by comparing Equation 2 with the displacements obtained by the FEM analysis. Finally, the
green circles illustrate the displacements obtained by means of Equation 3, and using the thrust force from FAST for
larger velocities than the ones experienced on the study cases.

Cut in wind speed Rated wind speed Cut out wind speed

120
a = 1/3
S-I S-II S-III
0.40 1.2
100 a = 1/2
0.36
0.32
0.28 1
80 0.24
0.8 LSW F u
δ/ht [%]
δ [cm]

60 0.16 V LW F u
0.6
0.12
LSW F s
40
0.4 V LW F s
0.08
LSW F n
20
a = 0.04 0.2 V LW F n
F AST
0 0
0 5 10 15 20 25

U 0 [m/s]

Figure 11: Representation of the generalized analytical model (GAM) for different axial induction coefficients for the NREL 5-MW wind turbine,
dashed lines. The red dashed line represents the model with the maximum theoretical axial induction factor, a=1/2, and the blue dotted-dashed line
represents the ideal induction coefficient, a=1/3. The different markers illustrate the displacement values computed with Expression 2 and using
the thrust force obtained with FAST for the different case studies. The error bar illustrates the error with respect to the FEM analysis. The green
circles depict the displacement obtained using Equation 3 and by using the thrust force obtained from FAST, at different wind speed.

From Figure 11, it can be concluded that the NREL 5 MW wind turbine has an induction factor that oscillates
between 0.32 and 0.24 between the cut-in wind speed and the rated wind speed, with the displacement values obtained
with the FEM and Equation 2 following the trend illustrated by the GAM. More interestingly the GAM is capable of
estimating the tower displacements for the operational region of the wind turbine (SII), up to its maximum displace-
ment which takes place at the rated wind speed, when the turbine thrust force is maximized. After reaching the rated
wind speed, the wind turbine control system feathers the blades to maintain a constant rotational speed (SIII). The
feathering of the blades reduces the total thrust, and therefore reduces the tip tower displacement.
14
5. Conclusions

In this work we have explored the usage of realistic wind turbine inflow conditions during the evolution of a
diurnal cycle and under two different scenarios for creating new knowledge about the dependence of turbine loads and
deformation with atmospheric stratification. Selected periods of a diurnal cycle have been used as input to FAST to
obtain the wind turbine loading under different stratification conditions. Finally, a 3-D finite element model has been
used to determine the tip tower displacements and angular deflection. This exhaustive numerical analysis serves as a
benchmark to develop a new, generalized analytical model to efficiently estimate wind turbine tower displacements
under a variety of structural and fluid-flow ABL stability conditions. Results from the FEM analysis are used to verify
the new model, showing a very good agreement.
Results from both the numerical and analytical predictions demonstrated that the lone standing wind turbines
under stable stratified flow leads to the largest tower displacements, with an associated small standard deviation.
Neutral and unstable stabilities illustrate larger standard deviations, with the very large wind farm case under unstable
conditions leading to the largest standard deviations. Also, it has been demonstrated that, while the vertical force is
the largest one acting on the top of the tower, the main contributor to the displacement is the thrust force at the rotor,
with the aerodynamic load along the tower being three orders of magnitude smaller. This demonstrates that the thrust
force resultant from the wind-tower interaction can be neglected. The FEM analysis revealed that the incorporation
of the thermal atmospheric boundary layer stability only affects the standard deviation of the data, but the average
displacements are only a function of the inflow wind conditions (which are dependent on thermal stratification).
Results also demonstrated that the mass imbalance due to the rotational effect of the blades has a minor effect on
the tower deflection. With all of these considerations, we were able to develop a generalized analytical displacement
model (GAM). Results from the exhaustive numerical study revealed that the GAM is capable of estimating the tower
displacements during operational conditions, up to its maximum displacement with good accuracy. Interestingly,
its maximum displacement takes place at the rated wind speed, when the turbine thrust force is maximized. After
reaching the rated wind speed, the wind turbine control system feathers the blades, reducing the overall thrust force
and therefore the tower displacement.
The GAM is suitable to obtain an estimate of the tower tip displacement by just knowing the turbine inflow wind
velocity, the structural characteristics of the tower, and the blades swept area with independence of the background
wind farm configuration and ABL stratification. The methodology presented here has the potential to become a basis
for design codes, where tower-tip displacements have to be taken into account during the design stages. Due to the
efficiency of the GAM predictions, the new model has also the potential to be used for tower optimization studies.

