You are on page 1of 11
Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. A98-16846 ATAA-98-0021 A 3-D Stall-Delay Model for Horizontal Axis Wind Turbine Performance Prediction ZHaonut Dut aND MICHAEL S. SELIG* Department of Aeronautical and Astronautical Engineering University of Ilinois at Urbana and Champaign Urbana, Ilinois 61801 Abstract Most design and analysis methods widely used for horizontal axis wind turbine performance predic- tion, such as the PROP code, are based on the traditional 2-D blade element/momentum theory (BEMT) methods, which are inadequate and under- predict the wind turbine rotor power output in the high-wind/peak-power condition, owing to effects of rotation on the wind turbine blade boundary layer. Although the deficiencies of the methods have been known for some time, this area has been ne- glected. The continued development of viable and well-established stall-regulated wind-turbine tech- nology makes this research topic timely and partic- ularly relevant to reducing the cost of wind energy. ‘The main aim of the present paper is to describe and analyze the fundamental flow phenomena that characterize the boundary layer on rotating blades, and to develop a preliminary stall-delay model that modifies the 2-D airfoil data so as to simulate the ED stall-delay effects. ‘The following steps were taken in the development of the model: 1) anal- ysis of the 3-D integral boundary- layer equations for a reference system rotating with the blade, 2) description of the effects of rotor rotation on the separation point and its causes, and 3) determina tion of a simple correction formula to obtain rotat- ing rotor lift coefficient C,(a) and drag coefficient Ce(a) data from measured 2-D airfoil data. ‘The preliminary 3-D stall-delay model consists of two key parameters (the ratio of local chord to local ra- dius c/r the ratio of rotation speed to freestream velocity A) and three empirical correction factors Copyright © 2908 by the American Insitute of Aeronau- tics and Astronautics, Inc. and the American Socity of Me- chanical Engineering. All rights reserved. Visiting Scholar. “Assistant Professor, Senior Member AIA. (a,b,d). The stall-delay model is consistent with the blade element/momentum theory method and the Viterna/Tangler model, and the 3D stall-delay model can be incorporated into the state of the art performance prediction codes, such as PROP. ‘Through comparison with the field test data, the new model for 3-D stall-delay shows good agree- ment between predictions and experiments. The new model should be of great use in existing codes for horizontal axis wind turbine design and analysis 1 Nomenclature = correction factor in stall-delay model 6 correction factor in stall-delay model c blade chord Ca = drag coefficient skin friction coefficient lift coefficient correction factor in stall-delay model boundary layer shape factor velocity gradient air pressure cylindrical coordinates blade radius Reynolds number separation point on airfoil surface ‘velocity components boundary layer edge velocity freestream velocity wind speed angle of attack stream angle air density boundary-layer momentum thickness &182 = boundary-layer displacement thickness Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. 6 = boundary-layer thickness d non-dimensional rotation speed Qr/Vac A = modified non-dimensional rotation speed OR//VE + ORY = viscosity’ ” non-dimensional height of boundary layer ‘ig = boundary-layer shear stress r Poblhausen parameter 9 = blade rotation speed 2 Introduction The accurate prediction of the peak rotor power for stall-regulated wind turbines is of great. prac- tical interest since it is used to size the gem erator and other mechanical components, which largely dictate the cost of the system and there- fore the cost of wind energy. Unfortunately, current, blade-element/momentum theory (BEMT) meth- ods, which are widely used in design codes, often under-predict the peak rotor power. For example, the NREL Combined Experiment (Phase II) turbine exceeded predictions by approximately 15-20%. ‘Numerous other examples exist, many of which have not been reported in the open literature. Despite @ general awareness of this shortcoming of BEMT methods, reports of new prototype tur bines