SSRN Id4496459
SSRN Id4496459
SSRN Id4496459
ed
Re-evolution Optimization and Multi Objective
Application on Horizontal Axis Wind Turbines
iew
ShengXian Huang1 Hua Yan1 HuiHe Qiu2 Ying Wang1, 2 *
1. Shanghai key laboratory of multiphase flow and heat transfer of power engineering, School of Energy and Power Engineering,
University of Shanghai for Science and Technology, Shanghai, China;
2. Department of Mechanical and Aerospace Engineering, The Hong Kong University of Science and Technology, Clear Water
Bay, Hong Kong, China.
v
(Email: [email protected] or [email protected])
re
Abstract: The design of specific airfoil for wind turbine is a crucial factor in the research of wind
turbine performance and stability. Based on bionic methods and numerical optimization methods,
a brand new airfoil design method is proposed without relying on any traditional airfoil and airfoil
design experience. This study proposes a re-evolution method (numerical optimization methods)
er
suitable for designing bionic airfoils based on the previously proposed bionic airfoil (Dol-Rot24°),
which was inspired by the Dall's Porpoise. According to this method, a new airfoil is successful
pe
designed that gives consideration to both aerodynamic performance and roughness sensitivity. The
research content mainly focuses on two aspects: new airfoil design and the new airfoil
performance verification. In terms of the new airfoil design, a surrogate model based on radial
basic function (RBF) is established for predicting aerodynamic performance and roughness
ot
sensitivity of airfoils. This surrogate model can significantly reduce the cost of bionic airfoil
optimization calculations, and control the time of 8000 optimization iterations within 1 minute.
Based on the actual engineering environment of a horizontal axis wind turbine, the multi-island
tn
genetic algorithm is used to adjust the profile of the Dol-Rot 24° airfoil. The optimized airfoil is
named Dol-Rot 24° opt. In the performance verification of the Dol Rot 24° opt airfoil, it is found
that compared with the Dol Rot 24° airfoil, the lift coefficient of the Dol Rot 24° opt airfoil is
rin
increased by 171.06%, and the lift-to-drag ratio is increased by 52.41% at most. Moreover, the
roughness sensitivity of the Dol-Rot 24° opt airfoil is also the best among the three airfoils (S809,
Dol-Rot 24°, Dol-Rot 24° opt). By modeling the three airfoils according to the NREL Phase VI
ep
wind turbine blade parameters and conducting numerical simulations, it is found that the newly
optimized wind turbine blades exhibit greatly improved energy efficiency. The purpose of this
study is to provide effective references for the design, bionic methods, and practical engineering
Pr
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
1. Introduction
Facing the severe ecological and environmental crises, actively developing wind power
ed
globally has become a consensus . To accelerate the high-quality development of wind energy,
[1]
improving the efficiency and operational stability of wind turbines has become the primary goal of
many scholars [2]. However, it must be noted that research on optimizing the performance of wind
iew
turbines should not be limited to improving their aerodynamic performance, but should be further
optimize the energy efficiency on the premise of strictly ensuring the stability or safety of wind
turbines.
As the basic component element of the blade, the optimization of blade aerodynamic
performance mainly focuses on optimizing the maximum lift-to-drag ratio [3]
and preventing flow
v
separation [4]
of the airfoil. Under high angle of attack conditions, fluid separation on the blade
re
surface is inevitable due to inherent unsteadiness [5]
. Therefore, it often leads to a significant
amount of energy loss and even poses a great threat to the safety of wind turbines [6]. Furthermore,
the operational stability of wind turbines also includes the sensitivity of the airfoil's surface roughness [7].
Wind turbine blades are exposed to complex high-altitude environments, and the surface
er
inevitably suffers from long-term corrosion or adhesion of pollutants such as dirt, dust, insect
bodies, etc [8]
. Thus, ensuring the efficient and stable operation of blades under the influence of
these pollutants is typically a key factor in evaluating blade performance.
pe
1.1 Aerodynamic optimization research of airfoil
To improve the aerodynamic performance of wind turbine blades, many researchers have
conducted numerous studies on various active [9] and passive [10]
flow control methods . However,
traditional active and passive flow control methods may have some unavoidable limitations in
ot
optimizing the aerodynamic performance of wind turbine blades . For example, additional
[6]
auxiliary structures in passive control may cause adverse interference with the flow field.
Meanwhile, in active control methods, huge energy consumption and complex mechanical
tn
structures also exist. Therefore, many researchers hope to imitate bionic structures with excellent
aerodynamic performance in nature, so as to create new airfoils or blades . By utilizing the
[11]
superior aerodynamic shape of these new bionic airfoils or blades, researchers aim to optimize the
rin
flow field structure and improve the aerodynamic performance of wind turbine blades while
minimizing the limitations of traditional active and passive flow control methods in practical
applications. Previous studies have demonstrated the effectiveness of using the reversed
engineering techniques based on bionic methods to obtain new airfoils with excellent aerodynamic
ep
performance quickly. These new airfoils can directly replace traditional airfoils and control the
flow field of wind turbine blades through their superior aerodynamic shape. This approach seems
to be an effective way of improving the efficiency of wind turbines and developing high-efficiency
Pr
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
obtain the contours of wings or bodies of organisms . In this method, the contours can be
[12]
directly applied to the design of new airfoils. This method has the advantage of quickly obtaining
ed
new airfoils with superior aerodynamic performance, and is uncomplicated to operate without
requiring designers to have extensive experience in airfoil design or bionic aerodynamic
performance optimization theory.
iew
For instance, Zhang et al. [13]
utilized the wavy fin limb of the humpback whale as the
foundation, and modify the equal cross-sectional blade's leading edge into a sinusoidal wave
similar to the the wavy fin. Lin et al. [14] also directly applied this wave structure to the fx63 airfoil
and designed horizontal axis wind turbine blades accordingly. In addition, similar methods also
include imitating the wavy tail edge of the owl [15] and the streamline head of the Dall's Porpoise [16].
v
Therefore, it is believed that this bionic method can significantly construct a new airfoil with
superior aerodynamic performance. Another method for obtaining biological contours is to extract
re
cross-sectional slices of the overall biological contour and directly use them as bionic new airfoil.
