You are on page 1of 19

WIND ENERGY Wind Energ. 2005; 8:457475 Published online 31 March 2005 in Wiley Interscience (www.interscience.wiley.com). DOI: 10.1002/we.

153

Research Article

Tip Loss Corrections for Wind Turbine Computations


Wen Zhong Shen*, Robert Mikkelsen and Jens Nrkr Srensen, Department of Mechanical Engineering, Technical University of Denmark, DK-2800 Lyngby, Denmark Christian Bak, Wind Energy Department, Ris National Laboratory, DK-4000 Roskilde, Denmark

Key words: blade element momentum theory; tip loss correction; wind turbine aerodynamics

As an essential ingredient in the blade element momentum theory, the tip loss effect of rotors plays an important role in the prediction of wind turbine performance. Various tip loss corrections based on the Prandtl tip loss function are analysed in the article. Comparisons with measurements and theoretical analyses show that existing tip loss correction models are inconsistent and fail to predict correctly the physical behaviour in the proximity of the tip. A new tip loss correction model is proposed that remedies the inconsistency. Comparisons between numerical and experimental data show that the new model results in much better predictions of the loading in the tip region. Copyright 2005 John Wiley & Sons, Ltd.

Introduction
To take into account the difference between the physics of an actuator disc with innitely many blades and an actual wind turbine or propeller with a nite number of blades, Prandtl introduced the concept of tip loss. In an appendix to the dissertation of Betz,1 Prandtl showed that the circulation of a real rotor tends to zero exponentially when approaching the blade tip. The blade element momentum (BEM) theory was later developed by Glauert2 as a computational tool to predict aerodynamic loading and power. The theory is based on onedimensional momentum theory in which forces are distributed continuously in the azimuth direction, corresponding to an innite number of blades with no tip loss. In order to make BEM computations more realistic, Glauert2 showed how the tip loss effect is integrated in a simple manner into the BEM model. He corrected the induced velocity in the momentum equations by exploiting that the ratio between the average induced velocity and the induced velocity at the blade position tends to zero by the expression developed by Prandtl. Further, Glauert assumed that the tip loss only affects the induced velocity but not the mass ux. A rened tip loss model was later introduced by Wilson and Lissaman,3 who suggested that the mass ow through the rotor disc should be corrected in the same manner as the induced velocity in the wake. This, however, leads to a formulation in which the orthogonality of the induced velocity to the relative velocity at the blade element is not satised. In order to satisfy the orthogonality condition, de Vries4 rened further the tip correction of Wilson and Lissaman by correcting the mass ux in the tangential momentum equation in the same way as in the axial momentum equation. Apart from BEM computations, advanced axisymmetric actuator disc (AD) models based on solutions to the axisymmetric NavierStokes/Euler equations have been introduced.58 In these models the kinematics is described by rst principle methods. Corrections for tip loss are still needed but have to be introduced in dif-

* Correspondence to: W. Z. Shen, Department of Mechanical Engineering, Fluid Mechanics, Technical University of Denmark, Building 403, DK-2800 Lyngby, Denmark. E-mail: shen@mek.dtu.dk Contract/grant sponsor: EFP-01, Research Programme for Renewable Energy, Danish Energy Agency Contract/grant sponsor: European Commission; contract/grant number: NNE5-CT-2002-00627.

Copyright 2005 John Wiley & Sons, Ltd.

Received 31 March 2004 Revised 29 November 2004 Accepted 30 December 2004

458

W. Z. Shen et al.

ferent ways from those used for BEM models. Recently, Mikkelsen et al.7 proposed a technique to model the tip loss in which the induced velocities are rst computed by the NavierStokes solver, after which they are corrected with the Prandtl tip loss function in the rotor plane before applying aerofoil data. The technique, however, is in principle a variant of the Glauert tip loss correction. Looking more closely into the various models, we nd that they all lack rigorous consistency when the tip of the blade is approached. By analysing the basic equations of the tip loss correction, we nd that the computed axial interference factor always tends to unity when approaching the tip. This implies that the axial velocity, independently of the pitch setting, tip shape, aerofoil type and operating conditions, always tends to zero at the tip. Furthermore, comparisons of BEM computations with experiments show that the Prandtl/Glauert tip loss correction method overestimates the loading close to the tip. As a rst attempt to derive a consistent tip loss correction model and overcome the difculties of the classical modelling of the tip region, a new tip loss correction model was recently proposed by the authors.9 Preliminary computations of the NREL experimental rotor10 gave very promising results. The NREL experimental rotor runs at relatively low tip speed ratios, in the range from 15 to 54. In the present work we analyse existing tip loss correction models and describe a more general model that also treats high tip speed ratios. The article is organized as follows. First we describe the commonly used tip loss correction models and analyse their inherent inconsistencies. Second, a new tip loss correction model is described. Finally, numerical results are compared with experimental results for the NREL experimental rotor10 and the Swedish WG 500 turbine.11