exceeding peak power predictions are not un- common. Consequently, new systems are sometimes cover-designed to be conservative. Of course, higher system costs are associated with this approach. An- other strategy is to run the turbine blades at a lower pitch setting (more toward stall). Although peak: power can be lowered in this way, the blades no longer run at the optimum pitch setting for maxi- mum annual energy production and the overall loads are increased. Moreover, operation at a pitch setting more toward stall amplifies the roughness losses,° which further decrease the annual energy produc- tion. Other common strategies to reduce the peak power involve using various boundary-layer devices, such as stall strips and surface roughness. Again these lead to poorer off-design performance. Clearly, the importance of predicting peak power is not likely to diminish, especially since the majority of wind turbines under development and in use worldwide today are stall regulated. Accurate prediction of peak power is therefore vital, The difficulties in predicting peak power stem from inadequate aerodynamic models of the 3-D 10 post stall effects. In particular, the primary uncer- tainty is related to the blade rotation effects, which are largest inboard. The effect of rotation is to pro- duce a so-called “stall delay,” the principle effect of which gives rise to a dramatic increase the section lift coefficient C; as compared with 2-D data. For in- stance, the local lift coefficient for several blade sta- tions on the NREL Combined Experiment turbine* is shown in Fig. 1 along with the corresponding 2-D data. Similar results on wind turbine blades have been documented elsewhere as well.5 A secondary, but still important, effect is an associated drag re duction, The increased lift and the drag reduction are the principle causes for the increased power out- put as compared with predictions based on 2-D air- foil data. ‘The 3D post-stall effect is not unique to wind tur- bine rotors, Himmelskamp” in his experiment on a propeller blade was the first to measure an increase in lif as compared with 2D data. Helicopter r0- tors in hover and in forward flight also experience 4a similar effect? For propellers and helicopter ro- tors, however, the 8-D post stall effects (more thrust ‘and less power) are beneficial and consequently have not be studied in any great detail. (In contrasting propellers and helicopter rotors with wind turbine rotors, it is important to keep in mind that for wind turbines the lift is directed to produce torque or power; whereas, for propellers it is directed to pro- duce thrust. Thus, @ 3-D lift enhancement produces ‘more power for a wind turbine and more thrust for a propeller.) The main reason for mentioning the sim- ilar 3-D post stall effects experienced on propellers ‘and helicopter rotors is that although the effects are well-known they have not been widely studied, and relatively few modern references to these effects can be found in the respective literature. It is clear that stall-delay on wind turbine blades comes from 3-D effects in which the rotation pays a key role in the phenomenon. Almost all of the literature related to this field suggests that the ef fects due to rotation are a significant factor in prop- erly predicting wind turbine performance and in de- signing more effective airfoils and blades for wind turbines." Over the last few decades, the effects of rotation on the boundary layer have been studied theoretical and experimental by several researchers, Among the earliest work published was that of Sears.! Over a period of approximately 10 yeats, several papers were published on laminar boundary layers in a rotating environment. Because of the Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. assumptions made by those researchers, the chord- wise and radial momentum equations were decou- pled. This left them with essentially a chordwise fixed-blade boundary layer problem and an inde- pendent radial equation for the spanwise boundary layer. Therefore their results were no different than the results obtained by researchers looking at the laminar boundary layer separation problem of the more conventional fixed wing (Le., problems with- out rotation effects) Later, other investigators attacked the same prob- Jem without making similar assumptions as to which ‘terms should be retained and which terms could be ignored. One of the most interesting was @ short technical note published by Banks and Gadd.”