Energy 36 (2011) 4919e4934

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

The Darrieus wind turbine: Proposal for a new performance prediction model based on CFD
Marco Raciti Castelli, Alessandro Englaro, Ernesto Benini*
Department of Mechanical Engineering, University of Padova, Via Venezia, 1, 35131 Padova, Italy

a r t i c l e i n f o
Article history: Received 18 December 2010 Received in revised form 18 May 2011 Accepted 20 May 2011 Available online 29 June 2011 Keywords: Darrieus wind turbine Vertical-Axis-Wind turbine Performance predition

a b s t r a c t
This paper presents a CFD model for the evaluation of energy performance and aerodynamic forces acting on a straight-bladed vertical-axis Darrieus wind turbine. The basic principles which are currently applied to BE-M theory for rotor performance prediction are transferred to the CFD code, allowing the correlation between ow geometric characteristics (such as blade angles of attack) and dynamic quantities (such as rotor torque and blade tangential and normal forces). The model is proposed as a powerful design and optimization tool for the development of new rotor architectures for which test data is not available. After describing and validating the computational model against experimental data, a full campaign of simulation is proposed for a classical NACA 0021 three-bladed rotor. Flow eld characteristics are investigated for several values of tip speed ratio, allowing a comparison among rotor operation at optimum and lower Cp values, so that a better understanding of vertical-axis wind turbines basic physics is obtained. 2011 Elsevier Ltd. All rights reserved.

1. Introduction and background The use of BE-M (Blade Element e Momentum) models for the design and analysis of vertical-axis wind turbines has aroused a large credit, not only in research and academic communities but also in industrial appliances, thanks to general acceptable accuracy of the results, as well as widely available literature and code simplicity. BE-M theory, rst introduced by Glauert [1] in order to predict the structural dynamics and performance of airplane propellers, was adjusted for vertical-axis wind turbine aerodynamics by Templin [2], obtaining the most elementary approach based on momentum theory, named single-streamtube model. Strickland [3] extended Templins approach by considering the single streamtube as composed of a number of adjacent and aerodynamically independent smaller streamtubes, thus allowing the use of airfoil characteristics based on local (rather than on average) Reynolds number. This approach, named multiplestreamtube model, was further developed by several authors [4e9] by placing two actuator discs behind each other, thus combining the multiple streamtube model prediction tool with the

* Corresponding author. Tel.: 39 0 49 8276767; fax: 39 0 49 8276785. E-mail addresses: marco.raciticastelli@unipd.it (M. Raciti Castelli), alessandro. englaro@studenti.unipd.it (A. Englaro), ernesto.benini@unipd.it (E. Benini). 0360-5442/$ e see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.energy.2011.05.036

double actuator disc theory. Through continuous renements [10,11] (including also the most recent optimization techniques, such as genetic algorithms [12]), BE-M based design tools proved to be able of accurately predicting Darrieus wind turbine performance and are to be considered as the basis of most of the performance prediction tools currently used in wind turbine industry. Aerodynamic performance prediction tools based on momentum theory, while presenting the advantage of a low computation effort, are nevertheless limited by the availability of extended airfoil databases (up to 180 ) which should also refer to rather low local Reynolds numbers (close to Re 105), which are typical of a vertical axis wind turbine operation [13,14]. Since most airfoil database available in literature is derived from aeronautical applications, hardly extending far beyond stall and also referring to relatively high Reynolds numbers (above Re 106), it has been recognized that there is a paucity of reliable airfoil data needed to describe the aerodynamics of wind turbine blades [15]. The airfoil section data requirements for application in vertical axis wind turbines are in fact broader in scope than those the aircraft industry usually concerns itself with. This limits the analysis on vertical axis wind turbines to a narrow number of blade proles, substantially some NACA series of symmetrical airfoils, for which extended database at low Reynolds numbers is available in literature [16]. Moreover, as pointed out by Burton et al. [17], dwelling upon considerations on the energy extraction process rather than

4920

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

Nomenclature Ai,j [359 90] matrix containing the local values of blade relative angle of attack as a function of both angular position and time As [mm2] rotor swept area Ati [1 90] vector containing the local values of blade relative angle of attack at time step ti as a function of azimuthal position j A.O.A. [rad] average value of blade relative angle of attack c [mm] blade chord cP(q) [-] differential pressure coefcient CP(q) [-] rotor instantaneous power coefcient Cp,ave [-] average power coefcient during a full rotor revolution Ct(q) [-] rotor instantaneous thrust coefcient CT(q) [-] rotor instantaneous torque coefcient Drotor [mm] rotor diameter Fx(q) [N] instantaneous rotor force in x direction Hdomain [mm] computational domain height Hrotor [mm] rotor height Hwindtunnel [mm] computational domain height, validation model j [-] azimuthal position along blades trajectory n [-] number of azimuthal positions which are not considered for the calculation of the relative value of local angle of attack N [-] number of rotor blades p(x,y) total pressure at point of coordinates (x,y) P(q) [W] rotor instantaneous power Pave [W] average power during a full rotor revolution Rrotor [m] rotor radius Re [-] blade Reynolds numbers Sdomain [mm2] computational domain section t [s] simulation time simulation time step ti [-] T [s] rotor period of revolution