Existing Tip Loss Correction Models


The BEM method is the most widely used technique for computing loads and power of wind turbines. The method is a one-dimensional approach based on the actuator disc principle (i.e. the forces are distributed evenly in annular elements without azimuth dependence). In order to simulate a real wind turbine with a nite number of blades, tip loss effects are introduced into the BEM method by using the Prandtl tip loss function.1 F= 2 B(R - r) 1 + l2 cos -1 exp p 2R

where l = WR/U is the tip speed ratio. The Prandtl tip loss function, which takes values in the range from zero at the tip to one at the root section, is derived by assuming that the wake consists of a system of straight vortex sheets. A more realistic and much more complicated model was introduced later by Goldstein,12 who used the inviscid screw surface structure of the wake to compute the circulation along an optimal rotor blade and compared it with the one obtained with the Prandtl tip loss function. For a four-bladed rotor, good agreement between the model of Goldstein and the Prandtl tip loss function is achieved. For a two-bladed rotor the results are in good agreement for tip speed ratios greater than l = 7, whereas large differences are found at smaller tip speed ratios (e.g. at l = 5 the difference is about 6%). An approximate formula for the Prandtl tip loss function was introduced by Glauert,2 namely F= 2 B( R - r) cos -1 exp p 2 R sin f R (1)

where fR is the ow angle at the tip. In order to make the formula easier to use in practical BEM computations, the tip loss function was further changed to F= 2 B( R - r) cos -1 exp p 2 r sin f (2)

where f = f(r) is the angle between the local relative velocity and the rotor plane. There are several ways of making tip loss corrections. In the following we summarize three well-known tip loss correction models and
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

Tip Loss Corrections

459

analyse the resulting expressions. Since all models are derivations of the model of Glauert, we discuss this model and its implications in detail. It should be noted that the following derivation is quite general and in principle covers all existing tip correction models.

BEM Analysis
In his pioneering work, Glauert2 deduced that the tip loss function F expresses the ratio between the average induced velocity and the induced velocity at the blade position, F = a /aB, where aB is the axial interference
2p

factor at the blade section and at the vortex sheet originating from the blade section, and a = 1 2p

adq and
0

a = a(r,q) are the average and local axial interference factors respectively. It should be noticed that, for a rotor with an innite number of blades, aB and a become identical. Further, Glauert introduced the tip loss as a cor rection to the interference factors aB and aB, using the Prandtl tip function F through the axial and angular momentum equations respectively. In the BEM theory the loading is computed using two independent methods, i.e. by a local blade element consideration using tabulated two-dimensional aerofoil data and by use of the one-dimensional momentum theorem. First, employing blade element theory, axial load and torque are written respectively as dT 1 2 = BFn = rcBVrel Cn dr 2 dM 1 2 = BrFt = rcBrVrel Ct dr 2 (3) (4)

where Fn and Ft denote the loading on each blade in the axial and tangential directions respectively, and Cn and Ct denote the corresponding two-dimensional tabulated force coefcients. From the velocity triangle at the blade element (see Figure 1) we deduce that sin f = U (1 - aB ) , Vrel cos f = Wr(1 + aB ) Vrel

Figure 1. Cross-sectional aerofoil element


Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

460

W. Z. Shen et al.

where the induced velocity is dened as Wi = (-aBU, -aBWr). Using the above relations, we get
2 Vrel = 2 U (1 - aB ) U (1 - aB )Wr(1 + aB ) = sin f cos f sin 2 f 2