? In this reference the authors derived a set of equations that were coupled through the Coriolis and centrifu- gal force terms. They went on to assume an ex- temal velocity with a linear adverse velocity gradi- cent. The solution to this problem showed that the laminar separation point was postponed because of ‘the rotation terms that coupled the equations. In fact they found that for the extreme inboard sta- tions, the boundary layer was completely stabilized against separation by a linear adverse velocity gra- dient. In recent work, Savino and Nyland?® performed flow visualization experiments on the NASA/DOE Mod-0 machine in order to determine flow patterns in the rotor wake and on the suction surface of the rotor blade, They found evidence of strong radial flow downstream of the separation line on the suc- tion surface of the blade. In addition, the location of the separation line was found to be 10 to 20% chord downstream of that for a non-rotating blade under corresponding fiow conditions. With the advent of high-speed computers and the emergence of more sophisticated analyses, includ- ing turbulent boundary layer models, CFD meth- ‘ods have been used to investigate this phenomenon. Narramore and Vermeland* used a full N-S equa- tions solver in conjunction with an algebraic tur- bulence model to simulate the aerodynamics on a helicopter rotor blade in hover at a high angle of attack. Their results showed that stall was delayed due to the effect of rotation, and the lift coefficient reached a much higher value than that in 2-D flow. Also they pointed out that the effect was particular pronounced for the inboard sections. In this current paper, the 3-D incompressible steady boundary layer equations in a cylindrical co- 1" ordinate system are employed to study the 3-D stall delay effects. The equations are simplified by an or- der of magnitude analysis after the work of Snel.4 This process gives rise to a set of equations that is uch simpler than the full 3-D equations. The non- linear convective terms can be used in 2D form, while the important effects of centrifugal and Corio- lis forces are retained. The formulation is consistent with the strip approach used by BEMT method. Using the simplified 3-D integral boundary-layer equations and assumed velocity profiles, a relation between the laminar separation point on the surface of the blade and the key parameters composed of the blade geometry and fiow condition has been set up in this paper. This simple relation points out thet the separation point is postponed with an increase in ¢/r and the rotation speed. These results confirm the earlier work of Banks and Gadd.!? Based the solution of the integral boundary-layer equations, preliminary 3-D stall-delay model is presented in this paper, which is composed of two key parameters (the ratio of local chord to local ra- dius ¢/r and the ratio of rotation speed to freestream velocity ) and three correction factors (empirical constants a,8,d). The 3D stall-delay model cor- rects the 2-D airfoil data C; and Cz, and can be integrated into codes based on the BEMT method, such as PROP. Results from the code are compared with FFA airfoil data, NREL Combined Experiment Rotor (CER) data and data on the Aerostar rotor. 3 3-D Boundary-Layer Equa- tions and Theoretical De- velopment 3.1 3-D Boundary-Layer Equations For the analysis of the flow over rotating blades, a cylindrical coordinate system attached to the blade is used, with the origin in the center of rotation as shown in Fig. 2. The velocity components are u,v and w in directions 0,r and 2, respectively. Undis- turbed freestream streamlines are circular arcs r constant. The $-D incompressible steady boundary- layer equations (continuity, r-momentum, and 6- ‘momentum equations) in cylindrical coordinates are given by Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. ude 7007 or udu 700 * Te has been assumed that the surface on which the boundary layer develops is in the plane of rotation 2=0. Thus, surface curvature in chordwise direc- tion has