Song et al. [17]
extracted cross-sectional slices of seagull wings at 40% extension and found that
this bionic airfoil can effectively improve the wind energy utilization efficiency of small wind
er
turbines. Chen et al. [18]
used a 3D laser scanner to obtain a point cloud of owl wings and divided
the the scanned 3D model into ten equidistant sections from the wing root to the tip. The each
section was obtained as a bionic new airfoil, and a bionic horizontal axis wind turbine blade was
pe
subsequently designed based on the bionic airfoils.
As indicated by the above content, wind turbine blade design based on bionic methods is a
valuable and efficient engineering application methodology. However, this airfoil design method
also has some drawbacks, that are nearly inevitable in all bionic methods applied in practical
ot
engineering applications. As shown in Fig. 1, the ancestors of dolphins migrated from land to the
ocean approximately 50 million years ago during the Eocene era [19]
, with their limbs gradually
deteriorating. With the marine environment as the main driving force for evolution, the body
tn
contours of dolphins gradually evolved into the most suitable form for their own survival.
rin
ep
Pr
Fig. 1 The mismatch between dolphin evolution environment and actual engineering
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
Specifically, there are inevitable differences between the actual operating conditions of wind
turbines and the living environment of organisms in the natural world, such as fluid medium,
ed
incoming velocity, Length, Reynolds number, etc. The good aerodynamic performance of
organisms may only exist in certain local conditions that are suitable for their own survival
environment.
iew
Additionally, in the natural world, each organ or external limb is an essential component, and
each organ or system is closely interconnected to form an organic whole. From a purely
aerodynamic performance perspective, certain structures may have adverse effects on their
aerodynamic performance, even though these structures contribute to the overall biological
performance. Therefore, the bionic airfoil obtained directly through the reversed engineering may
v
exhibit negative effects such as reduced aerodynamic performance and unstable blade
performance under certain operating conditions. Therefore, the directly obtained bionic airfoil
re
needs to evolve again in conjunction with engineering practice.
1.2 Roughness sensitivity research of airfoil
The research on preventing the aerodynamic performance degradation of wind turbine blades
er
mainly focuses on the sensitivity of airfoil leading edge roughness [20]
. Blade leading edge
roughness is a crucial source of power loss in wind turbines. Thus, the roughness sensitivity of
airfoils is often a key factor in measuring the performance of blades. As shown in Fig. 2, when
pe
defects or gravel appear on the surface of the airfoil, resulting in an increase in airfoil roughness,
the aerodynamic performance of the airfoil will significantly decrease.
Corten et al. [21]
identified insect corpses adhered to the blade leading edge as the
fundamental cause of the sharp decline in aerodynamic performance of the blades, and found that
ot
blade contamination can reduce the efficiency of wind turbines by up to 50%. Quantitative
research conducted by Yang et al. [22]
on offshore floating wind turbines revealed that after ice
buildup, the lift coefficient of the airfoil decreased by 34% at most while the drag coefficient
tn
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
Regarding the issue of performance degradation caused by blade leading edge or local
roughness, some traditional flow control methods have achieved significant improvements. For
ed
example, Zhu et al. [26] studied the combined effect of passive vortex generators and rough airfoils
on dynamic stall phenomena, revealing that passive vortex generators can suppress flow
separation efficiently and delay dynamic stall of the airfoil. Papi et al. [27]
used micro Gunny flap
iew
to improve the aerodynamic performance degradation caused by blade roughness. The study found
that if the size of the micro Gunny flap was appropriate, the operating state of the defective blade
could be closed to that of clean and undamaged blades, and even slightly improve the performance
of blade under design conditions.
Although traditional active and passive flow control methods are still effective in improving
v
the flow field caused by rough leading edges of airfoils. However, this article believes that the
insensitivity of the airfoil itself to roughness is the fundamental solution to the above problems.
re
Airfoils that are insensitive to roughness can maintain good power output of wind turbines even if
there are roughness variations near their leading edges. On this condition, large-scale flow
separation on the blade surface will not occur, further ensuring the operational safety and stability
er
of wind turbines, so as to providing sufficient time for maintenance personnel. For instance, the
S809 airfoil, the NREL Phase VI horizontal axis wind turbine blade cross section, has the
capability to provide both high aerodynamic performance and low sensitivity to roughness [28].
pe
1.3 Airfoil numerical optimization
objectives, including airfoil thickness and aerodynamic performance. Similarly, Saleem et al. [30]
adopted genetic algorithms to redesign the NACA-9415 airfoil contour to develop a new wind
turbine with better performance in terms of blade torque and turbine power. Furthermore, Liu et al.
[31]
studied a multi-island genetic algorithm to change airfoil parameters and contour to achieve
ep
dynamic stall suppression effects. In addition, there are also similar studies, such as Song et al. [32]
optimizing the aerodynamic performance of the original airfoil based on machine learning
optimization algorithms, and Du et al. [33] using Adam algorithm to optimize the airfoil, etc.
Pr
Therefore, the critical characteristic of numerical optimization is that the changes in design
variables are determined by the optimization algorithm instead of the designer's experience .
[29]
This advantage serves to eliminate local defects that could be unfavorable in engineering reality
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
while also significantly reducing the necessary requirement for designer experience. In conclusion,
the optimization method of adjusting the original airfoil contour using optimization algorithms
ed
combined with engineering reality is highly applicable to the further evolution of the bionic airfoil
discussed in this paper.
1.4 Present work and the innovations
iew
In response to the optimization of surface flow field of wind turbine blades, non-sensitive
design for blade roughness, and the mismatch issue in the practical application of bionic methods,
this paper proposes a multi-objective optimization method based on optimization algorithms to
solve the problem of further evolution of the bionic airfoil.
The studies of the entire paper is summarized in Fig. 3, and can be briefly described as
v
follows:
re
(1) A surrogate model based on radial basic function (RBF) is established for predicting the
aerodynamic performance and roughness sensitivity of airfoils. The Hicks-Henne function method
is adopted to parameterize the Dol-Rot 24° airfoil [16]. High-credibility calculation method (Fluent)
is used to obtain the aerodynamic performance data of the corresponding parameterized airfoil.
er
This data is used as training samples to construct the RBF surrogate model to replace
high-credibility calculation method (Fluent).
