Shape Optimum Design of Horizontal Axis Wind Turbine in Low Reynolds Number Range

Yukio Watanabe1, Hidetsugu Iwashita, and Masamitsu Ito

1
Department of Maritime Technology, Toba National College of Maritime Technology,
1-1 Ikegami-cho, Toba, 517-8501 Japan,
Phone: (+81) 599-25-8067, Fax: (+81) 599-25-8016, e-mail: [email protected]
2
Department of Social and Environmental Engineering, Hiroshima University
3
Department of Maritime Technology, Toba National College of Maritime Technology,

This study is concerned with a shape optimum design of a horizontal axis wind turbine (HAWT) blade
operating in low Reynolds number range (Re < 106). The objective is to maximize the power coefficient CP, rotor
power per air flow momentum through the rotor disk area, under the constraints on the torque and the angle of attack
in terms of chord length and twist angle distributions where the power-required is evaluated by boundary element
method (BEM). The profile drag is also considered in the optimization, because the effects cannot be ignored in a low
Reynolds number range. For calculation efficiency, the profile drag coefficient database is provided by B-spline
interpolation about Reynolds number and angle of attack in advance. On the other hand, BEM obtains the high
accuracy of aerodynamic prediction due to discrete sufficiently. However computational cost is too high to estimate
many design variables. Therefore B-spline approximation is adapted to the design variables, and the number of
design variables is decreased, the number of panel discretization is increased. In this study, we develop the new
design tool to use the aerodynamic analysis tool involving viscous and three-dimensional effects for prediction of the
small HAWT performance.

Nomenclature

A = Area of the rotor disk

c = Blade section chord length
CD = Drag coefficient
CL = Lift coefficient
CP = Power coefficient: 2Q ω/ρU 3A
CQ = Torque coefficient: 2Q/ρU 2AR
CT = Thrust coefficient: 2T/ρU 2A
D = Rotor disk diameter
D’ = Drag force per unit radius
Nb = Number of blade
Q = Rotor generated torque
R = Rotor disk radius
Re = Reynolds number: cW/ν
r = Radial coordinate
SH = Surface of the rotor disk
SW = Surface of the wake
T = Rotor generated thrust
U = Incident flow velocity
W = Local total velocity
x,y,z = Cartesian coordinates
α = Angle of attack
β = Blade twist angle
Γ = Circulation
ν = Kinematic viscosity
λ = Tip speed ratio: ωR/U
ρ = Fluid density
φ = Velocity potential
φ’ = Basic flow angle: arctan(U/ωr)
ω = Rotor angular velocity

Subscripts/Superscripts
TE = Trailing edge
+,- = Upper and lower surface
U,L = Upper and lower limit
 Introduction

A HAWT is a device for extracting kinetic energy from the wind, and it has caught on worldwide. Until now,
there has been an increased reliance on the use of computational methods to analyze and design HAWT. Many rotors
found on current generation HAWT systems are designed using combination of 2-D airfoil wind tunnel data and
combined Blade Element Momentum Theory (BEMT)[1]. A number of comprehensive computer codes using this
methodology are currently available. While computationally efficient and highly useful, these methods accurately are
incapable of modeling three-dimensional effects[2].
The small HAWT operates in low Reynolds number flow field affected strong viscous effects, and it is difficult
to predict the wind turbine performance. To solve this problem, it is considered to simulate fully three-dimensional
viscous flow field around HAWT[3]. However, the computation times and costs involved when employing the
Navier-Stokes flow solvers, even with the latest generation of supercomputers, remain prohibitive for routine work in
the applications of present study. On the other hand, though several hybrid methods, for example, CFD and vortex
panel method coupling[4] or BEM and boundary layer theory coupling[5], are suggested, these are not widely used.
In this present paper, the aerodynamic analysis tool using boundary element method and the database of 2-D
aerofoil profile drag is applied to predict aerodynamic performance of small HAWT. The viscous region surrounding
the blades is modeled using the 2-D boundary layer theory, and the inviscid region at a distance from the blade
boundary layer is modeled using potential flow, boundary element method. The accuracy of the aerodynamic analysis
for the small HAWT is verified by comparing the calculated wind turbine performance and measurement data of the
wind tunnel experiment. In this study, we develop the new design tool to use the aerodynamic analysis tool involving
viscous and three-dimensional effects for prediction of the small HAWT performance.