Trotor(q) [Nm] rotor instantaneous torque TSR [-] tip speed ratio U [m/s] absolute wind velocity at blade position Ux [m/s] absolute wind velocity at blade position, component along x axis Uy [m/s] absolute wind velocity at blade position, component along y axis Ut [m/s] blade tangential speed at blade position Vtest section [m/s] mean wind velocity at rotor test section VN [m/s] unperturbed wind velocity at computational domain entrance w [m/s] relative wind velocity at blade position Wdomain [mm] computational domain width Wwind tunnel [mm] computational domain width, validation model a [rad] blade relative angle of attack in the plane of the airfoil cross section apos [rad] positive stall angle for NACA 0021 at Re 300.000 (assumed 0.21 rad) aneg [rad] negative stall angle for NACA 0021 at Re 300.000 (assumed 0.21 rad) ati,j [rad] local values of blade relative angle of attack as a function of azimuthal position j along blades trajectory for time step i 3sb [-] solid blockage correction factor d [ ] angle between absolute wind velocity at blade position and unperturbed wind direction q [ ] blade azimuthal coordinate J [rad] angular sector covered during a period of rotor revolution g [ ] angle between absolute wind velocity at blade position and blade translational speed at blade position r [kg/m3] air density (assumed 1.225 kg/m3) s [-] rotor solidity u [rad/s] rotor angular velocity

on the specic turbine design, all classical aerodynamic tools are unable to visualize the basic structure of the ow eld inside the rotor volume. The limitations in low Reynolds number accurate aerodynamic databases can be overcome by using advanced CFD (Computational Fluid Dynamic ) codes, which can outank the lack of airfoil data thanks to their inherent ability to determine the aerodynamic components of actions through the integration of the NaviereStokes equations in the neighborhood of the wind turbine blade prole. A rst attempt of improving the efciency of a vertical-axis wind turbine through the application of the emerging CFD capabilities was performed by Vassberg et al. [18] through the simulation of the dynamic motion of a turbine blade spinning about a vertical axis and subjected to a far-eld uniform free-stream velocity ow eld. Also Ferreira et al. [19,20] presented a systematic CFD analysis of a two-dimensional straight-bladed rotor conguration. The effect of dynamic stall in a 2D single-bladed VAWT (Vertical-Axis Wind Turbine) was investigated, reporting the inuence of the turbulence model in the simulation of the vortical structures spread from the blade itself. Recently, Kumar et al. [21] proposed a low Reynolds number VAWT design and optimization procedure based on both CFD simulations and BE-M calculations, while Raciti Castelli and Benini [22] presented a model for the evaluation of energy performance and aerodynamic forces acting on a helical single-bladed VAWT

depending on blade inclination angle: the analysis was based on ve machine architectures, which were characterized by an inclination of the blades with respect to the horizontal plane in order to generate a phase shift angle of 0 , 30 , 60 , 90 and 120 between lower and upper blade sections, for a rotor with an aspect ratio of 1.5. DAlessandro et al. [23] developed a new computational approach, based on both CFD and wind tunnel measurements, for the simulation of a drag-driven VAWT energy performance, in order to gain an insight into the complex ow eld structures developing around a Savonius rotor. Raciti Castelli et al. [24] performed a numerical analysis validation campaign for a Darrieus micro-VAWT through a systematic comparison with wind tunnel experimental data. This work proved that it is possible to determine the best near-blade grid element dimension through statistical analysis of some indicators, such as the y parameter, in order to maximize the accuracy of the numerical prediction of rotor performance while maintaining a reasonable computational effort. The main goal of the present work is to propose a new straightbladed VAWT performance prediction model, based on numerical simulations, in order to determine the rotor power curve. The basic principles which are currently applied to the BE-M theory for rotor performance prediction are transferred to the CFD code, allowing the correlation between ow geometric characteristics (such as blade angles of attack) and dynamic quantities (such as rotor torque and power). In particular:

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

4921

 the kinematic and dynamic quantities which are classically determined from the Momentum section of a BE-M code have been calculated using CFD, thus being able to better investigate the ow eld close to the rotor by considering the effects of dynamic stall, streamtube expansion and velocity components normal to unperturbed wind direction;  the kinematic and dynamic quantities which are classically determined from the Blade Element section of a BE-M code have been calculated using CFD, thus being able to overcome the lack of low Reynolds number accurate aerodynamic databases thanks to the inherent ability of the CFD code in determining the aerodynamic components of actions through the integration of the NaviereStokes equations in the neighborhood of the blade prole [25]. After describing the computational model and the relative validation procedure, a full campaign of simulation is proposed for a classical NACA 0021 three-bladed rotor. Flow eld characteristics are investigated for several values of tip speed ratio, allowing a comparison among rotor operation both at optimum and lower values of angular velocity, so that a better understanding of vertical-axis wind turbines basic physics is obtained.