Acknowledgements

This research has been supported with the start-up funds provided by the Mechanical Engineering Department
at University of Utah to Assistant Professor Marc Calaf. The authors would like to acknowledge the computational
support provided by the Center for High Performance Computing (CHPC) at University of Utah. The authors would
like to thank Dr. Jason Jonkman and Bonnie Jonkman from the National Renewable Energy Laboratory.

References
[1] G. Watson, B. Hill, F. Courtney, P. Goldman, S. Calvert, R. Thresher, E. Assimakopoulos, J. Lyons, B. Bell, A framework for offshore wind
energy development in the united states, Massachusetts Technology Collaborative (MTC).
[2] Y. Bazilevs, M.-C. Hsu, I. Akkerman, S. Wright, K. Takizawa, B. Henicke, T. Spielman, T. E. Tezduyar, 3d simulation of wind turbine rotors
at full scale. part i: Geometry modeling and aerodynamics., International Journal for Numerical Methods in Fluids 65 (1-3) (2011) 207–235.
doi:10.1002/fld.2400.
[3] Y. Bazilevs, M.-C. Hsu, J. Kiendl, R. Wüchner, K.-U. Bletzinger, 3d simulation of wind turbine rotors at full scale. part ii: Fluid–structure
interaction modeling with composite blades, International Journal for Numerical Methods in Fluids 65 (1-3) (2011) 236–253.
[4] H.-J. Bang, H.-I. Kim, K.-S. Lee, Measurement of strain and bending deflection of a wind turbine tower using arrayed fbg sensors, Interna-
tional journal of precision engineering and manufacturing 12 (13) (2012) 2121–2126.
[5] R. E. Wilson, P. Lissaman, Applied aerodynamics of wind power machines, [Corvallis, Or., Oregon State University], 1974.
[6] E. W. Peterson, J. P. Hennessey Jr, On the use of power laws for estimates of wind power potential, Journal of Applied Meteorology 17 (3)
(1978) 390–394.
[7] I. Lavassas, G. Nikolaidis, P. Zervas, E. Efthimiou, I. Doudoumis, C. Baniotopoulos, Analysis and design of the prototype of a steel 1-mw
wind turbine tower, Engineering structures 25 (8) (2003) 1097–1106.

