Professional Documents
Culture Documents
www.springerlink.com/content/1738-494x(Print)/1976-3824(Online)
DOI 10.1007/s12206-015-0616-x
Aerodynamic shape optimized design for wind turbine blade using new airfoil series
Quan Wang1,*, Jun Wang1,*, Jin Chen2, Song Luo3 and Jinfeng Sun1
1
School of Mechanical Engineering, Hubei University of Technology, Wuhan, 430068, China
2
State Key Laboratory of Mechanic Transmission, Chongqing University, Chongqing, 400030, China
3
Clean Energy Research Institute, Department of Mechanical & Aerospace Engineering, University of Miami, Coral Gables, FL 33146, USA
(Manuscript Received February 12, 2014; Revised February 12, 2015; Accepted March 5, 2015)
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Abstract
This paper introduces a new airfoil series to optimize the aerodynamic shape of wind turbine blades. It is verified that the CQU-A air-
foil series exhibits high aerodynamic performance by using the wind tunnel experimental data and RFOIL. The geometry of a 2 MW
wind turbine blade with new airfoil families is designed preliminarily based on the shape of Tjaereborg 2 MW rotor. A multi-objective
optimized model combining the maximum power coefficient of the wind turbine with minimum area of the blade surface is proposed for
the pitch regulated wind turbine. An optimized code is developed based on the corrected blade element momentum (BEM) theory and
particle swarm optimization (PSO) algorithm. The optimization results show that, compared with that of the original rotor and the Tjae-
reborg rotor, not only the power coefficient and annual power production is improved, but also the area of the blade surface is reduced.
The decreased area indicates that the mass of the optimized blades is reduced. It is beneficial for increasing the fatigue life and reducing
cost of composite materials if the internal structure of the wind turbine blades is unchanged. Furthermore, the load of the blade root is
effectively controlled by using this alternative optimization program.
Keywords: Wind turbine blade; New airfoil series; Maximum power coefficient; Multi-objective PSO; Blade root load
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1
2. Aerodynamic model dT = B r cVrel2 F1Cn dr , (5)
2
Glauret [15] developed the original BEM theory that com- 1
dM = B r cVrel2 F1Ct rdr , (6)
bines blade element theory and momentum theory. From then, 2
the BEM equations and Prandtl tip loss correction have been
refined for many times to obtain better prediction of wind where (Cn, Ct) are the 2D force coefficients and F1 is the cor-
turbine performance [16]. relation between the 2D force coefficients and the 3D force
Fig. 1 shows a cross section of a rotor blade. The rotating coefficients on the blade. B is the number of blades. r is the
direction of the blade is clockwise. The axial velocity and the density of air. c is the chord length. The Cn, Ct and F1 are
Q. Wang et al. / Journal of Mechanical Science and Technology 29 (7) (2015) 2871~2882 2873
given as:
2 B( R - r )
F= cos -1[exp(- )] . (13)
p 2r sin j
Fig. 3. CQU-A airfoil series.
By Eqs. (5), (6), (11) and (12), the final formulas of the in-
duction factors become
2 + Y1 - 4Y1 (1 - F ) + Y12
a= , (14)
2(1 + FY1 )
1
b= , (15)
(1 - aF )Y2 / (1 - a ) - 1
-2.5 0.2
RFOIL RFOIL
Experiment 0.18 wind tunnel:smooth
-2
0.16
-1.5
0.14
-1
0.12
CP
CD
-0.5 0.1
0.08
0
0.06
0.5
0.04
1 0.02
1.5 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -10 -5 0 5 10 15 20
x/c a/o
Fig. 5. Comparison of the measured and predicted pressure coefficient Fig. 7. CQU-A180 measured and predicted drag coefficient for smooth
distribution at = 6, Re = 3.0106. surface flow at Re = 3.0106.
1.5
1.5
1
1
0.5
CL
CL
0.5
0
0
Fig. 6. CQU-A180 measured and predicted lift coefficient for smooth Fig. 8. A comparison of lift coefficients for CQU-A180 and NACA-
surface flow at Re = 3.0106. 64-418.
determine the wind tunnel reference pressures and estimate 1.690, which is practically identical to the predicted value
the turbulence level. The airfoil was equipped with 92 pres- (1.667) for smooth flow. The shape of lift coefficient (CL)
sure taps, all of which were placed along the chord at the cen- versus was approximately linear before maximum lift coef-
tre line of the model in a staggered alignment, so that distur- ficient (CL,max). The stall was well defined and the post-stall
bance from upstream taps could be minimized. The pressure region was smooth. The minimum CD was measured as
taps were carefully distributed in a manner through which the 0.0059, which agrees quite well with the predicted value of
expected pressure gradients can be adequately resolved. The 0.00592. There is a clear drop of CL in the post stall region, as
numerical predictions were calculated using RFOIL. can be expected due to the high value of CL,max. Although there
Fig. 5 shows the numerical and measured pressure distribu- is a little difference between the calculation and the test in the
tions of the CQU-A180 airfoil for smooth surface flow. The post-stall region, the trends of curve for lift coefficient and
measured and the numerical prediction showed desirable drag coefficient in deep stall are the same. Therefore, the re-
agreement. The suction side pressure distribution of this airfoil sults calculated by RFOIL are reasonable.