Inserting these expressions into equations (3) and (4), we get


2 dT rBcU (1 - aB ) = Cn dr 2 sin 2 f 2

(5) (6)

dM rBcU (1 - aB )Wr 2 (1 + aB ) = Ct dr 2 sin f cos f Next, applying axial momentum theory, the axial load is computed as dT = dr
2p

r(U
0

- Uwake )UD rdq

(7)

where UD = U(1 - a) is the axial velocity in the rotor plane and Uwake = U(1 - 2a) is the axial velocity in the ultimate wake. Performing the integration and introducing the denition of the F function, recalling that a = a(q,r) and F(r) = a /aB, we get dT 2 = 4prrU aB F(1 - e 1a B F) dr where e1 = a2 aa (8)

Applying the moment of momentum theorem, we get dM = dr


2p

rV
0

q wake

UD r 2 dq

(9)

where Vqwake = 2Wra is the induced tangential velocity in the far wake. Performing the integration, we get dM = 4pr 3 rWU aB F(1 - e 2 aB F) dr where e2 = aa a a (10)

The Tip Loss Correction of Glauert


In the model of Glauert,2 only the induced velocity is corrected, hence e1F and e2F are both assumed to be unity. Equating equation (5) to equation (8), and equation (6) to equation (10), the nal expressions for the interference factors read aB = aB = 1 4 F sin f (sCn ) + 1
2

(11) (12)

1 4 F sin f cos f (sCt ) - 1

Copyright 2005 John Wiley & Sons, Ltd.

Wind Energ. 2005; 8:457475

Tip Loss Corrections

461

where s = Bc/2pr. The coefcients (Cn, Ct) are related to the lift and drag coefcients (Cl, Cd) by Cn = Cl cos f + Cd sin f and Ct = Cl sin f -Cd cos f respectively. (Cl, Cd) depend on the local aerofoil shape and are obtained using tabulated 2D aerofoil data corrected with 3D rotating effects. In the following we analyse the tip loss correction of Glauert, introduced in equations (11) and (12), assuming that the Prandtl tip loss function F tends to zero at the tip and that the considered wind turbine blade has a nite chord length at the tip (i.e. s 0). Introducing the tip loss correction equations (11) and (12) into the ow angle relation tan f = U(1 aB)/[Wr(1 + aB )], we get U Wr 1sCt 4 F sin f cosf U 4 F sin f cos f - sCt = tan f sCn Wr 4 F sin2 f + sCn 1+ 4 F sin2 f U 4 F sin f - s (Cl tan f - Cd ) tan f Wr 4 F sin f tan f + s (Cl + Cd tan f ) tanf T = 0 tanf T = Cd - Cl WR U Cd - Cl l = Cl + Cd WR U Cl + Cd l

tan f =

(13)

Introducing Cl and Cd into this equation, we obtain the expression tan f = (14)

Letting r R, noting that F 0 and s 0 at the tip, we get two solutions: (15a) (15b)

where fT denotes the ow angle at the tip of the blade. In the following, our discussion is divided into different cases according to whether the drag coefcient Cd is zero. 1. Assuming Cd 0, a simple limit value analysis of equations (11) and (12) shows that both equations (15a) and (15b), in general, exhibit the same tendencies: aB1 aB - 1 implying that the relative velocity becomes zero at the tip. From a physical point of view the tip vortex is created at the blade tip and then convected into the wake. Zero velocity at the tip means that the tip vortex stays at the tip. Thus, whether the solution satises equation (15a) or equation (15b), the resulting ow eld becomes unphysical near the tip. 2. On the other hand, since the drag force does not contribute to the induced velocity physically, Cd may be omitted when calculating induced velocities. In this case, Cd is put equal to zero in equations (11) and (12), and the expression for the ow angle, equation (14), reads tan f = U 4 F sin f - sCl tan f tan f Wr 4 F sin f tan f + sCl (16)

To analyse the solution at the tip (r R), four cases are considered. (a) fT tends to zero at the tip, i.e. the solution satises equation (15a). Cl is then obtained from aerofoil data and will generally not be zero at the tip. From equations (11) and (12) the interference factors at the tip read aB1 aB - 1
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