been neglected with respect to the bound- ary layer thickness. About the z-axis, the rotation speed {is with respect to an inertial system, where 1 Js the distance to the center of rotation. ‘The shear stresses 7, and 72 (either laminar or turbu- Tent) are in the @- and r-direction, respectively. ‘The momentum equations include a number of inertia force-type terms. The #-momentum equation con- tains the Corilisforce term 2v and coordinate curvature term uv/r. ‘The r-momentum equation only contains the centrifugal-force term. TThis term is composed of what formally are a coordinate curva- ture term u/r, a Coriolis force component (—2uStr) and the formal centrifugal force (9r) resulting from the rotating of coordinate system 3.2 Transformation to —_ Integral Boundary-Layer Equations ‘The integral boundary-layer equations for a throe- dimensional boundary layer in an incompressible flow have been developed by many researchers.!> ‘The technique presented in this paper is an exten- sion developed by Snel" for a rotating system. The technique includes an order of magnitude analysis of the 3-D boundary layer equations, which gives rise to a set of equations that is much simpler than the full 3-D equations. In the formulation, the non- linear convective terms can be used in their 2-D form, while the important 3-D effects of centrifu- gal and Coriolis forces are retained. ‘The solution of this set of equations is much simpler than that of the full 3-D integral boundary-layer equations"? and Navier-Stokes equations. Moreover, such a formu- lation is consistent with the strip theory approach used by the BEMT method. ‘The simplified integral boundary-layer equations in 6- and r-directions in the rotating coordinate sys- tem are expressed as 12 oe. fi Ov & be Fett Eye +22 (2a) 502 | 262 Bue _ Cy Gs te os he 1 2076, | Wb) _ 6 Bue 4746 -8- SS FTE) St (28) where the integral boundary-layer parameters are defined as!® (8a) (3b) (8c) (3d) (Se) (se) 3.3 Solution of Integral Boundary- Layer Equations Considering the complexity of the problem, a lam- inar boundary layer distribution on the surface of the blade is assumed as the first step in the cur rent research. The widely used Poblhausen velocity profiles and associated closure relations!® are intro- duced to solve the integral boundary-layer equations for 6-direction velocity profiles given by u ceeaseenae, ia G7 On~2n +8) + Go Bn? + Bn? ~ 4) (4a) ‘The assumed cross-flow velocity profiles!® are given as = t98e(1—n)? (a) where 1) is non-dimensional boundary layer height (n = 2/6), Tis Poblhausen shape parameter, and By is the limiting stream angle. Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. From these assumed velocity profiles, the inte- ral boundary-layer parameters Eqs. (Se-e) can be expressed as (6a) (sb) (60) 6a) be 304 756 r 5 = 1980 (Zags + i0a06 * Tram0) 9) ‘The boundary conditions are given by 2=0u (6a) 2 = 00, Uy = teo(1 — ka) (6) where k is a constant representing the velocity gra- dient and s is arc length measured from the leading edge of the blade. Thus, an external flow with a lin- ear adverse velocity distribution is assumed. This approach of solving the laminar case for a linear ad- verse velocity distribution is similar to that taken by Banks and Gadd.}? Tntegration of the integral boundary-layer equa- tions up to the point of laminar separation (Cy = 0) yields an equation of the form ks = f(A,s/r) (7) where A = Qr/use and ks is the separation factor corresponding to the point of laminar separation. 4 Separation Factor Model Equation (7) represents the relation between the separation factor ks which determines the separa- tion point on the rotating surface and the two key parameters, the ratio of local arc length to local ra- dius s/r, and the ratio of the local rotation speed to freestream velocity 2 The separation factor ks was calculated for dif ferent A and s/r values, and the results are shown in Fig. 8. Several important trends can be gleaned from the figure. First, for @ given value of A, the separation factor ks increases with increasing s/r If is kept constant, k must increase with a reduc- tion in r. This means that the separation point is B postponed due to rotation, and at the same flow condition (the same k) the distance s to separation