Content uploaded by Krzysztof Rogowski

Author content

All content in this area was uploaded by Krzysztof Rogowski on May 23, 2019

Content may be subject to copyright.

TASK QUARTERLY vol. 22, No 3, 2018, pp. 195–209

VELOCITY FIELD AROUND DARRIEUS

WIND TURBINE ROTOR USING ACTUATOR

CELL MODEL AND OTHER CFD METHODS

KRZYSZTOF ROGOWSKI AND KLAUDIA ROGOWSKA

Warsaw University of Technology

The Institute of Aeronautics and Applied Mechanics

Nowowiejska 24, 00–665 Warsaw, Poland

(received: 4 May 2018; revised: 11 June 2018;

accepted: 29 June 2018; published online: 6 July 2018)

Abstract: The main purpose of this work is to analyze the usefulness of the active cell model

(ACM) developed by the author of this article to estimate the ow eld around a single-bladed

vertical-axis wind turbine (VAWT) with the Darrieus-type rotor. The obtained ow velocity elds

were compared with the experimental values taken from the literature available on the Internet.

Additionally, the ow elds around the rotor and the aerodynamic forces were determined using

the following approaches: the -RNG turbulence model, the scale-adaptive simulation (SAS) and

the laminar model. The velocity proles behind the turbine rotor obtained with all numerical

approaches are consistent with the experiment. The aerodynamic blade loads obtained using

numerical methods also appear to be satisfactory.

Keywords: CFD, vertical-axis wind turbine, aerodynamic blade loads, wake modeling

DOI: https://doi.org/10.17466/tq2018/22.3/d

1. Introduction

1.1. Vertical-axis wind turbines

Even though the share of the vertical-axis wind turbine (VAWT) is currently

decreasing in Poland, the number of patents related to them continues to increase.

There is also a growing interest in this topic in the world as evidenced by the

number of publications.

The archetype of the VAWT is considered to be a Persian windmill (Fi-

gure 1a) from the 2nd century BC used to drive a quern. This device resembled

a water wheel in which half of the rotor was shielded from the wind. The source

of torque in these devices was the drag dierence on the windmill blades. Such

turbines are called drag-driven devices [1].

196 K. Rogowski and K. Rogowska

Devices with the Darrieus rotor operate on a completely dierent princi-

ple. Darrieus wind turbines are an alternative to traditional horizontal-axis pro-

peller-type machines. This lift-driven device proposed by a French aeronautical

engineer, Georges Jean Marie Darrieus was patented by U. S. Patent Oce in

1931 [2]. According to the inventor, the rotor should have properly curved bla-

des. The shape of this curvature should be close to a rotating rope. This shape

is called the Troposkien shape (from Greek: – turning; – rope).

A rotor blade having such a shape only transfers the tensile and compressive

stresses. The aerodynamic torque in this device comes from an aerodynamic force

generated on the rotor blades that have aerodynamic proles. The silhouette of

a typical two-bladed Darrieus wind turbine is shown in Figure 1b. There are also

modications of this rotor, Figure 1c shows, for example, a rotor with straight

blades. It is a VAWT created in 1980 in the well-known company McDonnell Do-

uglas. The French engineer’s idea was forgotten for almost half a century, then it

was discovered in the mid-1970s, mainly in the USA and Canada. At that time,

extensive research work was undertaken, resulting in the creation of wind farms in

California in Tehachapi, Altamont Pass and San Gorgonio. About 500 machines

of this type were built there with the rated powers of 150–300kW. In 1987, the

largest wind turbine of this type was built – the Canadian EOL with an installed

capacity of 3.8MW [1].

The Darrieus concept was again forgotten when the idea of oating wind

turbines appeared in recent years. The low centre of gravity of the rotor is the

reason why the Darrieus wind turbine is better for oshore applications than

a traditional wind turbine with a horizontal axis of rotation [3].