Table 1 Main geometrical features of the tested model. Drotor [mm] Hrotor [mm] N [-] Blade prole c [mm] Spoke-blade connection s [-] 1030 1 (2D simulation) 3 NACA 0021 85.8 0.25 c 0.5

Rotor azimuthal position was identied by the angular coordinate of the pressure centre of blade No. 1 midsection (set at 0.25$c for NACA 0021 airfoil), starting between the 2nd and 3rd Cartesian plane octants, as can be seen in Fig. 2. 3. Spatial domain discretization As the aim of the present work was to reproduce the operation of a rotating machine, the use of moving sub-grids was necessary. In particular, the discretization of the computational domain into macro-areas has led to two distinct sub-grids:  a rectangular outer zone, determining the overall calculation domain, with a circular opening centered on the turbine rotational axis, which was identied as Wind Tunnel sub-grid, xed;  a circular inner zone, which was identied as Rotor sub-grid, rotating with rotor angular velocity u.

2. Model geometry Fig. 1 shows a schematic description of the adopted survey methodology, consisting in the coupling of an analytical code to a solid modeling software (capable of generating the desired rotor geometry depending on the desired design geometric parameters) which is linked to a nite volume CFD code for the calculation of rotor performance. CFD results are post-processed using a second analytical code for calculation of rotor main kinematic and dynamic characteristics. The aim of the present work was to numerically analyze the aerodynamic behavior of a three-bladed Darrieus VAWT operating at different angular velocities for a constant wind speed of 9 m/s. The main geometrical features of the tested rotor are summarized in Table 1. The solidity parameter s is dened as Nc/Rrotor, as suggested by Strickland [3].

3.1. Wind tunnel sub-grid Fig. 3 shows the main dimensions and the boundary conditions of the Wind Tunnel sub-grid area. The computational domain width was set almost 80 rotor diameters so that the value of solid blockage correction factor [26] was less than 0.32% and would cause an increase in the cube of the velocity at rotor test section of less than 1% with respect to the value of unperturbed wind velocity at computational domain entrance, in formulas:

3sb 1=4

As Drotor $Hrotor 1=4 Sdomin Wdomain $Hdomain

(1) (2)

Vtest section 1 3sb VN

Fig. 1. Scheme of the survey methodology.

Fig. 2. Azimuthal coordinate of blade midsections centre of pressure.

4922

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

Fig. 3. Main dimensions of the Wind Tunnel sub-grid area (dimensions in mm). Fig. 4. Schema of Rotor sub-grid area (dimensions in mm).

Since the estimation of the correct value of wake blockage was rather difcult, and being the overall calculation domain over-sized by imposing the requirement on the cube of the velocity at the test section, the wake blockage was not considered in these calculations. In order to allow a full development of the wake, Ferreira et al. [19] placed inlet and outlet boundary conditions respectively 10 diameters upwind and 14 diameters downwind with respect to rotor test section for a wind tunnel CFD simulation. As the aim of the present work is the simulation of a turbine operating in open eld conditions and because of the huge domain width necessary to avoid solid blockage (as discussed above), inlet and outlet boundary conditions were placed respectively 37 rotor diameters upwind and 60 rotor diameters downwind with respect to rotor test section. Two symmetry boundary conditions were used for the two side walls. The circumference around the circular opening, centered on the turbine rotational axis, was set as an interface, thus ensuring the continuity in the ow eld. An unstructured mesh was chosen for the Wind Tunnel sub-grid, in order to reduce engineering time to prepare the CFD simulations. 3.2. Rotor sub-grid The Rotor sub-grid is the uid area simulating the revolution of the wind turbine and is therefore characterized by a moving mesh, rotating at the same angular velocity of the turbine. Its location coincides exactly with the circular opening inside the Wind Tunnel sub-grid area and centered on the turbine rotational axis. Fig. 4 shows the main dimensions and the boundary conditions of the Rotor sub-grid area. It is good engineering practice to provide that the mesh on both sides of the interface (Rotor sub-grid and Wind Tunnel sub-grid areas) has approximately the same characteristic cell size in order to obtain faster convergence [27]. An isotropic unstructured mesh was chosen for the Rotor subgrid, in order to guarantee the same accuracy in the prediction of rotors performance during the revolution e according to the studies of Commings et al. [28] e and also in order to test the prediction capability of a very simple grid. Considering their features of exibility and adaption capability, unstructured meshes are in fact very easy to obtain, for complex geometries, too, and

often represent the rst attempt in order to get a quick response from the CFD in engineering work. All blade proles inside the Rotor sub-grid area were enclosed in a control circle of 400 mm diameter. Unlike the interface, it had no physical signicance: its aim was to allow a precise dimensional control of the grid elements in the area close to rotor blades, by adopting a rst size function operating from the blade prole to the control circle itself and a second size function operating from the control circle to the whole Rotor sub-grid area, ending with grid elements of the same size of the corresponding Wind tunnel subgrid elements. An interior boundary condition was used for control circle borders, thus ensuring the continuity of the cells on both sides of the mesh. A growth factor of 1.2 was set from the surface of the control circle to Rotor sub-grid, thus expanding grid size from 4 mm to 10 mm, as shown in Fig. 5.

Fig. 5. Rotor sub-grid mesh.

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

4923

Fig. 7. Grid points clustering close to trailing edge. Fig. 6. Size functions applied to control circle elements.