is increased as the radius r decreases. Second, the separation factor ks is decreased with decreasing 2. ‘Thus, it can be concluded if the wind velocity is kept as a constant while the rotation speed is increased then 2 is increased and the separation point s moves downstream. Third, another important result from Fig. 3 is that all of the curves tend to the limit- ing value 0.1267 for which s/r ~ 0.08. Thus, when s/r < 0.08 rotation has little effect on the separa- tion point, and the solution approaches that of the fixed blade 2-D boundary layer problem. ‘As a first step in the development of a post-stall delay model based on the trends indicated in Fig. 3, the ks-curves are approximated by an equation of ‘the form (sre B= (any ® where a, b, and d are correction factors. From the numerical solution yielding the results shown in Pig. 5, the correction factors have been found to vary from approximately 0.8 to 1.2 for a,b and 0.4 to 1.0 for d. An example is shown in Fig, 4 for which Bq. (8) is used and compared with the nu- ‘merical solution represented by Eq. (7). In this case the parameters a,b, and d are all set to unity. For a 2-D flow without rotation, the separation factor is ke = 0.1267; whereas, for a 3-D blade in rotation, the separation factor becomes ks > 0.1267 and asymptotes to ks = 0.1267. Thus, for a $D flow with rotation, the increase in ks as compared with the 2-D case without rotation is given by L6(s/r) a ~ (s/x)4/% 01267 5 (¢/r)7 This formula (after a slight modification which fol- lows) forms the basis ofthe key function that is used to empirically correct 2-D airfoil data for rotation cflects that are present in the -D rotating flow In light of the empirical and approximate nature of the correction, the distance parameter s/r is re- placed by the blade geometry parameter ¢/r, viz (efx)6” O4267 B+ (cfr) ks = 1.6(s/r)> Alks 1 (80) 1 (9) 5 3-D Correction to 2-D Air- foil Data Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. It is clear from the theory that boundary layer separation is postponed because of rotation, which implies that the stall angle of attack with rotation effects is larger than that in the 2-D flow. The corre- sponding lift coefficient Cy is increased and the drag, coefficient C, is reduced. Therefore, the stall delay model should account for these changes in the 2-D Cj and Cy airfoil data. Moreover, to be readily im- plemented in BEMT methods, such as that used in the PROP code, a simple correction to 2-D airfoil data for 3-D rotation effects would be most advan- ‘tageous. Ibis assumed that the 3-D airfoil data are approx: imately equal to that obtained in 2-D wind tunnel tests plus an increment in C; and e decrement in Ca s0 as to simulate rotation effects on each radial station of the wind turbine blade. If the AC; and Cg denote the increments and the decrements, the 3D airfoil data can be expressed as Cisp = Crap + AC, (10) Caap = Caan ~ ACa (108) Based on the stall-delay principle described pre- viously, as well as basic airfoil theory, a correction formula is suggested as Ci = filCig~ Ci.20) (11a) ACe= felCean - C50) (ay) where Ciy = 2n(a:~ a9), Cao = Caan for a= 0 ‘The functions fy and fy are modeled after that for the separation factor Bq. (9b), viz data and are 1 [1.6(¢/r) a (c/r) #2 . al C287 b+ (e/r)FF - 1 [1.6(¢/r) a ~ (c/n) 8 fs alana oe eae - i A=OR//V24 OR? (126) where a, b, and d are empirical correction factors. In this paper, a, 6, and d are set to unity. Also, the parameter A has been modified and replaced by a ‘modified tip speed ratio A, and furthermore for the rag factor fa a factor of 2 was introduced multiply- ing A. These changes were introduced to provide a better fit to the existing experimental data as dis- cussed in the next section 4 To summarize, a new 3-D stall-delay model has been developed, which is specifically developed for integration into the PROP code. This model con- sists of two key parameters — one is ¢/r which char- acterizes the rotor geometry, the other is A which accounts for the effects of rotation. Finally, the new ‘model is given by the following relations. Crap = Crap + fC» ~ Cao) Casp = Caz ~ fa(Ca20 ~ Cap) (13a) (235) Prediction and Comparison with Experimental or Nu- merical Calculation Data In the section, the post-stall model is applied to several existing data sets, namely the FFA 5WPX, wind turbine blade, the NREL Combined Experi- ‘ment Rotor (CER), and AeroStar 7.5-m blade. 