(a) (b) (c)

Figure 1. (a) Persian windmill silhouette [1]; (b) Darrieus wind turbine [1];

(c) H-type Darrieus wind turbine [1]

Velocity Field around Darrieus Wind Turbine Rotor197

1.2. State of the art

Currently, the most well-known tool for analyzing the aerodynamic perfor-

mance of the Darrieus rotor is the double-multiple streamtube (DMS) model. It

is a method that uses two actuator disks working in tandem. One of these disks

represents the upwind part of the rotor and the second the downwind part. This

method takes into account one component of the ow only, therefore, it is suitable

for rotors with low solidity. The calculation of aerodynamic forces is possible due

to the combination of the actuator disc theory with the blade element theory [4].

However, the aerodynamic blade loads obtained with this method are not satis-

factory [5]. Strickland et al. [6] proved in their report that an aerodynamic wake is

determined wrongly when these methods are used. The same authors have deve-

loped a much more accurate approach based on vortex equations. This approach

is still used, but it is more computationally expensive. An interesting approach

is the actuator cylinder model developed by Madsen [3]. In this approach the

curvilinear surface is used instead of a at surface of the actuator disc as in the

case of the DMS approach. On this circular surface Volume forces are applied in

the radial direction.

Nowadays Computational Fluid Dynamics (CFD) methods have been de-

veloped very strongly. They solve the average Navier-Stokes equations whereas

turbulence is solved by means of additional models called turbulence models. Ho-

wever, these methods are very costly computationally because they require mode-

ling using dense grids [7, 8]. Although the rotor calculation shown in Figures 1b

or 1c is theoretically possible using CFD methods, this is not usually done. This is

because these methods incorrectly determine the critical attack angle [9]. In ad-

dition, the use of a large number of processors is not economical. This determines

the need to look for new solutions. The presented article is an extended version of

Rogowski’s article [10]. The article presents a comparison of speed distributions

obtained by means of various numerical approaches and the actuator cell model

(ACM) approach developed by Rogowski [10].

2. Wind turbine

2.1. Wind turbine rotor

The wind turbine rotor presented in this paper consists of only one straight

blade with the NACA 0012 airfoil. The chord length of the airfoil, , is 9.14cm. The

rotor radius, , was established to be 0.61 m giving the rotor solidity, ,

of 0.15. The most important parameter determining the rotor operation is the

so called tip speed ratio, TSR. This parameter is dened as a non-dimensionless

ratio of the rotor blade velocity (the tangential velocity of the blade), 𝑡, to the

undisturbed ow velocity, 0. The tip speed ratio can be written as:

TSR 𝑡

0

0

where: is the angular velocity of the rotor. In the case of the examining rotor the

tip speed ratio value is 5.0. This means that the tangential velocity of the rotor is

198 K. Rogowski and K. Rogowska

ve times larger than the undisturbed ow velocity. According to Paraschivoiu [4],

the optimum value of the rotor power coecient, 𝑝, for rotors that have a low

solidity value () is obtained for TSRs of 4–5. The rotor power coecient

is dened as: 𝑝power extracted by rotor

For a typical Darrieus-type rotor, the maximum value of 𝑝is around 0.4.

Therefore, in these simulations a rotor operating at the optimal TSR was tested.

For optimal TSRs, the so called, secondary eects are negligible [4]. Moreover,

local angles of attack do not exceed static critical angles of attack, thanks to this,

the phenomenon of dynamic stall does not appear. According to Paraschivoiu [4]

the secondary eects are associated with e.g.: the presence of struts, the rotor

geometry, the presence of spoilers, etc. During simulations the angular velocity of

the rotor was 0.75rad/s and the undisturbed ow velocity was 0.091m/s.

The numerical experiment performed in this work corresponds to the

experiment of Strickland et al. [6]. This experiment was performed in a water

towing tank. The selected operating uid was water due to the lower angular

velocity of the rotor at the same tip speed ratio. In addition, studies conducted

in water made it easier to visualize the ow eld. More information about the

experiment settings can be found in two reports and in a scientic article [6,

11, 12]. Figure 2 presents a sketch of a one-bladed rotor.

Figure 2. Silhouette of one-bladed vertical axis wind turbine

2.2. Aerodynamic characteristics of the rotor

The Darrieus wind turbine is a lift-driven device consisting of a number of

curved blades. A rotor with straight blades can be called a Darrieus-type rotor.