462

W. Z. Shen et al.

Table I. Implications of the Glauert tip loss correction model. In case 1, Cd is included, whereas it is neglected in cases 2(a)2(d) Parameters Case 1 Cd 0 1 -1 0 0 Any Case 2(a) Cl 0 1 -1 0 0 0 Case 2(b) F/Cl 0 1 -1 0 0 <0 Case 2(c) Cl/F 0 0 0 U WR >0 Case 2(d) Cl = hF Finite Finite Finite Finite Any

aT aT z Vrel,T q Vrel,T fT

(b) If F goes to zero at the tip with a speed faster than Cl (i.e. F/Cl 0), or Cl does not tend to zero, from equation (16) we get tan fT = -WR/U = -l. From equations (11) and (12) the interference factors at the tip read aB1 aB - 1 (c) If F goes to zero at the tip with a speed slower than Cl, (i.e. Cl/F 0), from equation (16) we get tan fT = U/(WR) = 1/l. From equations (11) and (12) the interference factors at the tip read aB0 aB 0 implying that the induced velocity is zero at the tip. From aerofoil data, however, Cl will in general be non-zero, implying a contradiction with the assumption that Cl is zero at the tip. (d) If F goes to zero at the tip with the same speed as Cl, (Cl = hF, where h is a constant that may depend on the tip speed ratio l), equation (16) reads 1= U 4 F sin f - sCl tan f 1 4 sin f - sh tan f = WR 4 F sin f tan f + sCl l 4 sin f tan f + sh (17)

From this expression the ow angle is seen to depend on the constant h and the tip speed ratio l. The corresponding interference factors at the tip can be computed from equations (11) and (12). The various cases and their implications are summarized in Table I. From the above considerations it is deduced that 2(d) is the only physically realistic case and that the Glauert correction is consistent only if Cl tends to zero with the same speed as F tends to zero. For a rotor with prescribed loading, as in the original case of Prandtl where an optimum circulation was assumed, a load distribution obeying Cl ~ F may be specied a priori. However, when using aerofoil data, this may not be the case. In order to analyse more closely the consequences of using the Glauert tip loss correction in combination with aerofoil data, an iterative procedure for solving equation (16) is constructed as follows: tan f n +1 = 1 4 F n sin f n - sCln tan f n tan f n l 4 F n sin f n tan f n + sCln (18)

where n is the iteration index. Now we analyse the situation near the blade tip. Choosing a point near the tip such that the Prandtl tip loss function Fn is sufciently small but not zero, we start with an arbitrary ow angle f0 which is the initial guess
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

Tip Loss Corrections

463

to the procedure (18). From tabulated aerofoil data we get in general a non-zero lift coefcient Cl. The relation (18) is then rewritten approximately as (as Fn goes to zero) tan f n +1 = 1 -sCln tan f n 1 tan f n = - tan 2 f n l l sCln
2 n -1

The solution of the above iteration can be written in the form tan f 0 tan f = - l
n

tan f 0

for n 1

Letting n , it is clear that fn 0 for all initial ow angles f0 satisfying the condition that |tan f0| < l. Thus, using tabulated aerofoil data, the solution inevitably converges towards zero ow angle and a nite lift coefcient at the tip. This situation corresponds exactly to case 2(a) above and implies that, independently of pitch setting, tip shape, aerofoil type or operating conditions, the relative axial velocity always tends to zero at the tip.

The Tip Loss Correction of Wilson and Lissaman


In the work by Wilson and Lissaman3 the concept of circulation was employed in order to reformulate the tip loss correction relations. Since circulation basically is generated by lift, they only considered the lift force in their analysis. For the tangential interference factor aB the obtained form is the same as that of the Glauert tip loss correction. For the axial induction factor aB, however, the mass ux is corrected in the same way as for the induced velocity. Thus e1 and e2F in equations (8) and (10) are both put equal to one and the nal formulae read aB F sCl = 1 + aB 4 cos f (19) (20)

(1 - aB F)aB F (1 - aB )
2

sCl cos f 4 sin 2 f

A similar analysis to the one done for the Glauert tip correction shows that, if the chord at the tip is not zero, the axial interference factor aB tends to one and the ow angle f tends to zero when approaching the tip, whereas the lift coefcient Cl does not tend to zero. Thus a similar inherent inconsistency exists for this model.