(2) The Dol-Rot 24° airfoil profile is further evolved by using a multi-island genetic
pe
algorithm to obtain a new airfoil with high aerodynamic efficiency and insensitivity to
leading-edge roughness. By employing the Pareto multi-objective optimization design method,
based on the Dol-Rot 24° airfoil profile, a new airfoil profile with optimal performance and
meeting all constraints can be obtained as a novel airfoil profile for this study.
ot
(3) Aerodynamic efficiency verification and principle analysis are conducted on the
optimized new airfoil. A comparative analysis is conducted on the aerodynamic performance of
the optimized new airfoil (Dol-Rot 24° opt), Dol-Rot 24° airfoil, and S809 airfoil to verify
tn
whether the new airfoil designed based on the bionic optimization method proposed in this paper
can achieve the better flow field control effects without significant negative effects.
(4) The sensitivity of the optimized new airfoil (Dol-Rot 24° opt) to leading-edge roughness
rin
is verified and analyzed. The roughness sensitivity characteristics of the three airfoils are
compared. In addition, a new rough band model is proposed to further ensure the authenticity of
the roughness sensitivity optimization results of the airfoil.
(5) The energy conversion efficiency of the optimized new airfoil (Dol-Rot 24° opt) on wind
ep
turbines is also evaluated. Three-dimensional models of the optimized new airfoil (Dol-Rot 24°
opt), Dol-Rot 24°, and S809 are created according to the requirements of the NREL phase VI
horizontal-axis wind turbine. The energy conversion efficiency of these three blade designs is
Pr
compared to verify the effectiveness of the new airfoil designed based on the bionic re-evolution
method proposed in this study for flow field control.
The above five points are the research work of the entire paper, and the flow chart is shown in Fig. 3.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
v iew
re
Fig. 3 Research process
using numerical optimization methods, a bionic airfoil profile can be evolved into a new airfoil
suitable for practical operating environment of wind turbines.
(3) In terms of the further evolution of bionic airfoil, a set of surrogate model for
tn
airfoil optimization process, including the horizontal axis wind turbine model, roughness
approximation model, and optimization algorithm model.
2.1 Wind turbine model description
Pr
The NREL Phase VI horizontal axis wind turbine model completed by the US Renewable
Energy Laboratory is adopted [28]. The wind turbine is composed of two blades, each with an S809
airfoil as the basic unit of transverse cross-section, and a rotor rotation radius R is 5.029m, as
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
shown in Fig. 4 (a) below. Based on the blade model data provided by the experiment [28], the 3D
blade model is obtained as shown in Fig. 4 (b)
ed
v iew
re
(a) NREL Phase VI wind turbine blade [28]
er
pe
The chord length of all airfoils is denoted as c, as shown in Fig. 5. The roughness model used
in this paper refers to the research of Gharali et al. [25] and Ge et al. [20], which directly remove the
leading edge of the airfoil, as shown in Fig. 5. It should be noted that Han et al. [34]
found that
rin
erosion gradually expanded from the pressure side to the suction side on the stationary point.
Similarly, in the studies of Cappugi et al. [35]
and Papi et al. , there have been cases where the
[36]
rough area in the pressure surface is greater than that in the suction surface, as shown in Fig. 6.
Therefore, the roughness model used in this paper is shown in Fig. 5, where x=0.510% c is taken
ep
on the pressure surface and x=0.475% c is taken on the suction surface, in order to achieve a
roughness length slightly greater than that of the suction surface.
Pr
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
v iew
Fig. 5 S809 Airfoil and Roughness Approximation Model
re
er
pe
Fig. 7 demonstrates the further evolution process of bio-inspired airfoil combined with
practical engineering applications. This process utilizes the Dol-Rot 24° airfoil as the original
tn
design and aims to optimize the airfoil's aerodynamic performance and roughness sensitivity. The
research is mainly divided into two parts: building a surrogate model and global optimization of
airfoil based on multi-island genetic algorithms. The surrogate model refers to the approximate
mathematical model that replaces the complex and time-consuming numerical analysis [38]
in the
rin
optimization design process. The surrogate model is utilized in this paper to replace the traditional
CFD calculation, so as to quickly obtain the aerodynamic performance results of the original
airfoil and the defective airfoil. The overall steps of the bio-inspired airfoil further evolution are
ep
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
v iew
re
er
pe
number of parameters for input into the surrogate model, and the surrogate model will provide the
prediction results of the corresponding airfoil performance. This study adopts the Hicks-Henne
function perturbation method with high accuracy and stability . This method overlays various
[39]
rin
parameter shape functions on the original airfoil to achieve parameterized expression of the new
airfoil shape, and has been widely reported in recent years [40-42].
Divide the new airfoil into two parts: the upper surface and the lower surface, with the
leading and trailing edge points as the dividing points. The geometric shape of the wing profile of
ep
the new airfoil is represented by the reference airfoil, profile function, and control parameters,
respectively. The basic principle of the Hicks-Henne method is to add a change amount on the
longitudinal coordinate of the original airfoil profile shape.
Pr
10
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
Where y(x) represents the longitudinal coordinates of the new airfoil; y0 (x) is the longitudinal
coordinates of the original airfoil; n is the number of basis functions for the shape function; ci and
ed
fi(x) are the basis functions for the i-th control coefficient and its corresponding basic function,
respectively. The expression of the basis function is:
x 0.25 (1 x )e 20 x i 1
iew
f i ( x ) sin( x e (i ) ) 1 i n (2)
x 0.75 (1 x )0.1 e 20 x in
Where e(i ) ln 0.5 / ln xi , (0 xi 1) . In this case, n=5, with corresponding xi values of 0.1,
0.3, 0.5, 0.7, and 0.9 for the control points on both upper and lower airfoil. By perturbing the
v
shape function method, the shape of the new airfoil can be obtained by changing the control
coefficient ci.
re
In Fig. 8, based on the Dol Rot 24° airfoil, the fitting airfoil is obtained by utilizing the
Hicks-Henne method with control coefficients ci of 0. From Fig. 8, it can be seen that the fitting
airfoil profile is basically consistent with the reference airfoil profile (Dol-Rot 24°), exhibiting
good smoothness and continuity at all positions.
er
pe
ot
tn
During the process of optimizing the airfoil profile, thousands of CFD calculations are often
rin
required. To address this issue, surrogate model are applied to airfoil profile optimization. The
surrogate model is an analysis model with low computational complexity that can approximate the
calculation results of high-precision models. In airfoil profile optimization, the surrogate model is
ep
specific steps, as shown in Fig. 9. First, the airfoil is parameterized based on the Hicks-Henne
method, and the optimal Latin superelevation method is used to obtain initial sample points within
11
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
the range of design parameters. Second, one set of initial sample points is transformed into the
contour of the new airfoil, providing a model for subsequent mesh generation and flow field
ed
calculation. Then, defect treatment is carried out on the airfoil model, and the mesh division and
CFD flow field solutions of complete and defective airfoil are obtained to provide real response
values at the sample points. Finally, repeat the above steps and a surrogate model can be
iew
constructed by using the obtained sample points and their corresponding real response values.