The principle of the rotor operation results from the creation of a lifting force

on its blades. The local velocity at the rotor, , is lower than the undisturbed

ow velocity. The relative velocity, , is the sum of the vector of the tangential

velocity of the blade and the velocity vector . The angle between the relative

velocity vector and the chord line is called the angle of attack, . This angle

changes with the azimuth angle, , and depends on the tip speed ratio and the

Velocity Field around Darrieus Wind Turbine Rotor199

blade Reynolds number. The azimuth angle uniquely determines the position

of the rotor and is measured as shown in Figure 3. The direction of the drag

is the same as the direction of the relative velocity , whereas the lift force is

perpendicular to this velocity. The aerodynamic force components, the lift force

and the drag, projected onto the normal and tangent directions give the normal

and tangential forces, respectively.

During the experiments of Strickland et al. [6] both aerodynamic blade

loads as well as velocity proles behind the rotor were measured. The velocity

proles were measured at the distance of one rotor diameter downstream behind

the rotor, as shown in Figure 3. The wake velocity component 𝑥is normalized

by the undisturbed ow velocity 0.

Figure 3. Two-dimensional rotor model; forces acting on the rotor blade and velocity vectors

3. Methods

In this work, numerical methods were used to assess the aerodynamic

properties of the rotor: aerodynamic forces and ow parameters in the rotor area.

The classic full CFD approaches were used as well as the method developed by

Rogowski [10] – the Actuator Cell Model (ACM).

3.1. Ful l CFD approach

The full CFD approach, for the purposes of this article, means the numerical

approach in which the boundary layer is modeled using a turbulence model and

an appropriate grid near the blade edges. The ow around the Darrieus rotor

(or a Darrieus-type rotor) cannot be considered as steady. Fluctuations in the

rotational torque of a one-, two-and even three-bladed rotor are so large that the

ow must be considered as unsteady. Therefore, the moving mesh technique was

utilized in all simulations. In this concept, the rotor is surrounded by a large square

stationary area, a computational domain. Moreover, there is a smaller circular

area in the vicinity of the turbine rotor which rotates during the simulation with

the same angular velocity as the rotor. The data between these two areas is

200 K. Rogowski and K. Rogowska

exchanged via the interface. Rogowski et al. proved in their previous works [13, 14]

that the ratio of the square side length to the rotor diameter should be at

least 10. Otherwise, the power coecient results may be somewhat overstated.

Interestingly, the result obtained by Rogowski et al. [13, 14] is right both for

drag-driven rotors (such as e.g. Savonius rotors) as well as for Darrieus-type

rotors. Figure 4 shows schematically the concept of the numerical approach and

the boundary conditions: the velocity inlet, the pressure outlet, the walls, and the

symmetry. The symmetry boundary condition can be used as a zero-shear stress

wall boundary condition.

Figure 4. Numerical approach and boundary conditions

In all the cases presented in this article, the rotor was modeled as a two-di-

mensional object consisting of one airfoil rotating with respect to the axis of

rotation. In the two examined cases, the two-dimensional Navier-Stokes (NS)

equations were taken into account for the RNG -turbulence model and for

the laminar model. In the case of the Scale-Adaptive Simulation approach, the

two-dimensional ow was considered using three-dimensional NS equations. The-

refore, the presented approach is essentially a 2.5D approach.

As mentioned in the previous paragraph, one of the approaches used was

the -RNG turbulence model. This model solves two variables: the turbulence

kinetic energy and the rate of its dissipation. This approach was developed using

the so called Re-Normalization Group methods, hence, the acronym RNG.

In the laminar approach, no turbulence model is considered. Two equations

of momentum and equation of continuity are solved only. Therefore, the presented

approach can be treated as a direct numerical simulation (DNS) approach.

The scale-adaptive simulation (SAS) model is an improved unsteady-ave-

raged Navier-Stokes (URANS) approach. It enables the solution of a turbulent

spectrum in an unstable ow. This is a model developed on the well-known shear

stress transport (SST) formulation.

More about the models used can be found, for example, in the ANSYS Fluent

documentation.