15
 [8] P. Uys, J. Farkas, K. Jarmai, F. Van Tonder, Optimisation of a steel tower for a wind turbine structure, Engineering structures 29 (7) (2007)
1337–1342.
[9] S. Bisoi, S. Haldar, Dynamic analysis of offshore wind turbine in clay considering soil–monopile–tower interaction, Soil Dynamics and
Earthquake Engineering 63 (2014) 19–35.
[10] J. M. Jonkman, Dynamics modeling and loads analysis of an offshore floating wind turbine, University of Colorado at Boulder, 2007.
[11] C. Sim, S. Basu, L. Manuel, On space-time resolution of inflow representations for wind turbine loads analysis, Energies 5 (7) (2012) 2071.
doi:10.3390/en5072071.
[12] J. Park, S. Basu, L. Manuel, Large-eddy simulation of stable boundary layer turbulence and estimation of associated wind turbine loads, Wind
Energy 17 (3) (2014) 359–384. doi:10.1002/we.1580.
[13] J. Park, L. Manuel, S. Basu, Toward isolation of salient features in stable boundary layer wind fields that influence loads on wind turbines,
Energies 8 (4) (2015) 2977. doi:10.3390/en8042977.
[14] S. Guntur, J. Jonkman, S. Schreck, B. Jonkman, Q. Wang, M. Sprague, Fast v8 verification and validation using experiments from aeroelasti-
cally tailored megawatt-scale wind turbine blades, Tech. rep., National Renewable Energy Lab., Golden, CO (US) (2016).
[15] P. Murtagh, B. Basu, B. Broderick, Along-wind response of a wind turbine tower with blade coupling subjected to rotationally sampled wind
loading, Engineering structures 27 (8) (2005) 1209–1219.
[16] T. J. Larsen, T. D. Hanson, A method to avoid negative damped low frequent tower vibrations for a floating, pitch controlled wind turbine, in:
Journal of Physics: Conference Series, Vol. 75, IOP Publishing, 2007, p. 012073.
[17] R. Fadaeinedjad, M. Moallem, G. Moschopoulos, Simulation of a wind turbine with doubly fed induction generator by fast and simulink,
IEEE Transactions on energy conversion 23 (2) (2008) 690–700.
[18] L. Guo, C.-M. Uang, A. Elgamal, I. Prowell, S. Zhang, Pushover analysis of a 53 m high wind turbine tower, Advanced Science Letters 4 (3)
(2011) 656–662.
[19] M. Grujicic, G. Arakere, E. Subramanian, V. Sellappan, A. Vallejo, M. Ozen, Structural-response analysis, fatigue-life prediction, and material
selection for 1 mw horizontal-axis wind-turbine blades, Journal of materials engineering and performance 19 (6) (2010) 790–801.
[20] K. S. Hansen, G. C. Larsen, S. Ott, Dependence of offshore wind turbine fatigue loads on atmospheric stratification, in: Journal of Physics:
Conference Series, Vol. 524, IOP Publishing, 2014, p. 012165.
[21] M. Holtslag, W. Bierbooms, G. Bussel, Wind turbine fatigue loads as a function of atmospheric conditions offshore, Wind Energy 19 (10)
(2016) 1917–1932.
[22] N. Bazeos, G. Hatzigeorgiou, I. Hondros, H. Karamaneas, D. Karabalis, D. Beskos, Static, seismic and stability analyses of a prototype wind
turbine steel tower, Engineering structures 24 (8) (2002) 1015–1025.
[23] M. M. Shokrieh, R. Rafiee, Simulation of fatigue failure in a full composite wind turbine blade, Composite Structures 74 (3) (2006) 332–342.
[24] A. Heege, J. Betran, Y. Radovcic, Fatigue load computation of wind turbine gearboxes by coupled finite element, multi-body system and
aerodynamic analysis, Wind Energy 10 (5) (2007) 395–413.
[25] L. Y. Pao, K. E. Johnson, A tutorial on the dynamics and control of wind turbines and wind farms, in: American Control Conference, 2009.