The surrogate model fitting method adopted in this paper is the RBF model. Compared with
commonly used feedforward neural networks, RBF neural networks can determine the
corresponding network topology structure based on specific problems. In additional, the RBF
model has the ability to fitting quickly.
v
re
er
pe
After verifying the prediction accuracy of the surrogate model in this paper, it is found that
ot
the prediction accuracy basically meets the requirements. The judgment criteria mainly include the
following two points: (1) In Fig. 10, with Fluent's prediction results as the standard and Xfoil's
prediction results as the reference, the prediction error of the RBF surrogate model used in this
tn
paper is significantly smaller than Xfoil's prediction error. (2) If Xfoil is used to predict the
aerodynamic performance of an airfoil, poor input coordinate distribution may result in the
inability to predict the aerodynamic performance of the airfoil, which means that Xfoil software
rin
has high requirements for the distribution of new airfoil input coordinate points. Especially in the
process of predicting the aerodynamic performance of defective airfoils, it is easy to cause the
optimization program error, leading to the whole optimization procedure failure.
ep
Pr
12
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
v iew
re
Fig. 10 Comparison of prediction errors between RBF surrogate model and Xfoil
algorithm can avoid this situation and has high reliability. Therefore, the multi-island genetic
algorithm is utilized, and the control parameters are shown in Table 1.
Table 1 The control parameters of mult-island genetic algorithm
tn
Parameter Values
Sub-Population Size 10
Number of Islands 20
rin
Number of Generations 40
Rate of Crossover 0.9
Rate of Mutation 0.01
Rate of Migration 0.3
ep
Interval of Migration 5
Through the evolution, hybridization, migration, and mutation of multi-island genetic
algorithms, 10*20*40 optimization iterations are completed, achieving large-scale optimization.
Pr
This method can avoid taking the local optimal solution as the final result, thereby ensuring that
the optimization achieves the global optimal result.
13
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
2.3.4 Objective function and constraint conditions
ed
(1) Design point: Re=1.0×106,AoA=10°;
(2) Stall point: Re=1.0×106,AoA=18°;
When designing the airfoil, under the objective of high lift-to-drag ratio, the lift coefficient of
iew
the airfoil at the stall point should be increased while reducing the roughness sensitivity.
Meanwhile, it is necessary to consider the geometric compatibility and stall characteristics of the
airfoil. Therefore, the design objectives is summarized as follows:
(1) The objective at the design point should maximize the lift-to-drag ratio under free
transition condition, i.e., (cl/cd)free is maximized.
v
(2) At both the design point and stall point, the roughness sensitivity of the airfoil leading
re
edge should be minimized, i.e., (Ssf,l/d+Ssf,cl) is minimized.
Transition SST turbulent model can accurately simulate the friction distribution on the
surface of the airfoil and determine the position of the transition point [45]
. Therefore, it is chosen
to calculate the lift-to-drag ratio of the airfoil under free transition condition. The sensitivity of
er
airfoil leading edge roughness calculation is numerically simulated using actual defects, as shown
in in Fig. 5. The lift-to-drag ratio is defined as the quotient of lift coefficient and drag coefficient,
namely Cl/Cd. At the design point, the roughness sensitivity of the airfoil is expressed as:
pe
| (cl / cd ) max,free (cl / cd ) max,fixed |
S sf ,cl / cd 100% (3)
(cl / cd ) max ,free
(1) At the design point, the lift coefficient of the airfoil should be as large as possible under
free transition conditions;
(2) At the stall point, the lift coefficient of the airfoil should be as large as possible under free
rin
transition conditions;
(3) The maximum thickness of the airfoil is between 17.9% c and 20.6% c;
3. Sensitivity analysis and Validation
ep
Before conducting numerical calculations on all airfoils, the reliability of the numerical
simulation method is validated by verifying the lift and drag coefficient of the S809 airfoil.
Pr
14
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
3.1 S809 Static airfoil
The S809 airfoil is the basic unit of the transverse section of the NREL phase VI horizontal
ed
axis wind turbine. In addition, the S809 airfoil is widely used in previous researches and plenty of
related literature [46-48]. Therefore, the S809 airfoil is chosen as the validation object.
3.1.1 2D Numerical settings
iew
The topological structure of the computational domain and boundary conditions are shown in
Fig. 11, following the same setting as previous researches . As shown in Fig. 11, an O-type
[11]
v
radius of 50c (i.e. 50 times the chord length of the airfoil). The numerical simulation at different
re
angles of attack can be achieved by changing the incidence angle of the incoming flow.
The inlet of the computational domain is the velocity inlet, and the outlet is the pressure
outlet boundary condition. To meet the actual operating conditions of the NREL Phase VI
horizontal axis wind turbine, the Reynolds number takes 1.0 106, and the chord length c of the
er
airfoil is taken as the dimensionless length 1. Based on the inflow Reynolds number, the inflow
velocity is calculated to be 14.6m/s. To ensure y+<1, the height of the first layer of grid near the
wall is set to 2.3 10-5m, and the growth ratio of the normal grid near the wall is controlled within
pe
1.1, as shown in Fig. 12.
Due to the inability of SST and other two equation turbulence models to obtain the transition
position of the airfoil surface flow field, the Transition SST four equation turbulence model is
adopted in the airfoil optimization process. This turbulence model, proposed by Meter et al. [50]
,
ot
contains two transport equations of intermittence factor and momentum thickness Reynolds
number, and considers various transition mechanisms such as natural transition, bypass transition
and separation induced transition. This model can provide accurate results in predicting the
tn
15
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
v iew
re
Fig. 12 Grid distribution of the airfoil
er
3.1.2 Grid sensitivity analysis
To eliminate the impact of grid quantity on the numerical simulation results, the
pe
independence of grid quantity is verified by changing the number of nodes on the airfoil surface.