The Tip Loss Correction of de Vries


De Vries4 pointed out that the tip loss correction of Wilson and Lissaman does not satisfy the orthogonality between the induced velocity and the relative velocity at the blade element. As a consequence, de Vries put e1 equal to one and derived an expression for e2 that satised the orthogonality condition. It should be mentioned that he also derived high-order tip loss corrections in which expressions for e1 and e2 were computed by assuming distributions of a = a(q,aB). The nal formulae become aB F(1 - aB F) sCl = (1 + aB )(1 - aB ) 4 cos f (21) (22)

(1 - aB F)aB F (1 - aB )
2

sCl cos f 4 sin 2 f

In practice, the tip loss correction of de Vries gives almost the same results as those obtained by the original correction of Wilson and Lissaman. Further, it contains the same inherent inconsistency as the other two correction models.
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

464

W. Z. Shen et al.

New Tip Loss Correction Model


For a real rotor with a nite number of blades the axial velocity at the tip in the rotor plane, where the tip vortex is generated and convected into the wake, is usually not zero. Thus the ow angle at the tip, fR, is in general not zero. When applying blade element theory, two-dimensional aerofoil data give likewise a non-zero force near the tip, as the angle of attack generally is nite. From a physical point of view, however, the force should tend to zero at the tip owing to pressure equalization, where air is transported from the pressure side to the suction side of the blade. This shows that a correction is needed for the aerofoil data near the tip. Generally, in BEM computations, aerofoil data are corrected for three-dimensional and rotational effects.13 However, such corrections are of another nature than what is needed near the tip, where the ow is largely attached and rotational effects are negligible. In order to include three-dimensional tip loss effects, we propose r to compute the resulting force coefcients (denoted Cn and Ctr ) as
r Cn = F1Cn

(23) (24)

Ctr = F1Ct

where the function F1 is introduced as a correction to the two-dimensional aerofoil data in the tip region. Equating equation (5) to equation (8), and equation (6) to equation (10), putting e1 and e2 equal to one, as in the tip correction model of de Vries, and using equations (23) and (24), we get aB F(1 - aB F) sCt = F (1 + aB )(1 - aB ) 4 sin f cos f 1

(1 - aB F)aB F (1 - aB )
2

sCn F1 4 sin2 f

where Cn and Ct are the force coefcients obtained directly from two-dimensional aerofoil data. To determine a correction function, we note that F1 has to obey the following behaviour. First, in the case of a rotor with an innite number of blades, no correction is needed and it attains a value F1 = 1. Likewise, in the case of a rotor running at an innite tip speed, F1 should also be equal to one. Finally, when approaching the tip, F1 should tend to zero in the same manner as F, i.e. it should basically follow equation (17). Altogether, these features lead to a form of F1 that is similar to the one proposed by Glauert2 for F (equation (1)): F1 = 2 B(R - r) cos -1 exp - g p 2 R sin f R (25)

where g is a coefcient that is to be determined. In general, the coefcient g depends on number of blades, tip speed ratio, chord distribution, pitch setting, etc. For simplicity, the function is set to be dependent only on the number of blades and the tip speed ratio, assuming the following form: g = exp[ - c1 (Bl - c2 )] where c1 and c2 are two coefcients that have been determined from experimental data. Since there are two coefcients, only data at two different tip speed ratios are needed. In order to cover a broad spectrum of tip speed ratio, we employ the data of the NREL rotor at a wind speed of 10 m s-1, corresponding to a tip speed ratio of 379, and the Swedish WG 500 rotor at a tip speed ratio of 14. Comparing computed and measured distributions of normal force near the tip, a simple curve t shows that c1 0125 and c2 = 21. In order not to degenerate the formula in the case where l tends to innity, the coefcient g is shifted with a small constant of 01. Thus the nal form of the g function reads g = exp[-0125( Bl - 21)] + 01 (26)

The coefcient g can take more complicated forms than equation (26), but the general tendency would be similar. In future work we expect to obtain a more general form of the g function, when more experimental and numerical (CFD) data with different chord distributions and pitch settings are available.
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

Tip Loss Corrections

465

In accordance with the F function of Glauert,2 the function F1 can also be put in the form F1 = 2 B(R - r) cos -1 exp - g p 2 r sin f (27)

The F1 function gives the change in loading near the tip when two-dimensional aerofoil data are applied. In the new model, putting e1 and e2 equal to one, the nal formulae of the interference factors become aB =
2 2 + Y1 - 4Y1 (1 - F) + Y 1 2(1 + FY1 )