ACC’09., IEEE, 2009, pp. 2076–2089.
[26] V. Kumar, J. Kleissl, C. Meneveau, M. B. Parlange, Large-eddy simulation of a diurnal cycle of the atmospheric boundary layer: Atmospheric
stability and scaling issues, Water Resources Research 42 (2006) 1–18. doi:10.1029/2005WR004651.
[27] H. M. Negm, K. Y. Maalawi, Structural design optimization of wind turbine towers, Computers & Structures 74 (6) (2000) 649–666.
[28] J. C. Nicholson, Design of wind turbine tower and foundation systems: optimization approach, The University of Iowa, 2011.
[29] M. Muskulus, S. Schafhirt, Design optimization of wind turbine support structures-a review, Journal of Ocean and Wind Energy 1 (1) (2014)
12–22.
[30] E. Bou-Zeid, C. Meneveau, M. Parlange, A scale-dependent Lagrangian dynamic model for large eddy simulation of complex turbulent flows,
Physics of Fluids 17 (2) (2005) 1–18. doi:10.1063/1.1839152.
[31] M. Calaf, M. B. Parlange, C. Meneveau, Large eddy simulation study of scalar transport in fully developed wind-turbine array boundary
layers, Physics of Fluids 23 (12) (2011) 126603.
[32] C.-h. Moeng, A Large-Eddy-Simulation Model for the Study of Planetary Boundary-Layer Turbulence, Journal of the Atmospheric Sciences
41(13) (1984) 2052–2062.
[33] J. D. Albertson, M. B. Parlange, Natural integration of scalar fluxes from complex terrain, Advances in Water Resources 23 (3) (1999)
239–252. doi:10.1016/S0309-1708(99)00011-1.
[34] A. S. Monin, A. M. Obukhov, Basic laws of turbulent mixing in the surface layer of the atmosphere, Tr. Akad. Nauk SSSR Geofiz. Inst.
English translation by John Miller, 1959. 24 (151) (1954) 163–187.
[35] M. B. B. W. Parlange, Regional shear stress of broken forest from radiosonde wind profiles in the unestable surface layer, Boundary-Layer
Meteorology 64(4) (1993) 355–368.
[36] E. Bou-Zeid, C. Meneveau, M. B. Parlange, A scale-dependent Lagrangian dynamic model for large eddy simulation of complex turbulent
flows, Physics of Fluids 17 (2) (2005) 1–18.
[37] V. Kumar, G. Svensson, a. a. M. Holtslag, C. Meneveau, M. B. Parlange, Impact of surface flux formulations and geostrophic forcing on
large-eddy simulations of diurnal atmospheric boundary layer flow, Journal of Applied Meteorology and Climatology 49 (2010) 1496–1516.
doi:10.1175/2010JAMC2145.1.
[38] G. Svensson, a. a. M. Holtslag, V. Kumar, T. Mauritsen, G. J. Steeneveld, W. M. Angevine, E. Bazile, a. Beljaars, E. I. F. de Bruijn, a. Cheng,
L. Conangla, J. Cuxart, M. Ek, M. J. Falk, F. Freedman, H. Kitagawa, V. E. Larson, a. Lock, J. Mailhot, V. Masson, S. Park, J. Pleim,
S. Söderberg, W. Weng, M. Zampieri, Evaluation of the diurnal cycle in the atmospheric boundary layer over land as represented by a variety
of single-column models: The second gabls experiment, Boundary-Layer Meteorology 140 (2) (2011) 177–206.
[39] M. Abkar, A. Sharifi, F. Porté-Agel, Wake flow in a wind farm during a diurnal cycle, Journal of Turbulence 17 (4) (2016) 420–441.
[40] V. Sharma, M. Parlange, M. Calaf, Perturbations to the spatial and temporal characteristics of the diurnally-varying atmospheric boundary
layer due to an extensive wind farm, Boundary-Layer Meteorology (2016) 1–28.
[41] G. Cortina, V. Sharma, M. Calaf, Investigation of the incoming wind vector for improved wind turbine yaw-adjustment under different