Fig. 13 shows the comparison of lift coefficient of airfoil at different grid quantities. It can be seen
that when the grid number exceeds 180000, the lift coefficient tends to be stabilize. For a balance
between accuracy and efficiency, the fourth set of mesh models was selected for subsequent airfoil
calculations.
ot
tn
rin
ep
To verify the correctness of the simulation process and the error rate of the simulation results
in this research, the simulation results (lift and drag coefficient) of the S809 airfoil will be
compared with the experimental results [28]
, Reinhart et al. , Ge et al.
[52]
, and Shi et al.
[20]
. The
[6]
16
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
comparison results are shown in Fig. 14. As illustrated in Fig. 14, it can be concluded that the
trend of the lift coefficient simulated in this paper is consistent with the experimental results [28]
ed
and closer to the experimental value than the lift coefficient simulation results of Reinhart et al. [52]
and Ge et al. [20]. In the case of small angles of attack, the simulation results of the drag coefficient
in this research are closer to the experimental values than the results of Shi et al. [6]
and Ge et al.
iew
. For high angles of attack, the simulation results of the drag coefficient in this research are
[20]
comparable to those of other authors, but there is a certain gap between them and the experimental
values.
In summary, the numerical calculation results in this research are more consistent with the
experimental values than the numerical simulation results of other authors. Therefore, the
v
numerical simulation method used in this research is deemed correct, and the error of the
numerical simulation results can also meet the requirements.
re
The above-mentioned grid is also used for the calculation of airfoil flow simulation in the
optimization process.
er
pe
ot
Similarly, to ensure the accuracy of the numerical simulation results of the three-dimensional
rin
wind turbine, verification work is carried out on the numerical simulation of the three-dimensional
wind turbine referring to the process of the two-dimensional airfoil mentioned above. The relevant
grid independence verification and comparison with experimental results have been elaborated in
previous published literature [11].
ep
the results obtained from applying these airfoils to horizontal axis wind turbines.
4.1 Airfoil optimization
17
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
By the established surrogate model, it takes about one minute to complete 10*20*40
optimization calculations. After removing the points where the airfoil thickness does not meet the
ed
requirements, the lift-to-drag ratio is extremely small, or the roughness sensitivity is high, a
filtered Pareto optimization solution set is obtained in Fig. 15. As shown in Fig. 15, the
distribution of the lift-to-drag ratio and roughness sensitivity are presented in this solution set.
iew
Among them, the lift-to-drag ratio is calculated at the airfoil design point (AoA=10°), and the
overall roughness sensitivity is the sum of the roughness sensitivity at the design point (AoA=10°)
and the stall point (AoA=18°), which is the sum of Formula 3 and Formula 4. In addition, Fig. 15
also includes the S809 airfoil and Dol-Rot 24° as references.
When selecting an optimized airfoil, the following principles should be followed: for
v
optimization based on the S809 airfoil, select the airfoil in the green box; for optimization based
on Dol-Rot 24 ° airfoil, select the airfoil in the yellow box. Since the yellow area is included in the
re
green area, the optimized airfoil is selected within the yellow area. In the yellow area, the
optimization target mainly considers the the lift-to-drag ratio value at the design, so the airfoil
with a larger lift-to-drag ratio value is preferred. After comprehensive consideration, the optimized
er
airfoil indicated by the red arrow in Fig. 15 is selected.
pe
ot
tn
Given the selected optimized airfoil, the corresponding airfoil parameters are applied to the
rin
Hicks-Henne shape function to generate the optimized airfoil profile shown in Fig. 16, named
Dol-Rot 24° opt.
ep
Pr
18
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
iew
Fig. 16 Comparison of airfoil profiles before and after optimization
v
4.2 Performance of static optimized airfoil
re
As shown in Fig. 17, the aerodynamic performance of S809 airfoil, Dol-Rot 24° airfoil, and
Dol-Rot 24° opt airfoil are compared. In Fig. 17, the lift coefficient of the Dol-Rot 24° opt airfoil
is significantly higher than that of the S809 and Dol-Rot 24° airfoils. Compared with the Dol-Rot
er
24° airfoil, the maximum lift coefficient of the Dol-Rot 24° opt airfoil is increased by 171.06%.
Note that the drag coefficient of the Dol-Rot 24° opt airfoil is slightly higher than that of the other
two airfoils at small angles of attack. However, at larger angles of attack, both the drag
pe
coefficients of the Dol-Rot 24° opt and Dol-Rot 24° airfoils are significantly lower than that of the
S809 airfoil. In terms of lift-to-drag ratio, the lift-to-drag ratio of the Dol-Rot 24° opt airfoil is
much higher than that of the other two airfoils. Compared with the Dol-Rot 24° airfoil, the
maximum lift-to-drag ratio of the Dol-Rot 24° opt airfoil is increased by 52.41%. It can be
predicted that using the Dol-Rot 24° opt airfoil as the profile of wind turbine blades will
ot
pressure contour in Fig. 18, the difference between the high-pressure area on the pressure side and
the low-pressure area on the suction side of the Dol-Rot 24° opt airfoil is always the maximum. In
addition, the integral area enclosed by the upper and lower surface pressure coefficients of the
19
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
Dol-Rot 24° opt airfoil is always the largest. This indicates that both the surface pressure contour
and the surface pressure coefficient results prove that the special profile of the Dol-Rot 24° opt
ed
airfoil can effectively increase the pressure difference between the upper and lower surfaces of the
airfoil, thereby increasing its lift coefficient.
v iew
re
(a1) S809 (a2) Dol-Rot 24° (a3) Dol-Rot 24° opt (a4) Pressure coefficient
(a) AoA=2°
er
pe
(b1) S809 (b2) Dol-Rot 24° (b3) Dol-Rot 24° opt (b4) Pressure coefficient
(b) AoA=6°
ot
tn
(c1) S809 (c2) Dol-Rot 24° (c3) Dol-Rot 24° opt (c4) Pressure coefficient
(c) AoA=10°
rin
ep
(d1) S809 (d2) Dol-Rot 24° (d3) Dol-Rot 24° opt (d4) Pressure coefficient
(d) AoA=14°
Pr
20
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
iew
(e1) S809 (e2) Dol-Rot 24° (e3) Dol-Rot 24° opt (e4) Pressure coefficient
(d) AoA=18°
v
In Fig. 19, the drag force of three airfoil is separated into pressure drag force and skin-friction
drag force. By combining the velocity streamline distribution of the airfoil surface and the skin friction
re
coefficient, the variation principles of the three airfoil's drag coefficient of are explored. The velocity
streamline distribution can well reflect the magnitude of the airfoil's pressure drag, while the skin
friction coefficient on the airfoil surface can prove the changes in skin-friction drag. As shown in Fig.