Velocity Field around Darrieus Wind Turbine Rotor201

3.2. Actuator cell concept

The actuator cell model (ACM) is an original approach developed by

Rogowski [10]. Thanks to the moving mesh technique, it is also a model that

allows analyzing unsteady ows. In this approach, the boundary layer around the

airfoil of the rotor blade is not modeled. The presence of the blade is taken into

account by means of momentum sources, which are introduced into the laminar

Navier-Stokes equations. In other words, in the ACM approach aerodynamic loads

do not result from the solution of uid motion equations. These loads can come, for

example, from the blade element theory. The technique devised by Rogowski [10]

is still being developed. In this article, the authors want to show that the proposed

method can correctly determine the velocity eld around the rotor and behind the

rotor based on the set aerodynamic force function (aerodynamic load as a function

of the azimuth). The function of aerodynamic forces used in the presented research

was created based on the experimental measurements of Strickland et al. [6]. For

a given rotor position determined using an azimuth , the aerodynamic force

components are interpolated and entered into the NS equations as:

𝑥,𝑦 𝑥,𝑦

𝑐

where: 𝑐is the volume of the mesh cell, 𝑥and 𝑦are aerodynamic force

components in the CFD solver system. The ACM model has been implemented

in the commercial solver, ANSYS Fluent, which uses the Cartesian coordinate

system. The geometrical model of the rotor tested in this work is related to the

Cartesian coordinate system as shown in Figures 2–4. In the Cartesian coordinate

system, the components of aerodynamic forces are expressed as:

𝑥𝑁sin𝑇cos

𝑦𝑁cos𝑇sin

The aerodynamic forces dened by means of Equations (4)–(5), presented in the

result section 4, have been normalized:

𝑥,𝑦 𝑥,𝑦

2

0

where: is the uid density.

3.3. Mesh distribution

Both the ACM approach and full CFD models require the use of calculation

grids. The grid used for the URANS approach with the -RNG turbulence model

and for the laminar model is shown in Figures 5a and 5b. This grid consists of

a structural mesh near the edge of the rotor blade and a non-structural mesh in

the remaining area. The use of the structural grid provides a better representation

of the boundary layer near the edge of the blade. During all simulations, the wall

parameter was kept less than 1. Figure 5c presents the grid used for a 3D

simulation with the SAS model. The grid distribution in the plane perpendicular

202 K. Rogowski and K. Rogowska

to the rotor axis of rotation is the same as in the case of 2D CFD simulations (the

same as presented in Figures 5a and 5b). The thickness of the three-dimensional

computational domain has a length equal to two chords of the blade. Figures 5d

to 5f present the mesh distribution for the ACM approach. Figure 5f shows the

densest resolution of the grid near the cell into which the momentum sources are

introduced. The number of items for the 3D mesh is 3.9 million, for the 2D mesh

it is 131 958 items, whereas in the case of the ACM model, the number of items is

43 707. All the grids presented in this article were thoroughly investigated due to

the cell density. A detailed description of these tests and a detailed description of

these grids can be found in the articles [7, 10].

(a) (b) (c)

(d) (e) (f)

Figure 5. Mesh for full CFD models (Figures a–c) and mesh for ACM model (Figures d–f)

4. Results

4.1. Aerodynamic blade loads

This chapter presents a comparison of the aerodynamic blade loads obtained

by means of: the laminar model, the SAS approach and the -RNG turbulence

model. The results obtained are compared with the experimental results of

Strickland at al. [6]. As mentioned in Chapter 3.2, in the case of the ACM

model, the aerodynamic blade loads were not determined. The components of

aerodynamic blade loads, in the Cartesian coordinate system, were presented

in a dimensionless form (according to Equation (6)) in Figure 6. The azimuth

changes from 0 to . However, Figure 6 shows the results for the azimuth in

the range from 0 to . It was done to show that the blade loads change the

same in each rotation of the rotor (they are repeated in every rotation). As can

Velocity Field around Darrieus Wind Turbine Rotor203

be seen in Figure 6, the results of the 𝑥force component obtained by means

of dierent numerical approaches are convergent with the experimental values.

Some discrepancies are visible in the case of the 𝑦component in the azimuth

range between 100 and . In the azimuth range up to , it seems that

the best results of the 𝑦force component are given by the SAS approach.