(28) (29)

aB =

(1 - aF)Y2 (1 - a) - 1

where Y1 = 4F sin2 f/(sCnF1) and Y2 = 4F sin f cos f/(sCtF1), with Cn and Ct being the force coefcients obtained directly from two-dimensional aerofoil data. It should be noted that the induced velocity in the new model is perpendicular to the relative velocity (i.e. tan f = Wr aB /(UaB)) if only Cl is responsible for the induction, in accordance with vortex theory. Including the drag in the computation of the induced velocity will jeopardize the orthogonality. When the axial interference factor aB becomes larger than approximately 03, the momentum theory is no longer valid and the so-called Glauert correction2 is needed. In the present work the Glauert correction is introduced into equation (8) by replacing the local thrust coefcient CT = 4aBF(1 - aBF) by a linear relation when aB becomes greater than a critical value ac. This results in the expression 4 aB F(1 - aB F) CT = 2 2 4[ac F + (1 - 2 ac F)aB F] where we put ac = 1/3. for aB ac for aB ac (30)

Numerical Results
In this section, computations using all four tip loss correction models will be carried out and compared with experimental data. For low tip speed ratios (l < 54) the NREL experimental rotor is chosen, whereas the Swedish WG 500 rotor is chosen for high tip speed ratios (l > 54).

NREL Experimental Rotor


As a rst consideration of the various tip corrections, the NREL rotor (NREL-S809) is chosen. The radius of the rotor is 503 m and it rotates at 7193 RPM. The blade sections consist of S809 aerofoils with a chord length of 03 m at the tip. The experiments were carried out in the worlds biggest wind tunnel at NASA Ames. The wind speed ranges from 7 to 25 m s-1, corresponding to l between 54 and 15. Further details can be found in Reference 10. As a pre-step to the BEM computations, two-dimensional aerofoil data were established from wind tunnel measurements.14 In order to construct a set of aerofoil data to be used for a rotating blade, the correction formula for rotational effects by Snel and van Holten13 is used for incidences up to 16. For higher incidences (>40), two-dimensional lift and drag coefcients of a at plate are usually too big and are reduced by a factor of 85%. For incidences between 16 and 40 the force coefcients are found by linear interpolation of the values at 16 and 40. The nal lift and drag coefcients at radial positions of 20%, 40%, 60%, 80% and 90% along the blade are plotted in Figure 2. The BEM computations are carried out using 30 blade elements distributed uniformly along the blade. Comparative BEM computations without tip loss correction and with the tip loss corrections of Glauert,2 Wilson and Lissaman3 and de Vries4 are rst computed. In Figure 3 the axial interference factor aB is plotted as a funcCopyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

466

W. Z. Shen et al.

2 20

20 Figure 2. Distributions of lift and drag coefcients for the NREL rotor at different radial positions

tion of radius. It is seen that the aB-values are almost identical in the inner part of the blade but diverge when approaching the tip. When using the model of Glauert, it is seen that the axial interference factor, independently of wind speed, tends to one, as predicted by our theoretical analysis. This value is much bigger than the value obtained without correction (aB = 017) or with the new tip loss correction model (aB = 023). Computed normal force distributions are compared with experimental data in Figure 4. We nd that all three classical models fail to predict the behaviour of the normal force in the tip region. In Figure 5, results using the new tip loss correction are compared with experimental data and with results using the tip loss correction of Glauert. The gure shows that the distribution of normal forces at r/R = 95% predicted by the new tip correction is in good agreement with experimental data for all wind speeds.
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

Tip Loss Corrections

467

Figure 3. Axial interference factor at blade section, aB, for ow past the NREL experimental rotor at a wind speed of 7 m s-1 computed by BEM code with Glauerts tip correction, with new tip correction and without tip correction

Finally, power curves computed with the new and with Glauerts tip loss correction are plotted against measurements in Figure 6. The new model predicts slightly better results than the Glauert model as compared with experimental data. Both models overpredict the power curve at wind speeds from 10 to 14 m s-1. A likely explanation is that the aerofoil data are subject to inaccuracies originating from the somewhat crude correction for rotating effects and for the inuence of Reynolds number. The power curve, however, is an integrated quantity consisting of contributions from all the local tangential forces. Thus only qualitative comparisons between experiments and computations are possible, since the results are inuenced by various factors such as disturbances from the hub and inaccuracies of the post-stall aerofoil data along the blade.