Methodology

Boundary Element Method

Wind turbine blade rotating at constant angular velocity ω is subject to the inflow velocity U. With the
assumption that the fluid surrounded the wind turbine is incompressible, inviscid and irrotational, there exists a
perturbation velocity potential φ which satisfies the Laplace equation.
The kinematics boundary condition should be
satisfied on the wind turbine blade surface SH.

[L] " 2# = 0 in Fluid domain

(1)
$#
[H] = %WI • n on S H
$n

The wake surface SW is assumed to have zero

thickness. The normal velocity jump and the pressure
! jump across SW is zero, while a jump in the potential is
allowed.

("p) onSW
(
= p + # p# = 0 )
(2)
& $% ) & $% ) & $% )# +

(" + = ( + #( + = 0
' $n * onSW ' $n * ' $n *
Fig.1 Coordinate Systems

The perturbation potential φ at the field point, P(x,y,z), is given by the three-dimensional Green formula.
!
"(P) %G( P,Q) % " (Q ) %G( P,Q)
# $$ " (P ) dS = # $$ G( P,Q) dS + $$ & (QTE ) dS (3)
2 SH %n SH %n SW
%n +

!
Green function, G(P,Q), is defined as

1 (4)
G(P,Q) = , r = ( x # x $ )2 + ( y # y $ )2 + ( z # z $ )2
4" r

and Q(x’,y’,z’) is field points. n is a unit normal vector to the integration surfaces with the positive direction pointing
into the field domain. Γ (QTE) is the potential jump at the trailing edge.
The !Kutta condition requires that the velocity at the trailing edge of the blade be finite. In the numerical
formulation of the problem, we will implement the Kutta condition by requiring that the pressures at the upper and
lower control points at the trailing edge are equal. This can be expressed as follows:

"p # p + $ p $ = 0 at trailing edge (5)

A direct solution of the resulting system of equations (3) and (4) is difficult due to the nonlinear character of equation
(3). Therefore an iterative solution algorithm, Newton Raphson method, is employed [5]. The initial guess, Γ=φ+ -φ -,
is the approximate Kutta condition for
! a potential flow panel solution, first introduced by Morino[6].

Profile Drag

The profile drag of a HAWT blade is considered in the optimization, because the variations of Reynolds
number affect to the HAWT in low Reynolds number range. In the present method, XFOIL[7], viscous flow airfoil
performance evaluation software, is adopted to estimate the profile drag of the HAWT blade.
In order to develop the aerodynamic analysis tool, 2-D airfoil performance database is prepared to use for
XFOIL, first. Next, local Reynolds number (Rel=cW/ν) at each radial position and Lift coefficient CL computed BEM
is evaluated. Finally, the angle of attack α3D involving three-dimentionsl effects and the profile drag coefficient
CDprofile at each radial position are interpolated due to input Rel and CL.

1 (6)
D"profile = # CD profile dsW 2
2

Substituting Eq. (6) into the profile drag coefficient CDprofile, we obtain the profile drag D’profile. The relations for
thrust Tprofile and torque Qprofile are
!
Tprofile = D"profilesin# " , Q profile / r = "D#profile cos $ # (7)

Figure 2 shows the relation between the HAWT profile drag and airfoil coefficient CL and CDprofile.

! !