showed a significant peak at the leading edge, which facili- The comparison of the aerodynamic performance for the
tates the pressure recovery towards the trailing edge, and CQU-A180 airfoil and NACA-64-418 airfoil (The airfoils of
meanwhile, secures the airfoil against separation. Despite the Tjaereborg blade are composed of NACA airfoil series) is
some after-loading, most of the loading is distributed over the performed to demonstrate that the newly designed airfoil has a
leading edge. better aerodynamic characteristic. The free transition model is
Figs. 6 and 7 show the measured CL and CD of the CQU- used to simulate the smooth surface flow, while the fixed tran-
A180 airfoil for smooth flow (free transition) compared with sition model is introduced to simulate the rough surface flow
numerical prediction. The maximum CL was measured as (The transition points for the suction and pressure sides are
Q. Wang et al. / Journal of Mechanical Science and Technology 29 (7) (2015) 2871~2882 2875
Table 1. A comparison of the aerodynamic performance of airfoils. Table 2. Basic parameters of the rotor.
Airfoil Smooth condition Rough condition Rated Rated Number Direction Blade Design tip
CL,max L/D,max CL,max L/D,max power speed of blades of rotation length speed ratio
a ()
10
200
5
150
0
0 5 10 15 20 25 30 35
R (m)
100
50
NACA-64-418:free transition
-50 NACA-64-418:fixed transition
CQU-A180:free transition
CQU-A180:fixed transition
-100
-10 -5 0 5 10 15 20
/()
240
Table 3. Distribution of airfoils along the blade span-wise and bound-
ary conditions. 220
200
Relative Boundary
= r/R Airfoil
thickness condition 180
6
Re = 1.010 , 160
0.04~0.08 100% cylinder
CL/CD , max
Ma = 0.15 140
Transition Re = 1.0106, 120
0.10~0.20 100%~40%
airfoil Ma = 0.15
100 CQU-A150
Re = 1.5106, CQU-A180
0.20~0.25 40% CQU-A400 80
Ma = 0.15 CQU-A210
6
CQU-A250
Transition Re = 1.510 , 60
CQU-A300
0.25~0.30 40%~35%
airfoil Ma = 0.15 40 CQU-A350
CQU-A400
Re = 2.5106, 20
0.30~0.35 35% CQU-A350 0 1 2 3 4 5 6 7 8 9 10
Ma = 0.15
Re 6
6 x 10
Transition Re = 2.510 ,
0.35~0.40 35%~30%
airfoil Ma = 0.15 Fig. 13. Maximum lift/drag ratios with the change of Reynolds number.
Re = 2.5106,
0.40~0.45 30% CQU-A300
Ma = 0.15
Transition Re = 2.5106, with Reynolds number from 1.0106 to 8.0106 for the
0.45~0.50 30%~25%
airfoil Ma = 0.15 CQU-A airfoil series in smooth condition. It can be seen that
Re = 3.0106, the maximum CL/CD increases with Re number. It is important
0.50~0.60 25% CQU-A250
Ma = 0.15 to take this change into account for designing a wind turbine
Transition Re = 3.0106, blade. The largest change in CL/CD appears between Re = 1.0
0.60~0.65 25%~21%
airfoil Ma = 0.15 106 and Re = 3.0106.
Re = 3.0106,
0.65~0.70 21% CQU-A210
Ma = 0.15
Transition Re = 3.0106,
4.2 Optimization model
0.70~0.75 21%~18%
airfoil Ma = 0.15
The most important factor of optimization is to locate all
Re = 2.5106, important parameters and a suitable object function. From
0.75~0.80 18% CQU-A180
Ma = 0.15
earlier optimization studies of wind turbines [1-3], the most
Transition Re = 2.5106,
0.80~0.85 18%~15% convenient method is to maximize the power coefficient or to
airfoil Ma = 0.15
minimize the cost of energy. In this study, a multi-objective
Re = 2.5106,
0.85~1.00 15% CQU-A150 function with maximizing power coefficient and minimizing
Ma = 0.15
the area of curve surface for a 2 MW wind turbine is intro-
x 10
6 duced. Another important point is to choose a set of suitable
10
rotor speed=10(rpm)
design variables and constraint condition.