6.1 FFA 5WPX Wind Turbine Blade ‘The Aeronautical Research Institute of Sweden (PRA) conducted aerodynamic measurements on a (VSH) 5WPX blede that formed part of a two- bladed 5.35 m diameter rotor. ‘The measurement program was carried out in the 12 x 16 m low speed wind tunnel of Chinese Aerodynamic Re search and Development Center." One of the blades ‘was equipped with 29 pressure taps at each of eight radial stations. The measured pressure distribution ‘were used to determine the corresponding lift coef- ficients. The rotor blade planform and location of pressure measurement sections are given in Ref. ‘The effect of blade rotation has been predicted for two blade stations, namely r/R = 30% and 55% which correspond to c/r=0.37 and 0.16, respectively. It should be noted that in this simulation, the 2- D airfoil data were not changed, and kept equal to the section at the 55% radial position. For the predictions, the conditions were Re = 0.5 x 10°, {1 = 158 rpm, and wind speed Vip = 8.8 m/s, Figure 5 compares the predicted Ci-a curves using the new stall-delay model with the corresponding data obtained from field test measurements made on the rotating blade. Also shown in the figure is ‘the corresponding 2-D airfoil data. In the figure “3- D prediction” refers to the predictions based on the Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. new model, and “S-D test” refers to the field test data. The agreement is very good between the pre- dictions and field test data. The effect of rotation is seen to delay stall and increase the lift coefficient to angles of attack past that of 2-D stall. Itis also seen that the effect of rotation drops rather quickly as the cr ratio is decreased. For the radial station with r/R = 0.30 and o/r = 0.37, the 3-D predicted Cr al- ‘ways increases for increasing of the angle of attack. The C; reaches a value of 1.6 at 30 deg as compared ‘with a 2-D value of 0.5 at the same condition. For the station r/R = 0.56 with c/r = 0.16, the 3D prediction and test date are as expected similar to the 2-D airfoil data since rotation has litte effect on the lift coefficient. This conclusion is consistent with the theoretical analyses and past research discussed previously.® 8 For the drag coefficient Cz, Fig. 6 compares the current 3D predictions with numerical predictions of Sorensen using an viscous-inviscid interaction code." Experimental data from field tests was not measured. Also shown in the figute is predicted 2-D data based on Navier-Stokes calculations." Tt can be seen the 3-D predictions are always lower than 2D, which suggests that the rotational effects in ‘general lower the drag coefficients. The current pre- dictions compare very well with the 3-D numerical simulation. 6.2 CER Combined Experiment Ro- tor ‘The National Renewable Energy Laboratory has ‘completed the initial tests of the “Unsteady Aerody- namics Experiment” aimed at providing a detailed understanding of horizontal-axis wind-turbine aero- dynamics. The experiment used a special three- blade, stall-regulated (constant speed), downwind, 10-m diameter wind turbine rotor. The rotor had a 0.475-m constant chord blades that made use of the NREL $809 airfoil along the entire span. The rotor speed was 72 rpm. A detailed description the the ex- perimental setup is given in Refs.1°. The resulting pressure data provided high-resolution aerodynamic lift and drag force measurements on the rotating blade. The stall delay model was used to predict the blade airfoil characteristics at three radial stations: 1/R = 30%, 47%, and 80%, corresponding to ¢/r = 0.301, 0.181, and 0.118, respectively. ‘The following conditions were used: Re = 0.5 x 108, © = 72 rpm, 15 Vig = 10 m/s. The 2-D airfoil data from wind tunnel tests is that given in Ref. 19. The comparisons be- tween the predicted results and test data are shown in Fig. 7. As with the FFA blade, a large depen- dence on radial positon is seen for the lift coefficient curves, with increasing maximum lift coefficients oc- curring at increasingly large angles of attack in mo ing from tip to root. It should be noted that at