Fig.2 Overview of profile drag

Results and Discussions

Computational Test Case: Wind Turbine Performance Prediction

In this study, the small HAWT model that specifications of rotor blade are shown in Table 1 is targeted. Figure
4 shows chord length and pitch angle distributions of wind turbine test model designed by BEMT. The experiment
data of the small HAWT model ware measured in the Hiroshima University wind tunnel[8]. Figure 3 shows overview
of wind tunnel experiment.
 Table1 Specifications of Rotor Blade

Diameter, D 1.0
Boss ratio 0.1
Number of blade, Nb 3
Design wind speed,U 10
Design tip speed ratio, λ 6
Wing section NACA642-415
Reynolds Number [at r/R=0.7] 1.2250×105

Fig.3 Wind Tunnel Test Fig.4 Blade Twist Angle and Chord Length Distributions

Profile Drag Effects for the Wind Turbine Performance

In order to validate the aerodynamic analysis tool comparison with the wind tunnel experimental data was done.
Figure 5 shows comparison of the wind turbine performance, CT, CQ and CP between experimental result and
calculated result. In figure 5, symbols for experimental result: curve lines; BEM calculated result: circles without
viscous effects; the inviscid viscous interaction analysis result: deltas.
The experimental result and calculated result including the profile drag show good agreement in the CT, CQ. In
the power coefficient CP, very important factor for optimization, some differences between calculated inviscid flow
results and experimental results are observed though the accuracy of calculated result seems to reach enough level by
considering viscous effects. And it can be seen that the power coefficient C P is overall larger in magnitude for
inviscid flow. To simulate the performance of a small HAWT operating in low Reynolds number flow field need to
consider the profile drag.

Fig.5 Comparison of Wind Turbine Performance, CT, CQ and CP

Shape Optimum Design

It is well known that modern wind turbine blades are designed with very complex optimality criteria involving
more than aerodynamic efficiency. In this work, the objective is to maximize the power coefficient CP to be varied
the constraint variables; chord length c and blade twist angles β. While blade twist angle distributions to satisfy
design attack of angles are determined in the BEMT. In the present method, the distributions of the blade twist angle
can be determined directly not to specify the attack angles at each radial position.
On the other hand, BEM obtains the high accuracy of aerodynamic prediction due to discrete sufficiently.
However computational cost is too high to estimate many design variables. To adopt B-spline approximation about
the design variables, the number of design variables can be smaller and the number of panel discretization can be
larger.
The rotor generated torque Q is required to be larger than the resistance of electric generator Qg in general.
Furthermore, chord length c is constrained adequate range because of structural reason. Since BEM cannot compute
the aerodynamic stall or separation flow, the effective angle of attack is constrained not to stall. Therefore, the
optimum problem is expressed as follows.

Maximize : CP (ci , " i )

subject to : Q(ci , " i ) # Qg (8)
L U
$ % $ j (ci , " i ) % $
j ( j = 1,2,L)
j
L U
c % ci % c
i i (i = 1,2,L)

Result of Design Procedure

!
In order to validate the design tool, a small HAWT design is performed. The small HAWT model is the same
specifications model. In this work, the rotor generated torque Q must be more than zero.

Optimum Design

A number of 6 discretization points are constrained as β and c respectively. In the aerodynamic analysis, to
estimate CP requires at least 15 discretization points. It is already mentioned, at each radial position not to be
specified along the HAWT blade, using B-spline interpolation provides β and c. The design is initialed with the
specified conditions showed Table. 1.