9 rotor speed=20(rpm)
rotor speed=25(rpm)
8
rotor speed=30(rpm) 4.2.1 Objective function
7 On a pitch regulated machine, it is possible to actively pitch
6 the entire blade and thus to change simultaneously the angles
of attack along its entire length. By pitching the entire blade it
Re
5
is possible to control the angles of attack and thus the power
4
output. This will make it possible to reach maximum power
3 coefficient Cp. Thus, one of the objective functions is taken as:
2
1 F1 ( X ) = max C p . ( ) (16)
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r/R According to the definition of the power coefficient, it can
Fig. 12. Reynolds number distributions along blade span with the be expressed as
change of rotor speed.
P
Cp = . (17)
has little influence on the aerodynamic performance for wind rV A / 2
3
{ }
F ( X ) = max m1F1 ( X ) + (1 - m1 ) - F2 ( X ) / k , (23)
8l 2 R
Cp = b(1 - a )r 3dr . (18)
R4 0 where 1 is weighting factor, 1[0, 1], and the optimization
result is better when it is set to 0.45. The scale factor k is set to
If the lift coefficient and drag coefficient are known, the Cp 100 to make sure that the two objective functions are the same
can be obtained by calculating the axial induction factor a and order of magnitude.
the tangential induction factor b. The aerodynamic characteris-
tic of the airfoils could be computed by RFOIL that is intro- 4.2.2 Design variables and constraints
duced in Sec. 3. To obtain a suitable shape of an MW-size wind turbine
An MW-size wind turbine blade with complicated curve blade, the geometry of the blade needs to be controlled by
surface is made of composite materials that lead to the differ- chord length, twist angle, airfoils, as well as rotational speed
ent density along its entire length. It is challenging to build the and pitch angle. The aerodynamic characteristic of the airfoils
mathematical model of blade mass by classical lamination are the lift and drag coefficients that depend on the angle of
theory. To simplify the model of the mass for wind turbine attack.
blades, the surface area may replace the mass prediction in the To illustrate the optimization model, the Tjaereborg 2 MW
case of the same internal structure. If the area for the blade rotor [5] with NACAXX airfoils is chosen for comparison.
curve surface is reduced, the mass of the rotor will be de- The initial geometry of the optimized model is based on the
creased. Therefore, another objective function is as follows: Tjaereborg rotor except for the new airfoil series, which is
called the original rotor. Once the chord and twist angle are
F2 ( X ) = min ( Ablade ) , (19) determined, the relative thickness is determined by spatial
n coordinate transformation. Therefore, the chord and twist an-
Ablade = B lim DS = B dS = B f (x,y) dr . (20) gle for each blade section can be seen as design variables.
l 0
i =1 S S
Fig. 14 shows the area calculating model of blade curve sur- X i min X i X i max i = 1, 2 . (24)
face. f(x, y) is the profile length function related to the position
of the blade section. It can be expressed as: To make sure that the designed blade exhibits high aerody-
namic characteristic and continuous smoothness of the surface,
m
a spline function is used to control the distribution of the chord
f (x,y) = (x i+1 -x i ) 2 + (yi+1 -yi ) 2 , (21)
i =1 and twist angle. Table 4 gives the constrained ranges of the
design variables.
where xi , yi are the three-dimensional coordinate data of the The total load has not only a normal component to the flow
blade section. According to the principle of blade geometry but also a tangential component in the rotational direction of
space coordinate transformation, (xi , yi) can be expressed: the blades. The tangential load component delivers the shaft
torque that turns the rotor. To characterize these loads it is
xi = c(x0 cos b + y0 sin b ) common to state the flapwise and the edgewise bending mo-
y = c(x sin b + y cos b ) , (22)
ments at a position close to the root of the blades. The flap-
i 0 0
3.5
the original chord length
3 the optimal chord length
2.5
C (m)
1.5
0.5
0
5 10 15 20 25 30
R (m)
12
the original twist angle
10 the optimal twist angle
q ()
6
0
Fig. 15. Flowchart of the optimized design algorithm.