19, with the change in angle of attack, the skin-friction drag of the three airfoils remains relatively
er
constant, and the change in pressure drag is the main cause of the variation in airfoil drag coefficient.
pe
ot
tn
As demonstrated in Fig. 19, the change in pressure drag is the main cause of the variation in
rin
airfoil drag coefficient. Therefore, Fig. 20 presents the vorticity contour and surface flow
streamline of the three airfoils to investigate the principle of the changes in airfoil drag
coefficients. In Fig. 20, when the angle of attack is less than 10°, there is no apparent flow
separation occurring on the three airfoils, meaning that there are no obvious separation vortices on
ep
the airfoil surface and thus no significant changes in pressure drag for these airfoils. When
AoA=14°, the S809 airfoil exhibits an obvious flow separation vortex structure on its upper surface,
while the Dol-Rot 24° and Dol-Rot 24° opt airfoils maintain better flow control without showing
Pr
any separation vortices. When AoA=18°, all three airfoils exhibit obvious flow separation vortices,
with the Dol-Rot 24° bio-inspired airfoil demonstrating the best flow control ability.
Considering that the tip-speed ratio and lift-to-drag ratio are the determining factors affecting
21
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
wind turbine efficiency, the optimization direction for the Dol-Rot 24° opt airfoil leans more
towards improving the lift-to-drag ratio, relaxing the optimization of drag coefficient changes and
ed
flow separation vortex control under high-angle-of-attack conditions.
v iew
(a1) S809 (a2) Dol-Rot 24° (a3) Dol-Rot 24° opt
re
(a) AoA=2°
er
pe
(b1) S809 (b2) Dol-Rot 24° (b3) Dol-Rot 24° opt
(b) AoA=6°
ot
tn
22
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
iew
(d1) S809 (d2) Dol-Rot 24° (d3) Dol-Rot 24° opt
(d) AoA=18°
v
The roughness sensitivity is also taken as one of the optimization objectives. The Dol-Rot
re
24° opt airfoil, S809 airfoil, and Dol-Rot 24° airfoil are compared for their leading edge roughness
sensitivity.
Fig. 21 presents the roughness sensitivity of the airfoil with roughness segment entity at both
the design point (AoA=10°) and the stall point (AoA=18°), following the same conditions as in Fig.
er
5. For the y-axis roughness sensitivity factors at the design point (AoA=10°) and the stall point
(AoA=18°), Formula (3) and Formula (4) are utilized, respectively. Fig. 21 shows that the Dol-Rot
pe
24° opt airfoil has lower roughness sensitivity than other airfoils at both the design point and stall
point. This indicates that this research successfully achieved the roughness sensitivity optimization
goal of the Dol-Rot 24° opt airfoil.
ot
tn
rin
To validate the roughness sensitivity optimization for the Dol-Rot 24° opt airfoil further, this
section adopts the roughness patch length definition from Ge et al. [20]
and redefines the leading
edge roughness patch lengths for all three airfoils, as illustrated in Fig. 22. By varying the
configuration of roughness patches, the performance of Dol-Rot 24° opt airfoil's roughness
Pr
23
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
iew
Fig. 22 Schematic diagram of Erosion 2
Following the Erosion 2 definition in Fig. 22, the leading edge of the three airfoils undergoes
the same treatment. Fig. 23 shows the roughness sensitivity of the three airfoils under Erosion 2
v
conditions, where Dol-Rot 24° opt airfoil is still the least sensitive to roughness. The leading edge
roughness sensitivity of Dol-Rot 24° opt airfoil is successfully proved by using two different
re
roughness models.
er
pe
ot
This section aims to verify the optimization effect of the Dol-Rot 24° opt airfoil on low-speed
shaft torque for wind turbines. The S809 airfoil previously used in the NREL Phase VI wind
turbine blades is substituted with the Dol-Rot 24 ° airfoil and the Dol-Rot 24° opt airfoil,
rin
respectively. The low-speed shaft torque of the three blades is compared in Fig. 24 below. As
illustrated in Fig. 24, the low-speed shaft torque of Blade 2 and Blade 3 is notably higher than that
of Blade 1 across the entire range of wind speeds. Particularly, when the incoming wind speed
U∞=15m/s, compared with Blade 1, the low-speed shaft torque of Blade 2 and Blade 3 increased
ep
by 38.86% and 71.50%, respectively. For most wind speeds, Blade 3 has a better low-speed shaft
torque than Blade 2, and when the incoming wind speed is U∞=25m/s, Blade 3 has the most
significant improvement, with an increase of 61.55% compared to Blade 2.
Pr
24
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
v iew
Fig. 24 Low-speed shaft torque of three blades
re
According to the law of energy conservation, the fluid flow velocity distribution in the wake
of wind turbines can directly reflect the blade efficiency of energy absorption. When the fluid flow
velocity in the wake distribution is low, it can be inferred that the blade's energy absorption
efficiency is high. To investigate the changes in the energy absorption efficiency of the three
er
blades mentioned above, the wake velocity contour of the three blades at various wind speeds are
compared, as shown in Fig. 25.
In Fig. 25, It can be observed that when the incoming wind speed is 15m/s, the low energy
pe
area of Blade 2 and Blade 3 is significantly wider than that of Blade 1, indicating that Blade 2 and
Blade 3 can absorb more wind energy from the incoming flow. This phenomenon conforms to the
law of energy conservation, which means that the larger the extraction amount in the upwind area,
the lower the energy content of the downwind airflow. When the incoming wind speed increases
ot
to 25m/s, the wake situation of the three blades is similar to that at 15m/s. Compared with the
wake of Blade 2, the wake of Blade 3 is further optimized.