Fig. 6 Design Procedure Image

The optimum design procedure adopted is FFSQP (FORTRAN Feasible Sequential Quadratic Programming)
[9]. FFSQP is a set of FORTRAN subroutines for the minimization or the maximum of a set of smooth objective
functions (possibly a single one or none at all) subject to nonlinear equality and inequality constraints, linear equality
and inequality constraints, and simple bounds on the variables. The Sequential Quadratic Programming algorithm is a
generalization of Newton’s method for unconstrained optimization in that it finds a step away from the current point
by minimizing a quadratic model of the problem.
 Constrained conditions of c and α are set, 10 mm < c < 100 mm, -2.0deg. + α0 < α + α0 < 2.0deg. + α0
(α0=β0-φ’0; β0, φ’0: initial blade twist and flow angle), for manufacturing and aerodynamic analysis. Figure 6 shows
design procedure image. Figure 7 describes computational iteration steps of the effective angle of attack. The
variables α1 ~ α6 in figure 7 are the effective angles at the radial positions, r/R=0.20, 0.30, 0.60, 0.80, 0.90, 1.00,
respectively. Shown in figure 7, iterative processes converge about 10 steps.
The resulting optimum design is shown in figure 8, which gives for each radial station: blade chord length and
twist angle. In figure 8, symbols are the initial design condition, and curve lines are the optimization result of present
design tools calculation. By nature, the tip chord length designed by BEMT is zero. However, the value of tip chord
length is not zero, and the blade shape result curve line is smooth. After optimization, power coefficient CP improves
about 4% compared with the initial condition.
The twist angle of optimum blade is larger at the tip and smaller around middle part of blade than the initial
blade twist angle. The chord length c converges to 10 mm at blade tip. Shape optimum design results indicate that
power coefficient CP is affected due to the variations of β and c around blade tip.

Fig.7 Iteration Steps of Effective Angle Fig.8 Comparisons of Optimum Design Result
and Initial Condition

Conclusions

In this study development of the shape optimum design tool of a small HAWT blade considering the profile
drag and three-dimensional effects is presented.
The accuracy of the aerodynamic analysis tool adopted to design procedure is verified by comparing the
calculated wind turbine performance and measurement data of the wind tunnel experiment. As a result of the shape
optimum designs, three-dimensional blade shape directly is provided and maximized power coefficient CP improves
about 4%. Therefore, the validity of the shape optimum designs tool is verifiable. In the future work, this design tool
will adopt comprehensive HAWT design including structural analysis.

References

1. Glauert H. Airplane Propellers, Ed. By Durand, W. Aerodynamic Theory, Div. L, Vol. 4, Peter Smith, Glouceste,
MA, 1976; 169-269.
2. Adkins CN, Liebeck RH. Design of Optimum Propellers, AIAA Aerospace Sciences Meeting Januar. 1983 ;
Reno/Nevada.
3. Sorensen NN, Michelsen JA, Schreck S. Navier-Stokes Predictions of the NREL Phase VI Rotor in the NASA
Ames 80 ft x 120 ft Wind Tunnel, Wind Energy, Vol. 5, No. 2-3, 2002; 151-169.
4. Schmitz S, Chattot JJ. A Parallelized Coupled Navier-Stokes/Vortex-Panel Solver, Journal of Solar Energy
Engineering, 2005; 127: 475-487.
5. Watanabe Y, Iwashita H, Ito M, Numerical Analysis of the Small Horizontal Axis Wind Turbine Performance
Taking Account of Viscous Effects, Renewable energy 2006 proceedings; Makuhari, 2006; 952-955
6. Kerwin JE, Kinnas S, Lee J-T, Shin W-Z. A Surface Panel Method for the Hydrodynamic Analysis of Ducted
Propellers. SNAME Transactions, 1987; 95:93-122.
7. Kuo CC, Morino L. Steady Subsonic Flow around Finite-Thickness Wings, NASA CR-2616, 1975
8. Drela M. XFOIL: An Analysis and Design System for Low Reynolds Number Airfoils, T. J. Muller, ed.,
Springer-Verlag, Lecture Notes in Engineering, 1989; 54
9. Yanagihara D, Iwashita H, et al.. Development of an Elastic Rotor-Blade for the Horizontal Axis Wind Turbine,
Proceedings of Wind energy symposium, 2004; 26:187-190.
10. Zhou JL, Tits AL, Lawrence CT. User’s Guide for FFSQP Version3.7: A FORTRAN Code for Solving
Constrained Nonlinear (Minimax) Optimization Problems, Generating Iterates Satisfying All Inequality and Linear
Constraints, Institute for System Research University of Maryland, SRC TR-92-107r2, 1997