station r/R = 0.80 the Cj increases because at this station ¢/r = 0.113 — a value which is o slightly larger than the limiting value of ¢/r = 0.08. Overall, the results shown in Fig. 7 indicate that the stall-delay model yields fairly good agreement ‘with the test data. In particular, in the post stall re gion for 20 deg < a < 30 deg, the agreement is very ‘good, but over the range 10 deg < a < 20 deg there is a larger discrepancy between the prediction and ‘est results. Thus, further refinement of the model is ‘warranted to properly simulate post-stall rotor flow fields. The effects of laminar-to-turbulent transition should likely be included in the model, which cur- rently has no dependence on Reynolds number and was developed based on only laminar boundary re- lations. 6.3 AeroStar HAWT In Ref, 20, Sorensen modeled the AeroStar 7.5- m blade using a three-level, viscous-inviscid inter- active model to predict the 3-D stall-delay effects In Fig. 8, the predicted C; distribution along the blade based on the current method is compared with ‘the mumerical predictions of Sorensen as well as cal- culated 2-D data. For this case, the rotor speed is 50.3 rpm with a wind speed of 10 m/s. First, it is seen that C predictions from the 3D stall- delay model are in close agreement the numerical predictions of Sorensen. Second, for the outer 40% of the blade span, the effect of rotation is not evi- dent; whereas, for the inner part of span, the 3-D values are much larger than those of the 2-D case. In Fig. 9, the drag coefficient distributions along the blade span are compared. Only a slight difference in the 3-D Cz distributions (current method and that from Sorensen) as compared with the 2-D case is seen because the correction factor fz is smaller than fi Figure 10 shows the predicted power curves for ‘the AeroStar blade as compared with the experi- mental data. It is seen that for wind speeds less than 10 m/s, both the blade-element theory and the Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. ‘&D stall-delay model are in good agreement with the measurements since for the most part the blade is not stalled. For higher wind speeds, however, the prediction based on 2-D data under-predicts the Power output as would be expected. In general, the results based on the 3-D stall-delay model are in good agreement with test data at higher wind speeds. Where the difference is largest at a wind speed of 12 m/s, this may be due to the over- prediction in lift over the range 10 deg < a < 20 deg as seen in Fig. 7. Nevertheless, the close agreement at peak power is encouraging. 7 Conclusions Based on the concept of the stall delay and basic airfoil theory, a preliminary 3-D stali-delay model has been presented in this paper, The model con- sists of two key parameters (the ratio of local chord to local radius c/r, the ratio of rotation speed to freestream velocity AA), and three empizical correc- tion factors (a, and d). This 3-D stall-delay model can be used to modify the airfoil data based on 2- D wind tunnel test, and it is consistent with the blade element/momentum theory (BEMT) method and Vitera/Tangler model, and it can be incor- porated into the state of the art simulation codes, such as PROP, which are currently applied to hor- ‘zontal axis wind turbine performance predictions. ‘Through comparison with the field test data, the new model for 3-D stall-delay shows the good agree ‘ment between predictions and experiments. The new model should be of great use in existing codes, for wind turbine design and analysis. 8 Acknowledgments The support of the National Renewable Bnergy Laboratory under Subcontract No. XCX-7-16466-01 is gratefully acknowledged. Also, the several discus- sions with J.L. Tangler of NREL proved to be quite helpful during the course of this work. 9 References {1] Hansen, A.C., and Butterfield, C-P., “Aerody- namics of HAWT,” Annu. Reu. Fluid Mech., Vol.25, 16 1993, pp. 115-149, (2] Tangler, J-L., and Selig. M.S., “An Evaluation cof an Empirical Model for Stall Delay Due to Rota- tion for HAWTS,” Presented at the Windpower'97 Conference, Austin, Texas, June 15-18, 1997, {3] Tangler, J.L., “Influence of Pitch, Twist, and ‘Taper on a Blade’s Performance Loss Due to Rough- ness,” AWEA Windpower'96 