Based on the comparison of the aerodynamic performance of the three airfoils (S809 airfoil,
tn
Dol-Rot 24° airfoil, and Dol-Rot 24° opt airfoil) and the analysis of the related flow field changes,
this paper believes that the changes in the wake field of the three blades shown in Fig. 25 are
mainly due to two aspects: 1. the increase of the blade lift coefficient; 2. The flow separation on
rin
25
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
v iew
re
(a) U∞=15m/s
er (b) U∞=25m/s
red box in Fig. 26. The smaller separation vortex on the blade's surface indicates a significant
reduction in the flow separation area on the new blade surface, thereby improving the blade's
energy absorption efficiency. Combined with Fig. 25, due to the effective control of the trailing
tn
edge shedding vortices of the two new blades, Blade 2 and Blade 3 are considered to be able to
better absorb wind energy from the main flow.
rin
ep
Pr
26
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
v iew
re
(a) Blade 1 er (b) Blade 2 (c) Blade 3
The optimization of blade lift coefficient can be directly reflected by the pressure coefficient
pe
on the blade surface. The surface pressure coefficient of the blade section (airfoil) at 70%, 75%
and 80% elongation of the blade is selected for comparison when the inflow speed is 15m/s. As
shown in Fig. 27, compared with the S809 airfoil, the pressure coefficient difference between the
pressure and suction surface of the Dol-Rot 24° airfoil significantly increases. In addition, the
surface pressure coefficients of the Dol-Rot 24° opt airfoil are further optimized. This indicates
ot
that when the bio-inspired airfoil is combined with engineering practice for further evolution, its
performance is more superior.
tn
rin
ep
27
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
iew
(b) 75% blade height
v
re
er
pe
cross section of Blade 2 (Dol-Rot 24°) and the cross section Blade 3 (Dol-Rot 24° opt) can
effectively control the flow separation on the blade surface, which is consistent with the shedding
tn
vortex results on the three blade surfaces in Fig. 26. This indicates that the Dol-Rot 24° and
Dol-Rot 24° opt airfoils can better absorb more wind energy from incoming flow.
rin
ep
Pr
28
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
ed
iew
(b1) S809 (b2) Dol-Rot 24° (b3) Dol-Rot 24° opt
(b) 75% blade height
v
re
(c1) S809
er (c2) Dol-Rot 24° (c3) Dol-Rot 24° opt
(c) 80% blade height
pe
Fig. 28 Velocity streamline at different cross-sections for three types of blades when U∞=15m/s
5. Conclusions
In terms of the wind turbine airfoil design and the drawbacks in the traditional bionic
optimization process, this study developed a new bio-inspired optimization integrated airfoil
ot
design scheme using surrogate models and multi-island genetic algorithms. Furthermore,
aerodynamic performance, roughness sensitivity, and application research on horizontal axis wind
turbines are conducted.
tn
being completely suitable for practical engineering applications. To address this issue, this study
proposes targeting different goals and constraints related to engineering practical environments, so
as to adjust the bio-inspired airfoil profile by combining optimization algorithms. In this way, this
study obtained the Dol-Rot 24° opt airfoil suitable for the NREL Phase VI wind turbine.
ep
(2) Through the parametric modeling of Dol Rot 24° airfoil, approximate solution of
surrogate model and multi-island genetic algorithm optimization, an airfoil optimization platform
is built. This platform achieves integrated design from geometric modeling, mesh generation, to
airfoil performance analysis and optimization.
Pr
(3) From the perspective of airfoil aerodynamic performance, compared with the Dol Rot 24°
airfoil, the maximum lift coefficient of the Dol-Rot 24° opt airfoil increased by 171.06%, and the
maximum lift-to-drag ratio increased by 52.41%. From the perspective of the roughness sensitivity,
29
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
compared with the S809 and the Dol-Rot 24°,the Dol-Rot 24° opt airfoil is the best among the
three types of airfoils.
ed
(4) Three types of airfoils are modeled according to the parameters specified for the NREL
Phase VI wind turbine blades, denoted as Blade 1, Blade 2, and Blade 3, respectively, and their
power generation characteristics are compared. The results show that Blade 2 and Blade 3 have
iew
higher low-speed shaft torque than Blade 1. Moreover, under the vast majority of wind speed
conditions, the low-speed shaft torque of Blade 3 is better than that of Blade 2, with a maximum
improvement of 61.55%.
Future work will mainly focus on two aspects:
(1) Improvements in the bio-inspired airfoil optimization process: increase the diversity of
v
roughness models for airfoil leading edge. Adding multiple roughness models can provide a
broader optimization direction for the optimization of the roughness sensitivity of the new airfoil,
re
thereby further reducing the roughness sensitivity of the new airfoil in the actual operating
environment.
(2) Supplementary work of the new airfoil in the performance verification of wind turbines:
er
add the performance verification of wind turbine blade in the case of defects in the leading edge.
Acknowledgment
This study was supported by the Shanghai Rising-Star Program (Grant No. 19QC1400200).
pe
References
[1] Rezaeiha A, Montazeri H, Blocken B. Characterization of aerodynamic performance of vertical
axis wind turbines: Impact of operational parameters. Energy Conversion and Management,
2018,169:45-77.
ot
[2] Wood R J, Lu P. Leading edge topography of blades – a critical review. Surface Topography:
variation of the angle of attack in a vertical-axis wind turbine: Journal of Physics: Conference
Series, 2020. IOP Publishing.
[6] Shi X, Shuai X, Li D, et al. Passive flow control of a stalled airfoil using an oscillating
micro-cylinder. Computers & Fluids, 2018,178.
ep
[7] Li X, Zhang L, Song J, et al. Airfoil design for large horizontal axis wind turbines in low wind
speed regions. Renewable Energy, 2020,145:2345-2357.
[8] Zare J, Hosseini S E, Rastan M R. NREL Phase VI wind turbine in the dusty environment. arXiv
preprint arXiv:2304.06285, 2023.
Pr
[9] Aubrun S, Leroy A, Devinant P. A review of wind turbine-oriented active flow control strategies.
Experiments in Fluids, 2017,58:1-21.
[10] González-Salcedo Á, Croce A, Arce León C, et al. Blade design with passive flow control
30
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
technologies. Handbook of Wind Energy Aerodynamics, 2020:1-57.
[11] Huang S, Qiu H, Wang Y. Aerodynamic performance of horizontal axis wind turbine with
ed
application of dolphin head-shape and lever movement of skeleton bionic airfoils. Energy
Conversion and Management, 2022,267:115803.
[12] Yan H, Su X, Zhang H, et al. Design approach and hydrodynamic characteristics of a novel bionic
airfoil. Ocean Engineering, 2020,216:108076.
iew
[13] Zhang M M, Wang G F, Xu J Z. Aerodynamic control of low-Reynolds-number airfoil with
leading-edge protuberances. AIAA journal, 2013,51(8):1960-1971.