4. The Control circle Being the area close to the blade proles, great attention was placed to the control circle. The computational grids were constructed from lower topologies to higher ones, adopting appropriate size functions, in order to cluster grid points near the leading edge and the trailing edge of the blade prole, so as to improve the CFD code capability of determining lift, drag and the separation of the ow from the blade itself. Two size functions were set inside the control circle, as shown in Fig. 6:  size function No. 1 started from the leading edge and inuenced inner and outer blade prole;  size function No. 2 started from the blade prole and inuenced the whole control circle area, as described in the previous section. Table 2 summarizes the main features of the mesh close to rotor blade. The trailing edge thickness was set to about 0.38 mm. The trailing edge itself was meshed using 10 grid points, as shown in Fig. 7, while Fig. 8 displays a view of the control circle grid. 5. Code validation Before analyzing the models described in the previous sections, a complete validation work based on wind tunnel measurements has been conducted. As shown in Fig. 9, the experimental setup consisted in a classical vertical-bladed Darrieus rotor made of aluminum and carbon bers, using a NACA 0021 blade prole with a chord length of 85.8 mm, which was tested in Bovisas low turbulence facility (Milan).

As can be seen from Fig. 10, a computational domain of rectangular shape has been chosen, having the same wind tunnel external sizes: the wall boundary conditions of the model consisted in two lateral walls spaced 2000 mm apart from the wind tunnel centerline (the wind tunnel measured 4000 mm in width and 3880 mm in height). The rotor axis was placed on the symmetry position of the wind tunnel section. 2D and 3D simulations were performed, in order to take into account the drag effect induced by the spokes, too. Only half of the experimental setup was modeled, due to its vertical symmetry: in this case a symmetry boundary condition was used at the computational domain midsection. Anyway, the geometrical features of the model did not allow other simplications to be performed. The effect of gravity on the rotor working curves has not been contemplated, being considered not inuential for the scope of the validation work. The main features of the validation model are summarized in Table 3. The correction due to wind tunnel blockage was not applied (both for experimental measurements and numerical computations), in order to minimize any sources of error due to a wrong

Table 2 Main features of the mesh close to rotor blade. Number of grid points on airfoil upper/lower surface Minimum grid points spacing (on airfoil leading edge) Maximum grid points spacing (on airfoil trailing edge) Growth factor from airfoil leading edge to airfoil trailing edge Growth factor from airfoil surface to Rotor sub-grid area 3600 0.015 mm 0.025 mm 1.005 1.1 Fig. 8. Control circle grid for NACA 0021 blade section.

4924

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934 Table 3 Validation model main features. Blade prole c [mm] Drotor [mm] Hrotor [mm] As [m2] s [-] N [-] Spoke-blade connection Hwind tunnel [mm] Wwind tunnel [mm] NACA 0021 85.8 1030 1456.4 1.236 0.5 3 0.5 c 4000 3800

After some corrections to take into account spoke drag, the average torque values, measured in the wind tunnel for a 9 m/s unperturbed wind speed at test section entrance (and different tip speed ratios), were compared with those obtained from CFD analysis. For more details about the validation procedure and tests, see Ref. [24]. Comparison of the computed results with experimental data showed that prediction obtained using Enhanced Wall Treatment k-3 Realizable turbulence model successfully reproduced most of the ow features associated with the revolution of the tested rotor. In particular, as can be seen in Fig. 11, the numerical code proved able to replicate the shape of the experimental curve and was able to accurately capture the maximum power coefcient tip speed ratio, thus offering a reliable alternative to the development of experimental tests, at least at a rst attempt. The discrepancies between the two curves, roughly constant over rotor operational angular range and corresponding to about one half of two-dimensional power coefcient for optimum tip speed ratio, are due to the combined effects of nite blade length and spokes drag, as described by Raciti Castelli et al. [24]. 6. Temporal discretization and convergence criteria The numerical analysis proposed in the present work kept some common points with the validation model, particularly as far as the blade prole and the rotor radius are concerned. Only the computational domain was enlarged in order to analyze the open eld operation of the turbine. The adopted commercial CFD package was Fluent 6.3.26, that implements 2-D Reynolds-averaged NaviereStokes equations using a nite volume-nite element based solver. The uid was assumed to be incompressible, being the maximum uid velocity in the order of 60 m/s. As pointed out by McMullen et al. [29], the calculation of unsteady ows in turbomachinery continues to present a severe challenge to CFD. During VAWT operation, the unsteadiness stems mainly from the relative motion of the rotating blade elds and has a fundamental period which depends both on the rate of rotation and the number of blade passage. For the proposed calculations, the temporal discretization was achieved by imposing a physical time step equal to the lapse of time the rotor takes to make a 1 rotation. An improved temporal discretization simulation did not show any signicant variation [24]. As a global convergence criterion, each simulation was run until instantaneous torque coefcient values showed a deviation of less than 1% compared with the equivalent values of the previous period, corresponding to a revolution of 120 due to rotor threebladed geometry. Residuals convergence criterion for each physical time step was set to 105. The present simulations required about 4 CPU seconds per physical time step. An average of about 25 sub iterations was necessary to make the solution converge at each physical time step. The calculations, performed on an 8 processor, 2.33 GHz clock

Fig. 9. Experimental setup for validation model.

estimation of the blockage of the wind tunnel itself. Furthermore, this choice had the signicant advantage of reducing the computational domain, allowing a saving in the total number of mesh elements. The correction of the friction resistive torque due to the bearings was taken into account. The 2D grid independence of the solutions was veried through simulations performed with grid sizes that varied from about 600,000 to more than 1,000,000 mesh elements. In order to test the code sensitivity to the number of grid points, three unstructured meshes were adopted for the Rotor sub-grid, while the Wind Tunnel sub-grid remained substantially the same.