Swedish WG 500 Rotor


As a second consideration of the various tip loss corrections, the Swedish WG 500 rotor equipped with two STORK blades is chosen. The blade radius of the rotor is 2675 m. The blade sections consist of a series of NACA 44XX aerofoils with a chord length of 0125 m at the tip. The experiments were carried out in the 12 m 16 m wind tunnel at CARDC Low Speed Aerodynamics Institute in China. Further details can be found in References 11 and 15. From wind tunnel measurements,15 two-dimensional aerofoil data corrected for rotational effects are available for incidences up to 40. For incidences higher than 40, two-dimensional at plate data are used. The nal lift and drag coefcients at radial positions of 20%, 40%, 60%, 80% and 90% along the blade are plotted in Figure 7. The BEM computations are carried out using 40 blade elements distributed uniformly along the blade. First, results of normal force (i.e. normal to the aerofoil section) are compared with experimental data11 in Figure 8. We nd that the three classical models fail to predict the normal force in the tip region. Next, BEM computations with the new tip loss correction are performed. The resulting forces are compared with experimental data and with results obtained with the tip loss correction of Glauert in Figure 9. It is seen that the normal force distribution predicted by the new tip loss correction is in good agreement with experimental data for the three considered tip speed ratios.
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

468

W. Z. Shen et al.

Figure 4. Comparison of normal forces computed by BEM code with various classical tip corrections for ows past the NREL experimental rotor at different wind speeds: (a) 7 m s-1 (l = 54); (b) 10 m s-1 (l = 38); (c) 13 m s-1 (l = 29); (d) 15 m s-1 (l = 25)

Finally, thrust coefcient and power coefcient curves computed with the new tip loss correction and that due to Glauert are plotted against measurements in Figure 10. The new model is seen to predict slightly better results than the Glauert model as compared with experimental data. It should be noted, however, that, as in the case of the NREL rotor, the power curve is the integral of local tangential forces acting on the rotor, and only qualitative comparisons are possible.

Concluding Remarks
A new tip loss correction model for aerodynamic computations of wind turbines and propellers has been developed. The correction is mathematically consistent and is valid for a wide range of tip speed ratios. Experimental results from the NREL experiment at low tip speed ratios (l < 54) and the Swedish WG 500 rotor at high tip speed ratios (l > 54) are chosen to validate the tip loss correction models. Comparisons between experimental data and the various correction models show that the new model better predicts the aerodynamic force distribution in the vicinity of the tip. A continuation of the work will be to derive a more general form
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

Tip Loss Corrections

469

Figure 5. Comparison of normal forces computed by BEM code with new tip correction and with Glauerts tip correction for ows past the NREL experimental rotor at different wind speeds: (a) 7 m s-1 (l = 54); (b) 10 m s-1 (l = 39); (c) 13 m s-1 (l = 29); (d) 15 m s-1 (l = 25)

of the introduced g function and to extend the technique to cope with tip loss corrections for actuator disc and actuator line models.16

Acknowledgements
This work was supported partially by EFP-01, the Research Programme for Renewable Energy under the Danish Energy Agency and by the European Commission in the Research Programme of ENERGIE4-G2 in the STABCON project (contract NNE5-CT-2002-00627). The authors would like to thank Martin O. L. Hansen for discussing different issues during the preparation of the manuscript, and the reviewers for their helpful comments, especially for pointing out an error in the proof of the inconsistency in the original manuscript.
Copyright 2005 John Wiley & Sons, Ltd. Wind Energ. 2005; 8:457475

470

W. Z. Shen et al.

Figure 6. Power performance of the NREL experimental rotor

Appendix: Nomenclature
a,a aB, aB a, a ac B c CT Cl,Cd Cn,Ct F F1 M R r T U Vrel z Vrel q Vrel W f fR, fT l axial and tangential induced velocity interference factors axial and tangential induced velocity interference factors at blade section average axial and tangential induced velocity interference factors critical axial induced velocity interference factor number of blades chord thrust coefcient lift and drag force coefcients normal and tangential force coefcients Prandtl tip loss function new tip loss function of aerofoil data moment radius of blade radial distance from rotor centre thrust wind speed relative velocity axial relative velocity tangential relative velocity angular velocity of rotor ow angle ow angle at tip tip speed ratio

Copyright 2005 John Wiley & Sons, Ltd.