Conference, Denver, CO, June 1996. is] Miller, M.S, Shipley, DB, Young, TS., Robinson, M.C., Luttges, M.W., and Simms, D.A., “The Baseline Data Sets for Phase II of the Com- bined Experiment,” NREL /TP-442-6915, July 1996. (5] Bjorck, A., Ronsten, G., and Montgomerie, B., “Aerodynamic Section Characteristics of a Rotating and Non-Rotating 3.375 m Wind Turbine Blade,” FFA TN 1995-03, March 1998. [6] Madsen, H.A., “Aerodynamics of a Horizontal- ‘Axis Wind Turbine in Natural Conditions,” Riso-M- 2903, Riso National Lab, Sept. 1991 [7] Himmelskamp, H., “Profile Investigations on a Rotating Airscrew,” MAP Volkenrode, Reports and ‘Translation No. 882, Sept. 1947. [8] Narramore, J.C., and Vermeland, R., “Navier- Stokes Calculations of Inboard Stall Delay Due to Rotation,” J. Aireraft, Vol. 29, No. 1, Jan-Feb. 10992, pp. 73-78. (9) Corrigan, J.J., and Schillings, J.J., “Empirical Model for Stall Delay Due to Rotation,” American Helicopter Society Aeromechanics Specialists Cont., San Francisco, CA, Jan. 1994. (20) Eggers, AJ., and Digumarthi, R.V., “Ap- proximate Scaling of Rotational Effects of Mean Aerodynamic Moments and Power Generated by the Combined Experiment Rotor Blades Operating in Deep-Stall Flow,” 11th ASME Wind Energy Sym- posiurn, SED-Vol. 12, 1992, pp. 33-43. [11] Sears, W.R., “Boundary Layers in Three- Dimensional Flow", Applied Mechanics Reviews, Vol. 7, No. 7, 1954, pp. 281-286 (02] Banks, W.HLH., and Gadd, G.E., “Delaying Effect of Rotation on Laminar Separation,” AAA J., Val. 1, No. 4, April 1963, pp. 941-942 [03] Savino, J.M. and Nyland, T.N., “Wind Ture bine Flow Visualization Studies,” AWEA Wind En- ergy Conference, San Francisco, CA, Aug. 1985, [i] Snel, H., Houwink, R., and Bosschers, J., “Sectional Prediction of Lift Coefficients on Rotat- ing Wind Turbine Blades in Stall,” ECN Report, EON-C-93.052, Dec. 1904. {15} Lakshiminarayana, B., and Covinden TR, Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. “Analysis of Turbulent Boundary layer on Cascade and Rotor Blades of Turbomachinery,” AIAA J, Val. 19, No. 10, 1981, pp. 1333-1341 16) Moore, F-K., Theory of Laminar Flows, Princeton University Press, Princeton, New Jersey, 1964, pp. 259-265, 17] Ronsten, G., “Static Pressure Measurement on @ Rotating and a Non-Rotating 2.375 m Wind ‘Turbine Blade, Comparison with 2D Calculation,” European Energy Conference, Amsterdam, 1991 [18] Jameel, H.M., “Study of Rotating Airfoil,” J. of Aircraft, Vol. 8, No. 4, 1992, pp. 314-316. (19] Simms, D.A., and Robinson, M.C., etc., “A Comparison of Baseline Aerodynamics Performance of Optimally. Twisted Versus Non-Twisted HAWT Blades,” 15th ASME Wind Energy Symposium, Jan. 28 - Feb. 2, 1996. [20] Sorensen, J.N., “Prediction of Three Dimensional Stall on Wind Turbine Blade Using ‘Three-Level, Viscous-Inviscid Interaction Model,” 1986, pp. 33-41, 4 Fig. 2 Coordinate system attached to rotating blade (Ref.14) as zo Bas Q 2 é. = 5 : fos i z . he Angle ot Attack (egres) Fig 1 Blade norma ce cosine test ata (Ret. 3) a 0 00s 0x 015 02 Parameter st 02503 Fig. 3 Relationship between parameter ks and s/t Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. 0x8 os rate ous 20a, Rat — ZAREST 2-2 peioson Soo. 16 OW7F a 208, ° 4 ec. Ret.11, cira0.16 © A Tog Est f raat 1 Ets, a RAT eas spovf ire gos} sSemherth eas pny tse Ll gos ; 2s Bors g E F 02 one bos 0.13 ot as ore ° oon 01 01s 02 02s 03 Cee spe 0 neg seen omens Parameter oe Angle of Attack (depres) Fig. 4 Simulation results of Eq (8) 18 Fig. 6 FFA airfoil drag coefficient 2D Cal, Rol4 — 3D presiction, /=0.16 -—= S Dies Beli ota018 © oh=0.37 obiataat ones? 16 2.0 wind tunnel data — 2D predeton, dre0.1t — 0 toet data, i011 > go q Sos S § 2 Sos § os oe ° oe ess 80 Angle ot Atack degrees) Angie ot Atak (epee Fig SFA sri it cote 18 Fig. 7$809 airfoil it coetficient Copyright © 1997, American Institute of Aeronautics and Astronautics, Inc. 20 Cal, Ret'20 — 30 Gal, Rel 20 = 3. preston Lift Coefficient aeona Fig. 8 Lift coefficient distribution along wing span yom B 005 -D Cal., Ref.20 — Bose sesso poe & 001 ° ego ae co recto m Fig. 9 Drag coefficient distribution along wing span Power Output (kW) 19 2.0 Cal, Rel20 — SDiest Ret20 © . ‘3 preacton * ol 2. 6 8 0 2 4 6 18 ‘Wind Speed (m/s) Fig. 10 Aerostar rotor performance

You might also like