Fig. 10. Validation model, 3D computational domain.

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

4925

Fig. 11. Comparison between wind tunnel measured (red) and numerically determined 2D power curve (blue) for a wind speed at test section entrance of 9 m/s (correction due to wind tunnel blockage was not considered). (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

Fig. 13. Blade velocity vectors. The following blade characteristic velocities are shown: absolute wind velocity at blade position U (red), inverse of blade tangential speed at blade position -Ut (blue) and relative wind velocity w at blade position (green). (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

where d is given by:

frequency computer, required a total CPU time of about 4 days for each simulation.

d arctg Uy =Ux

(4)

7. Performance aerodynamic model The proposed performance analysis is based upon a simplied aerodynamic model, consisting in the analysis of kinematic and dynamic quantities every 4 rotor azimuthal position along blades trajectory, as exemplied in Fig. 12. For each azimuthal position, the main aerodynamic components of actions (such as rotor torque, power and the forces acting on the blade proles) can be directly determined by the CFD code through the integration of the NaviereStokes equations in the neighborhood of the wind turbine blade prole. Blade relative velocity is determined as the vector sum of the absolute wind velocity U and the inverse of blade tangential speed -Ut at blade position, being blade position identied by the coordinate of its centre of pressure, as exemplied in Fig. 13. Being q blade azimuthal coordinate and d the angle between absolute wind velocity at blade position and unperturbed wind direction, the angle g between absolute wind velocity at blade position and the inverse of blade translational speed at blade position can be written as:

being Ux and Ux the components of absolute wind velocity at blade position along x and y axes. The angle between the relative velocity and the tangent to the blade trajectory at each azimuthal position is therefore given by:

a arctg

U sin g U cos g Ut

(5)

and is named as blade relative angle of attack in the plane of the airfoil cross section. As reported in Fig. 12, 90 reference azimuthal positions (spaced 4 each other) were identied along the circumference representing the trajectory of the blades. For each position and for each temporal time step the Ux and Uy components were determined, allowing the calculation of the characteristics blade angles d, g and q. Being the value of blade tangential speed:

Ut uRrotor

(6)

it is possible to determine the values of the relative angle of attack a from Eq. (5) as a function of both blade azimuthal position and time. For each time step ti, the vector containing the local values of blade relative angle of attack as a function of azimuthal position j along blades trajectory can be written as:

g qd

(3)

Ati ati;0+ ati;0+ ; ati;4+ ; ati;8+ ; .; ati;356+

0 356 .

(7)

ranging j from to Some of ati,j values are not determinable because of the passage of the rotor blade (there is no ow eld at the considered azimuthal position along blades trajectory). These values are registered as not a number by the code and are not considered in the calculation. By composing all Ati vectors during a full rotor revolution, a matrix can be obtained, which is representative of blade relative angles of attack during rotor operation as a function of both angular position and time:

2 Ai;j

6 at 2;0+ 6 6 6 at 3;0+ 4 .

at1;0+

at1;4+ at2;4+ at3;4+

.

at1;8+ at2;8+ at3;8+

.

at1;16+ at2;16+ at3;16+

.

atn;4+

atn;8+

atn;16+

atn;0+

. . . . .

at1;356+ at2;356+ 7 7 at3;356+ 7 7

. 5

3 (8)

atn;356+

Fig. 12. Reference azimuthal position along blades trajectory.

being t1 e tn the number of time steps contained in a full rotor revolution. Figs. 14 and 15 display the evolution of blade relative angle of attack as a function of time for two different azimuthal positions along blade trajectory.

4926

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934 Table 4 Resulting average values of local angle of attack for 40 azimuthal position as a function of the number of azimuthal positions (both before and after the peak values) which are not considered for the present calculation. n 3 4 5 6

a [ ]
0.118257 0.113347 0.113611 0.113839

Deviation with respect to case n 4 [%] 4.33 0.00 0.23 0.43

Cr;ave

Pave 3 1=2rAs VN

(9)

for an incident wind speed of 9 m/s and as a function of the tip speed ratio, dened as:
Fig. 14. Evolution of the instantaneous values of blade relative angle of attack as a function of time for 40 azimuthal position, the red rectangles evidence the number of azimuthal positions which are not considered for the calculation; TSR 2.33.