16
 atmospheric and wind farm conditions, Renewable Energy 101 (2017) 376–386.
[42] G. Cortina, M. Calaf, Turbulence upstream of wind turbines: A large-eddy simulation approach to investigate the use of wind lidars, Renew-
able Energy 105 (2017) 354–365. doi:10.1016/j.renene.2016.12.069.
[43] G. Cortina, R. B. Cal, M. Calaf, Distribution of mean kinetic energy around an isolated wind turbine and a characteristic wind turbine of a
very large wind farm, Physical Review Fluids 074402 (2016) 1–18. doi:10.1103/PhysRevFluids.1.074402.
[44] G. Cortina, V. Sharma, M. Calaf, Wind farm density and harvested power in very large wind farms: A low-order model, Phys. Rev. Fluids 2
(2017) 074601. doi:10.1103/PhysRevFluids.2.074601.
[45] G. Cortina, Y. Karkera, L. Ibarra, M. Calaf, Loads on Tall Wind Turbine Towers under different Atmospheric Stability Conditions. Submitted.
[46] V. Sharma, M. B. Parlange, M. Calaf, Perturbations to the spatial and temporal characteristics of the diurnally-varying atmospheric boundary
layer due to an extensive wind farm, Boundary-Layer Meteorology (2016) 1–28doi:10.1007/s10546-016-0195-0.
[47] M. Calaf, C. Meneveau, J. Meyers, Large eddy simulation study of fully developed wind-turbine array boundary layers Large eddy simulation
study of fully developed wind-turbine array, Physics of Fluids 22, 015110 (2009). doi:10.1063/1.3291077.
[48] J. Meyers, C. Meneveau, Optimal turbine spacing in fully developed wind farm boundary layers, Wind Energy 15 (2) (2012) 305–317.
[49] M. Abkar, F. Porté-Agel, Mean and turbulent kinetic energy budgets inside and above very large wind farms under conventionally-neutral
condition, Renewable Energy. 70 (2014) 142–152.
[50] V. Sharma, M. Calaf, M. Lehning, M. Parlange, Time-adaptive wind turbine model for an LES framework, Wind Energydoi:10.1002/we.
[51] N. Ali, G. Cortina, N. Hamilton, M. Calaf, R. Cal, Turbulence characteristics of a thermally stratified wind turbine array boundary layer via
proper orthogonal decomposition, Journal of Fluid Mechanics 828 (2017) 175–195.
[52] N. Ali, N. Hamilton, G. Cortina, M. Calaf, R. B. Cal, Anisotropy stress invariants of thermally stratified wind turbine array boundary layers
using large eddy simulations, Journal of Renewable and Sustainable Energy 10 (1) (2018) 013301.
[53] Y.-T. Wu, F. Porté-Agel, Atmospheric turbulence effects on wind-turbine wakes: An les study, Energies 5 (12) (2012) 5340–5362. doi:
10.3390/en5125340.
[54] J. M. Jonkman, M. L. B. Jr., Fast user’s guide, Technical report no. nrel/el-500-38230, National Renewable Energy Laboratory (2005).
[55] J. Jonkman, S. Butterfield, W. Musial, G. Scott, Definition of a 5-mw reference wind turbine for offshore system development, Technical
report no. nrel/tp-500-38060, National Renewable Energy Laboratory (2009).
[56] B. R. Resor, Definition of a 5-MW/61.5m Wind Turbine Blade Reference Model, Tech. Rep. April, Sandia National Laboratories (2013).
[57] E. Simley, L. Y. Pao, P. Gebraad, M. Churchfield, Investigation of the impact of the upstream induction zone on lidar measurement accuracy
for wind turbine control applications using large-eddy simulation, Journal of Physics: Conference Series 524 (1) (2014) 012003.
[58] H. Hibbitt, B. Karlsson, P. Sorensen, Abaqus v6. 12 documentation—abaqus analysis user’s manual, Providence (RI): Dassault Systè”mes
Simulia.
[59] R. Gasch, J. Twele, Wind power plants: fundamentals, design, construction and operation, Springer Science & Business Media, 2011.
[60] A. C. Fitch, J. K. Lundquist, J. B. Olson, Mesoscale Influences of Wind Farms throughout a Diurnal Cycle, Monthly Weather Review 141
(2013) 2173–2198.
[61] A. Sathe, J. Mann, T. Barlas, W. Bierbooms, G. Bussel, Influence of atmospheric stability on wind turbine loads, Wind Energy 16 (7) (2013)
1013–1032.
[62] K. S. Hansen, R. J. Barthelmie, L. E. Jensen, A. Sommer, The impact of turbulence intensity and atmospheric stability on power deficits due
to wind turbine wakes at horns rev wind farm, Wind Energy 15 (1) (2012) 183–196. doi:10.1002/we.512.
[63] F. P. Beer, R. Johnston, J. Dewolf, D. Mazurek, Mechanics of Materials, Sixth Edition, McGraw-Hill, 2012.
[64] W. Rankin, On the mechanical principles of the action of propellers, Trans. Inst. Naval Architects 6 (1865) 13–39.
[65] R. E. Froude, On the part played in propulsion by differences of fluid pressure, Trans. Inst. Naval Architects 30 (1889) 390.
[66] J. F. Manwell, J. G. McGowan, A. L. Rogers, Wind energy explained: theory, design and application, John Wiley & Sons, 2010.

17

View publication stats