5 10 15 20 25 30
R (m)
1 R V (1 - a ) wr (1 + b )
MT = r B 0 cCt rdr M T ,max , (26) 5. Results and discussion
2 0 sin j cos j
The optimization results converge after the number of itera-
where M T ,max is the maximum torsion that is taken to be the tions is up to 200. The optimization results are better when the
maximum thrust of the original rotor. weighting factor 1 is 0.25. Fig. 16 shows the comparison of
the distribution for the optimization chord length and the
4.2.3 Optimization algorithm original chord length. The distribution of the optimization
An optimized procedure for wind turbine blades can be chord length exhibits nonlinear characteristics compared with
achieved by combining PSO algorithm and BEM theory. the original one. Moreover, the chord length of each blade
Based on the improved BEM theory, giving a certain initial section is shorter than that of the original blade, especially in
condition and setting the tolerance, the axial induction factor the middle of the blade. This is because that one of the objec-
and the tangential induction factor may be computed from Eqs. tive functions is minimum area of the wind turbine blade sur-
(14) and (15). Then, the load on the blade and the power of the face. The area of the wind turbine blade is reduced by PSO
rotor can be obtained. A multi-objective PSO algorithm [20- algorithm. Fig. 17 shows the comparison of the distribution
22] is introduced to optimize the geometry of the new wind for the optimized twist angle and the original twist angle. The
turbine blade. new distribution of the optimization twist angle also performs
The basic parameters of PSO are: inertial factor w, 0.85; nonlinear compared with the one of the original twist angle.
learning factor, C1 = C2 = 0.5; variable dimension, n = 16; The twist angle at the root is larger than that of the original
population size, 50; the maximum iteration, 200; weighting blade. Whereas, the change of the twist angle near the tip is
factor, 0.25; and the scale factor, 500. The flow chart of the not significant. The blade coordinate data could be obtained
optimization procedure that combines a multi-objective PSO after the chord, twist angle and airfoils determined by coordi-
Q. Wang et al. / Journal of Mechanical Science and Technology 29 (7) (2015) 2871~2882 2879
Table 5. Comparison of the performance for the new design blade and 8000
Maximum power
0.47658 0.48866 0.50830 4000
coefficient Cp,max
Annual power 3000
production 4.224106 4.321106 4.406106
AEP (kWh) 2000
0 800
Tjaereborg rotor
-1
original rotor
700
30 optimized rotor
25 600
root torsion(KN.m)
20
500
15
400
10
300
5
0
2 1 0 200
(a) Original blade
0.5 100
2 3 4 5 6 7 8 9 10 11 12
0 tip speed ratio l
-0.5
Fig. 20. Comparison of the rotor torsion for the new design blade and
30
the traditional blade.
25
2.5
x 10 blade element momentum theory and PSO algorithm is devel-
oped. A novel shape of a 2-MW wind turbine blade with the
new airfoil series is designed by this optimization code. Re-
2
sults indicate that it is possible to boost the rotor power per-
formance by optimizing the chord and twist angle based on
Rotor power P(w)
1.5 the original rotor. The new rotor with new airfoil series not
only exhibits more maximum power coefficient, more annual
1
power production but also lower root load in comparison with
the Tjaereborg rotor and original rotor in normal operating
condition. Therefore, this novel optimization method with
0.5
Tjaere rotor new airfoil series is helpful for a modern MW-sized wind
Original rotor
Optimal rotor
turbine blade. However, the blade is designed and optimized
0
5 10 15 20
on a normal operating condition. If the stall stability and local
Wind speed U(m/s) stress concentration of the blade is considered, the optimized
Fig. 23. Comparison of rotor power. model will be more complicated. Therefore, in the next step,
these ideas will be considered in the optimization code.
[19] W. A. Timmer and A. P. Schaffarczyk, The effect of Quan Wang received his B.S. in Mecha-
roughness at high reynolds numbers on the performance of nical Engineering in 2007 from Yangtze
aerofoil DU 97-W-300Mod, Wind Energy, 136 (2004) 295- University, his M.S. and Ph.D. in Me-
30. chanical Engineering in 2009 and 2013,
[20] W. A. Timmer and V. T. Prjom, Summary of the delft respectively, from Chongqing University.
university wind turbine dedicated airfoils, J. of Solar En- He now does teaching and research related
ergy Engineering, 125 (2003) 488-496. with mechanical work. His research inter-
[21] J. Kennedy and R. Ebcrhart, Particle swarm optimization, ests cover mechanical system optimization
Proceedings of the IEEE Intemational Conference on Neu- design, aerodynamic and structural design of wind turbine blade.
ral Networks (1995) 1942-1948.
[22] Y. Shi and R. Eberhart, Modified particle swarm opti- Jun Wang is a professor in Mechanical
mizer. In Proceedings of the IEEE international conference Engineering at HuBei University of Tech-
on evolutionary computation, Piscataway, IEEE (1998) nology. He received his B.S. in Aircraft
69-73. Design from BeiJing University of Aero-
[23] A. Chatterjee and P. Siarry, Nonlinear inertia weight nautics and Astronautics (2002), and M.S.
variation for dynamic adaptation in particle swarm optimi- in Mechanical Engineering from Tenne-
zation, Computers and Operations Research, 33 (3) (2006) ssee Technological University (2007) and
859-871. his Ph.D. in Mechanical Engineering from
Tennessee Technological University (2010). His research interests
include mechanisms, robotics, manufacturing and renewable energy.