TSR uRrotor =VN

(10)

Figs. 17e24 show the distribution of the instantaneous torque coefcient, dened as:

The passage of the rotor blade is visible as a peak in the instantaneous value of blade relative angle of attack. The average value of local angle of attack, due to the presence of the rotating turbine inside the ow eld at each time step, was determined by eliminating n azimuthal positions (4 in the present case, for a total angular spacing of 16 ) before and after the peak values (where the ow eld is locally perturbed by the passage of the rotor blade) and by calculating the arithmetic mean of the remaining angular values. The width of the angular sector perturbed by the local passage of rotor blade is a function of both rotor geometrical features (such as blade chord and, therefore, rotor solidity) and spatial discretization of the numerical analysis (such as the chosen number of azimuthal position analyzed along blades trajectory). From Figs. 14 and 15 and Table 4 it can also be seen that, being the distribution of blade relative angles of attack evenly spaced around their mean value (with the exclusion of the perturbed angular sectors), the proposed methodology is independent from the chosen value of n. 8. Results and discussion Fig. 16 represents the evolution of rotor average power coefcient, dened as:

CT q

Trotor q 2 1=2rAs Rrotor VN

(11)

and of the relative angle of attack, determined as described in the previous section, as a function of azimuthal position for a single rotor blade and for TSR ranging from 1.44 to 3.30. Also both positive and negative stall angles for a NACA 0021 airfoil at Re 300,000 are represented. As can be clearly seen, blade relative angles of attack decrease passing from lower to higher TSR values, because of the raising inuence of blade translational speed in the near-blade ow eld. It can also be seen that the maximum torque values are generated during the upwind revolution of the turbine and for azimuthal position where rotor blades are experiencing very high relative angles of attack, even beyond the stall limit, as can be seen in Figs. 25 and 26, representing relative velocity contours and relative pathlines for the angular position (92 ) of maximum instantaneous torque coefcient (and therefore power, being constant the angular velocity) and for the optimum value of TSR (2.33). As can be seen from Fig. 20, the azimuthal positions of maximum power extraction along blade trajectory are located

Fig. 15. Evolution of the instantaneous values of blade relative angle of attack as a function of time for 96 azimuthal position, the red rectangles evidence the number of azimuthal positions which are not considered for the calculation; TSR 2.33.

Fig. 16. Evolution of rotor average power coefcient as a function of TSR for an incident wind speed of 9 m/s; rotor operation at 8 different TSR values was numerically simulated.

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

4927

Fig. 17. Evolution of instantaneous torque coefcient and relative blade angle of attack as a function of blade azimuthal position for a single rotor blade; TSR 1.44.

Fig. 18. Evolution of instantaneous torque coefcient and relative blade angle of attack as a function of blade azimuthal position for a single rotor blade; TSR 1.68.

4928

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

Fig. 19. Evolution of instantaneous torque coefcient and relative blade angle of attack as a function of blade azimuthal position for a single rotor blade; TSR 2.04.

Fig. 20. Evolution of instantaneous torque coefcient and relative blade angle of attack as a function of blade azimuthal position for a single rotor blade; TSR 2.33.

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

4929

Fig. 21. Evolution of instantaneous torque coefcient and relative blade angle of attack as a function of blade azimuthal position for a single rotor blade; TSR 2.51.

Fig. 22. Evolution of instantaneous torque coefcient and relative blade angle of attack as a function of blade azimuthal position for a single rotor blade; TSR 2.64.

4930

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

Fig. 23. Evolution of instantaneous torque coefcient and relative blade angle of attack as a function of blade azimuthal position for a single rotor blade; TSR 3.09.

Fig. 24. Evolution of instantaneous torque coefcient and relative blade angle of attack as a function of blade azimuthal position for a single rotor blade; TSR 3.30.

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

4931

aerodynamic forces as a function of rotor blade azimuthal position should be further investigated. As can be seen, although the average single blade power coefcient is quite low, the instantaneous power coefcient locally (close to 92 azimuthal position) reaches the Betzs limit. Fig. 27 displays the distribution of the instantaneous power coefcient as a function of time for the three rotor blades and for optimum TSR, showing the contribution of each blade to overall rotor performance. Three peaks of the instantaneous power coefcient can be seen, being three periods contained in a full rotor revolution, each developing in an angular sector of:

J 2p=N 2p=3rad
Fig. 25. Relative velocity contours [m/s] and relative pathlines for rotor blade passing through 92 azimuthal position (wind is coming from the left) for optimum tip speed ratio (TSR 2.33).

(13)

inside the 4th and 5th Cartesian plane octants, as schematized in Fig. 27, showing the distribution of the instantaneous power coefcient, dened as:

Cp q

P q 3 1=2rAs VN

(12)

as a function of azimuthal position for a single rotor blade and for optimum TSR (2.33). This is probably due to the combination of a great energy extraction exerted by the rotor blade (due to the upwind operation of the rotor blade itself) and a relative high lever arm with respect to the rotor axis. Anyway, the distribution of

Once more, although the average rotor power coefcient is lower (0.429), the instantaneous power coefcient locally (close to 92 , 212 and 336 azimuthal position) exceeds the Betzs limit for almost 40 of rotor revolution, reaching a maximum value of 0.664. As a matter of fact, according to Betzs law, no turbine could capture more than 59.3 percent of the kinetic energy ux associated to wind [30]. Nevertheless, as pointed out by Jonkman [31], among the assumptions of Rankine-Froude actuator disc theory, the static pressure on the boundary of the streamtube portion enclosing the rotor disc should be equal to unperturbed ambient static pressure. This hypothesis is not veried for the present work, as can be seen from Figs. 28 and 29, representing the contours of the differential pressure coefcient (the atmospheric pressure value was set to 0 Pa), dened as:

Cp q

px; y 2 1=2rVN

(14)

Fig. 26. Evolution of instantaneous power coefcient as a function of blade azimuthal position for a single rotor blade. The location of maximum power extraction along blade trajectory, inside the 4th and 5th octants, is highlighted; TSR 2.33.