[14] Lin Y, Chiu P. Influence of leading-edge protuberances of fx63 airfoil for horizontal-axis wind
turbine on power performance. Sustainable Energy Technologies and Assessments,
2020,38:100675.
[15] Gao R, Chen K, Li Y, et al. Investigation on aerodynamic performance of wind turbine blades
v
coupled with airfoil and herringbone groove structure. Journal of Renewable and Sustainable
re
Energy, 2021,13(5):53301.
[16] Huang S, Hu Y, Wang Y. Research on aerodynamic performance of a novel dolphin head-shaped
bionic airfoil. Energy, 2021,214:118179.
[17] Song L, Tian K, Jiao X, et al. Design and optimization of seagull airfoil wind energy conversion
device. International Journal of Green Energy, 2021,18(10):1046-1063.
er
[18] Chen K, Yao W W, Wei J H, et al. Bionic coupling design and aerodynamic analysis of horizontal
axis wind turbine blades. Energy Science & Engineering, 2021,9(10):1826-1838.
[19] McCurry M R, Pyenson N D. Hyper-longirostry and kinematic disparity in extinct toothed whales.
pe
Paleobiology, 2019,45(1):21-29.
[20] Ge M, Zhang H, Wu Y, et al. Effects of leading edge defects on aerodynamic performance of the
S809 airfoil. Energy conversion and management, 2019,195:466-479.
[21] Corten G P, Veldkamp H F. Insects can halve wind-turbine power. Nature, 2001,412(6842):41-42.
[22] Yang X, Bai X, Cao H. Influence analysis of rime icing on aerodynamic performance and output
ot
[24] Ehrmann R S, Wilcox B, White E B, et al. Effect of surface roughness on wind turbine
performance.Sandia National Lab.(SNL-NM), Albuquerque, NM (United States), 2017.
[25] Gharali K, Johnson D A. Numerical modeling of an S809 airfoil under dynamic stall, erosion and
high reduced frequencies. Applied Energy, 2012,93:45-52.
rin
[26] Zhu C, Qiu Y, Feng Y, et al. Combined effect of passive vortex generators and leading-edge
roughness on dynamic stall of the wind turbine airfoil. Energy Conversion and Management,
2022,251:115015.
[27] Papi F, Melani P F, Alber J, et al. Potential of Mini Gurney Flaps as a Retrofit to Mitigate the
ep
CO.(US), 2001.
[29] Li X, Yang K, Bai J, et al. A new optimization approach to improve the overall performance of
thick wind turbine airfoils. Energy, 2016,116:202-213.
31
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
[30] Saleem A, Kim M. Aerodynamic performance optimization of an airfoil-based airborne wind
turbine using genetic algorithm. Energy, 2020,203:117841.
ed
[31] Liu J, Chen R, Lou J, et al. Deep-learning-based aerodynamic shape optimization of rotor airfoils
to suppress dynamic stall. Aerospace Science and Technology, 2023:108089.
[32] Song X, Wang L, Luo X. Airfoil optimization using a machine learning-based optimization
algorithm: Journal of Physics: Conference Series, 2022. IOP Publishing.
iew
[33] Du X, He P, Martins J R R A. Rapid airfoil design optimization via neural networks-based
parameterization and surrogate modeling. Aerospace Science and Technology, 2021,113:106701.
[34] Han W, Kim J, Kim B. Effects of contamination and erosion at the leading edge of blade tip
airfoils on the annual energy production of wind turbines. Renewable energy, 2018,115:817-823.
[35] Cappugi L, Castorrini A, Bonfiglioli A, et al. Machine learning-enabled prediction of wind turbine
energy yield losses due to general blade leading edge erosion. Energy Conversion and
v
Management, 2021,245:114567.
re
[36] Papi F, Balduzzi F, Ferrara G, et al. Uncertainty quantification on the effects of rain-induced
erosion on annual energy production and performance of a Multi-MW wind turbine. Renewable
Energy, 2021,165:701-715.
[37] Bartolomé L, Teuwen J. Prospective challenges in the experimentation of the rain erosion on the
leading edge of wind turbine blades. Wind Energy, 2019,22(1):140-151.
er
[38] Han Z H, Zhang K S. Surrogate-Based Optimization[M]. Real-World Applications of Genetic
Algorithms, 2012.
[39] Hicks R M, Henne P A. Wing design by numerical optimization. Journal of Aircraft,
pe
1978,15(7):407-412.
[40] Liu R, Hua Y, Zhou Z, et al. Prediction and optimization of airfoil aerodynamic performance
using deep neural network coupled Bayesian method. Physics of Fluids, 2022,34(11):117116.
[41] Yang G, Da Ronch A, Drofelnik J, et al. Sensitivity assessment of optimal solution in
aerodynamic design optimisation using SU2. Aerospace Science and Technology,
ot
2018,81:362-374.
[42] Tang X, Yuan K, Gu N, et al. An interval quantification-based optimization approach for wind
turbine airfoil under uncertainties. Energy, 2022,244:122623.
tn
by setting micro-plate in front of the blade leading edge. Renewable Energy, 2021,179:2309-2321.
[47] Stival L J, Brinkerhoff J R, Vedovotto J M, et al. Wake modeling and simulation of an
experimental wind turbine using large eddy simulation coupled with immersed boundary method
alongside a dynamic adaptive mesh refinement. Energy Conversion and Management,
Pr
2022,268:115938.
[48] Liu X, Lu C, Liang S, et al. Vibration-induced aerodynamic loads on large horizontal axis wind
turbine blades. Applied Energy, 2017,185:1109-1119.
32
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459
[49] Yi P, Wang Y, Sun X, et al. The effect of variations in first-and second-order derivatives on airfoil
aerodynamic performance. Engineering Applications of Computational Fluid Mechanics,
ed
2017,11(1):54-68.
[50] Menter F R, Langtry R B, Likki S R, et al. A correlation-based transition model using local
iew
[51] Langtry R B, Menter F R, Likki S R, et al. A correlation-based transition model using local
[52] Reinhart T, Medida S, Thomas S, et al. Computation of two-dimensional wind turbine airfoil
characteristics using advanced turbulence and transition modeling methods and a
GPU-accelerated Navier-Stokes solver: 32nd ASME Wind Energy Symposium, 2014.
v
re
er
pe
ot
tn
rin
ep
Pr
33
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4496459