Wind Energ. 2005; 8:457475

Tip Loss Corrections

471

2 20

20 Figure 7. Distributions of lift and drag coefcients for the Swedish WG 500 rotor at different radial positions

Copyright 2005 John Wiley & Sons, Ltd.

Wind Energ. 2005; 8:457475

472

W. Z. Shen et al.

Figure 8. Comparison of normal forces computed by BEM code with various classical tip corrections for ows past the Swedish WG 500 rotor at different tip speed ratios: (a) 544; (b) 824; (c) 14

Copyright 2005 John Wiley & Sons, Ltd.

Wind Energ. 2005; 8:457475

Tip Loss Corrections

473

Figure 9. Comparison of normal forces computed by BEM code with new tip correction and with Glauerts tip correction for ows past the Swedish WG 500 rotor at different tip speed ratios: (a) 544; (b) 824; (c) 14

Copyright 2005 John Wiley & Sons, Ltd.

Wind Energ. 2005; 8:457475

474

W. Z. Shen et al.

Figure 10. Thrust and power performance of the Swedish WG 500 rotor

Copyright 2005 John Wiley & Sons, Ltd.

Wind Energ. 2005; 8:457475

Tip Loss Corrections

475

References
1. Prandtl L, Betz A. Vier Abhandlungen zur Hydrodynamik und Aerodynamik. Gttinger Nachr.: Gttingen, 1927; 8892. 2. Glauert H. Airplane propellers. In Aerodynamic Theory, Durand WF (ed.). Dover: New York, 1963; 169360. 3. Wilson RE, Lissaman PBS. Applied aerodynamics of wind power machines. Oregon State University Report NSF/RA/N74113, 1974. 4. De Vries O. Fluid dynamic aspects of wind energy conversion. AGARD Report AG-243, 1979; chap. 4: 150. 5. Srensen JN, Myken A. Unsteady actuator disc model for horizontal axis wind turbines. Journal of Wind Engineering and Industrial Aerodynamics 1992; 39: 139149. 6. Srensen JN, Shen WZ, Munduate X. Analysis of wake-states by a full-eld actuator disc model. Wind Energy 1998; 1: 7388. 7. Mikkelsen R, Srensen JN, Shen WZ. Modelling and analysis of the ow eld around a coned rotor. Wind Energy 2001; 4: 121135. 8. Masson C, Smaili A, Leclerc C. Aerodynamic analysis of HAWTs operating in unsteady conditions. Wind Energy 2001; 4: 122. 9. Shen WZ, Mikkelsen R, Srensen JN, Bak C. Evaluation of the Prandtl tip correction for wind turbine computations. Proceeding of 2002 Global Windpower Conference and Exhibition, Paris, 2002. http://www.ewea.org/ 10. Giguere P, Selig MS. Design of a tapered and twisted blade for the NREL combined experiment rotor. NREL Report SR-500-26173, National Renewable Energy Laboratory, Golden, CO, 1999. 11. Ronsten G, Dahlberg J, Meijer S, He DX, Chen M. Pressure measurements on a 535 m HAWT in CARDC 12 16 m wind tunnel compared to theoretical pressure distributions. Proceedings of European Wind Energy Conference and Exhibition (EWEC 89), Glasgow, 1989; 729735. 12. Goldstein S. On the vortex theory of screw propellers. Proceedings of the Royal Society 1929; 123A: 440465. 13. Snel H, van Holten T. Review of recent aerodynamic research on wind turbines with relevance to rotorcraft. AGARD Report CP-552, 1995; chap. 7: 111. 14. Hand MM, Simms DA, Fingersh LJ, Jager DW, Cotrell JR, Schreck S, Larwood SM. Unsteady aerodynamics experiment phase VI: wind tunnel test congurations and available data campaigns. NREL Report TP-50029955, National Renewable Energy Laboratory, Golden, CO, 2001. 15. Bjrck A, Ronsten G, Montgomerie B. Aerodynamic section characteristics of a rotating and non-rotating 2375 m wind turbine blade. FFA Report TN-1995-03, Aeronautical Research Institute of Sweden, 1995. 16. Srensen JN, Shen WZ. Numerical modelling of wind turbine wakes. Journal of Fluids Engineering 2002; 124: 393399.

Copyright 2005 John Wiley & Sons, Ltd.

Wind Energ. 2005; 8:457475

You might also like