4932

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

Fig. 27. Evolution of instantaneous power coefcient as a function of time for the three rotor blades. The areas in red correspond to the overcoming of the Betzs limit; TSR 2.33.

Fig. 28. Contours of differential pressure coefcient cp(q) close to rotor disc (wind is coming from the left) for 92 azimuthal position (maximum local Cp(q)); TSR 2.33.

in correspondence to the rotor disc for 92 azimuthal position (maximum local Cp(q) value, exceeding the Betz limit) and 32 (minimum local Cp(q) value). A sudden pressure coefcient drop can be seen, concerning the whole rotor disc and the surrounding ow region, especially for

maximum local Cp(q) azimuthal position, where the thrust exerted from the rotor to the ow is almost doubled with respect to the case of minimum local Cp(q) azimuthal position, as can be seen from Table 5, reporting rotor instantaneous thrust coefcient, dened as:

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934

4933

Fig. 29. Contours of differential pressure coefcient cp(q) close to rotor disc (wind is coming from the left) for 32 azimuthal position (minimum local Cp(q)); TSR 2.33.

Table 5 Rotor thrust coefcient for 92 and 32 azimuthal positions; TSR 2.33. Azimuthal position [ ] 92 (maximum local Cp) 32 (minimum local Cp)


Ct q

Fx q 2 1=2rAs VN

(15)

Ct(q)[-] 1.12 0.69

for 92 and 32 azimuthal positions. The sudden pressure coefcient drop in the rotor disc region can be better seen from Fig. 30, showing a 3D representation of pressure coefcient contours for the Rotor-sub-grid zone for 92 azimuthal position (maximum local Cp(q)) and TSR 2.33 (local pressure peaks/drops in correspondence of the three rotor blades should not be considered). This phenomenon could be considered the main responsible for the local increase of the instantaneous power coefcient and should be further investigated.

Fig. 30. 3D representation of pressure coefcient contours for the Rotor-sub-grid zone, 92 azimuthal position (maximum local Cp(q)); TSR 2.33. The black arrow represents the direction of unperturbed wind.

4934

M. Raciti Castelli et al. / Energy 36 (2011) 4919e4934 [6] Paraschivoiu I, Delclaux F. Double multiple streamtube model with recent Improvements. J Energy 1983;7(3):250e5. [7] Paraschivoiu I. Double-Multiple streamtube model for Studying vertical-axis wind turbines. J Propulsion Power JulyeAugust 1988;4:4. [8] Mc Coy H, Loth JL. Up- and downwind rotor half Interference model for VAWT. 2nd Terrestrial Energy Systems Conference. Colorado Springs, CO: AIAA; December 1e3, 1981. [9] Mc Coy H, Loth JL. Optimization of Darrieus turbines with an upwind and downwind momentum model. J Energy 1983;7(4):313e8. [10] Lanzafame R, Messina M. Power curve control in micro wind turbine design. Energy February 2010;35(2):556e61. [11] Kishinami K, Taniguchi H, Suzuki J, Ibano H, Kazunou T, Turuhami M, et al. Study on the aerodynamic characteristics of a horizontal axis wind turbine. Energy AugusteSeptember 2005;30(11e12):2089e100. [12] Ceyhan O, Ortakaya Y, Korkem B, Sezer-Uzol N, Tuncer IH. Optimization of horizontal axis wind turbines by using BEM theory and genetic algorithm, 5th Ankara International Aerospace Conference. Ankara: METU; 17e19 August, 2009. AIAC-2009-044. [13] Zhang, J., Numerical modeling of vertical axis wind turbine (VAWT), Master thesis, Department of Mechanical engineering, Technical University of Denmark, December 20, 2004. [14] Claessens, M.C., The design and testing of airfoils for application in Small vertical axis wind turbines, Master of Science Thesis, Faculty of Aerospace engineering, Delft University of Technology, November 9, 2006. [15] Paraschivoiu I. Wind turbine design with emphasis on Darrieus concept. Polytechnic International Press, ISBN 2-553-00931-3; 2002. p. 76. [16] Sheldal, R. E., Klimas, P. C., Aerodynamic characteristics of Seven symmetrical airfoil sections through 180-degree angle of attack for use in aerodynamic analysis of vertical axis wind turbines. 1980. SAND 80e2114, Unlimited Release, UC-60. [17] Burton T, Sharpe D, Jenkins N, Bossanyi E. Wind energy handbook. John Wiley & Sons, Ltd; 2001. [18] Vassberg JC, Gopinath AK, Jameson A. Revisiting the vertical-axis windturbine design using advanced computational uid dynamics. AIAA Paper; 2005:0047. 43rd AIAA ASM, Reno, NV. [19] Simao Ferreira CJ, Bijl H, van Bussel G, van Kuik G. Simulating dynamic stall in a 2D VAWT: modeling strategy, verication and validation with particle image velocimetry data, the Science of making torque from wind. J Phys Conf Ser 2007;75. [20] Simao Ferreira CJ, van Bussel G, Scarano F, van Kuik G. 2D PIV Visualization of dynamic stall on a vertical axis wind turbine. AIAA; 2007. [21] Kumar V, Paraschivoiu M, Paraschivoiu I. Low Reynolds number vertical axis wind turbine for Mars. Wind Eng June 2010;34(4). [22] Raciti Castelli M, Benini E. Effect of blade inclination angle on a Darrieus wind turbine. Turbo Expo Technical Conference. Glasgow, Scotland, UK: ASME; June 14e18, 2010. GT2010-23332. [23] DAlessandro VD, Montelpare S, Ricci R, Secchiaroli A. Unsteady aerodynamics of a Savonius wind rotor: a new computational approach for the simulation of energy performance. Energy August 2010;35(8):3349e63. [24] Raciti Castelli M, Pavesi G, Battisti L, Benini E, Ardizzon G. Modeling strategy and numerical validation for a Darrieus vertical axis micro-wind turbine. Vancouver, British Columbia, Canada: ASME 2010 International Mechanical Engineering Congress & Exposition; November 12e18, 2010. IMECE201039548. [25] Raciti Castelli, M., Analisi numerica delle prestazioni di una micro-turbina eolica ad asse verticale modello Darrieus, PhD Thesis (in Italian), Universit di Padova, Italy, 2010, pp. 193e194. [26] Bradshaw P. Experimental uid Mechanics. Cambridge University Press; 1964. [27] Fluent Inc., Fluent Users Manual. 2006. pp. 52, 54, 59, 71, 143. [28] Cummings RM, Forsythe JR, Morton SA, Squires KD. Computational challenges in high angle of attack ow prediction. Progr Aerosp Sci 2003;39(5): 369e84. [29] McMullen M, Jameson A, Alonso JJ. Acceleration of convergence to a periodic steady state in turbomachinery ows, 39th AIAA aerospace sciences meeting & exhibit. Reno, NV: AIAA; January 8-11, 2001. 2001e0152. [30] Betz A. Das Maximum der theoretisch moglichen Ausnutzung des Windes durch Windmotoren. Z das gesamte Turbinenwesen 1920;26:307e9. [31] Jonkman JM. Modeling of the UAE wind turbine for renement of Fast_AD. NREL/TP-500-34755; December 2003. p. 7.

9. Conclusions and future work In this paper, a numerical model (based on the application to CFD of the aerodynamic principles which are currently applied to BE-M theory for rotor performance prediction) for the evaluation of energy performance and aerodynamic forces acting on a straightbladed vertical-axis Darrieus wind turbine has been presented. A simplied aerodynamic model (based on the analysis of kinematic and dynamic quantities at discrete and xed rotor azimuthal positions along blades trajectory) was also presented, allowing the correlation between ow geometric characteristics (such as blade angles of attack) and dynamic quantities (such as rotor torque and blade tangential and normal forces). The results of a 2-D full campaign of simulations were proposed for a classical NACA 0021 three-bladed rotor. Through the analysis of the distribution of the instantaneous torque coefcients and of the relative angles of attack as a function of azimuthal position for a single rotor blade, ow eld characteristics were investigated for several values of tip speed ratio, allowing a comparison between rotor operation at optimum and lower Cp values. The obtained results have shown the reduction of blade relative angles of attack passing from lower to higher TSR values, due to the increasing inuence of blade translational speed in the near-blade ow eld. It has also been shown that the maximum torque values are generated during the upwind revolution of the turbine and for azimuthal positions where rotor blades are experiencing very high relative angles of attack, even beyond the stall limit. The azimuthal positions of maximum power extraction along blade trajectory have been located inside the 4th and 5th octants, probably due to the combination of a great energy extraction exerted by the rotor blade (due to the upwind operation of the rotor blade itself) and a relative high lever arm with respect to the rotor axis. The distribution of aerodynamic forces as a function of rotor blade azimuthal position should be further investigated. Finally, it has been shown that, although the average rotor power coefcient is rather lower, the instantaneous power coefcient locally exceeds the Betzs limit three times per rotor revolution. This phenomenon, probably caused by a sudden pressure coefcient drop concerning the whole rotor disc and the surrounding ow region (thus violating the assumptions of Rankine-Froude actuator disc theory) should be further investigated. References
[1] Glauert H. Airplane propellers, aerodynamic theory, vol. 4. New York: Dover Publication Inc; 1963. Division L, 169e360. [2] Templin RJ. Aerodynamic performance of vertical-axis wind machines. ASME; 1975. Paper 75-WA/ENER-1. [3] Strickland, J.H.: The Darrieus turbine: a performance prediction model using multiple streamtube. 1975. SAND 75e0431. [4] Read S, Sharpe DJ. An extended multiple streamtube theory for vertical axis wind turbines. Department of M.A.P. Engineering Report. United Kingdom: Kingston Polytechnic, Kingston Upon Times; 1980. [5] Paraschivoiu I. Double-Multiple streamtube model for Darrieus wind turbines. 2185. NASA Conference Publication; May 1981.