The results of aerodynamic blade loads presented in Figure 6 obtained with

the laminar model are amazingly good. We can see that there are considerable

oscillations of aerodynamic blade loads resulting from the vortex structures that

are formed on the blade surfaces, averaged by e.g. the RNG -model. However,

these aerodynamic loads oscillate around the experimental values or around values

calculated by other methods used.

(a) (b)

Figure 6. Aerodynamic blade load components. Comparison between full CFD approach and

experimental data [1]

4.2. Velocity proles

Figure 7 presents the proles of the ow velocity component 𝑥obtained

at a distance of one rotor diameter after the rotor. The velocity components were

normalized by the undisturbed ow velocity 0. The comparison of numerical and

experimental results conrms the eectiveness of the methods used for analyzing

the aerodynamic wake downstream behind the rotor. The ACM model also gives

very good quantitative velocity results.

4.3. Static pressure distribution

As can be seen in Figure 6, the blade load component values in the azimuth

range between 180 and 360 are rather smaller compared to the values of these

forces in the remaining rotor area. In the trade literature, the rotor area in the

azimuth range between 0 and is commonly referred as the upwind part of

the rotor while the remaining part is called the downwind part of the rotor. In the

upwind part of the rotor, the ow is less disturbed than in the downwind part.

Figures 8–9 present the static pressure distributions for the upwind and

downwind parts of the rotor, respectively. The pressure dierences between these

204 K. Rogowski and K. Rogowska

Figure 7. Velocity proles downstream behind the rotor: experiment (black circles), ACM

(blue solid lines), -RNG (red dashed lines), laminar (magenta dashed lines), SAS (green

dashed lines)

areas are expressive. Figures 8f9 also show the pressure distributions obtained

by various numerical methods. If we compare the pressure distributions for the

azimuth of , they look very similar. This is related to the components of the

aerodynamic blade loads 𝑥and 𝑦, which in the range of about –are

almost the same for all calculation methods and for the experiment (cf. Figure 6).

The ACM model indicates pressure results similar to the laminar model and SAS.

Employing the -RNG turbulence model the pressure is more smooth in the

whole azimuth range. The results obtained by means of the laminar model and

the SAS model indicate the appearance of ow instabilities near the trailing edge

of the azimuth of . This is not visible in the case of the two-equation -RNG

model. Many works indicate the possibility of ow detachment in this part of the

rotor area, e.g. [15]. Depending on the rotor, the operating conditions (tip speed

ratio) of this phenomenon may be more or less strong. This is a surprising eect

because, according to previous observations [4], for the optimal tip speed ratios,

the local blade angles of attack should be small and the stall phenomenon should

not occur.

4.4. Vorticity magnitude

Another method of assessing the velocity eld used in this work is to com-

pare the vorticity magnitude distributions (Figure 10). The vorticity describes the

Velocity Field around Darrieus Wind Turbine Rotor205

Figure 8. Static pressure distributions for rotor upwind part

tendency of the ow eld to rotate. In uid mechanics the vorticity pseudovector

describes the local spinning motion of the ow near some point. Mathematically,

in the Cartesian coordinate system, this vector can be dened as:

𝑧

𝑦

𝑥

𝑧

𝑦

𝑥

where: is the del operator, is the velocity eld. In the case of a two-dimensional

ow this vector has only a component:

𝑦

𝑥

𝑧

Figure 10 shows contour maps of the vorticity magnitude, i.e. the square root

of the vorticity vector described by Equation (8). The history of the vorticity

magnitude shown in this gure indicates that the velocity eld in the upwind

part of the rotor is slightly disturbed. In the second part of the rotor, the

interaction between the aerodynamic wake created by the rotor blade in the

upwind part and the same blade is visible (wake-blade interaction). Comparing

the aerodynamic wakes provided by the SAS and ACM models we can observe an

amazing compatibility. It seems that it is much more reliable than in the case of

the -RNG model. The aerodynamic wake provided by the laminar model is, as

expected, heavily disturbed. This model is not equipped with the “mathematical

206 K. Rogowski and K. Rogowska

Figure 9. Static pressure distributions for rotor downwind part

mechanism” of dissipative turbulent kinetic energy as in the case of the RNG

model.

To better understand the phenomenon of blade-wake interaction, the ad-

vantages of the SAS model, which in the case of an unstable ow behaves like

a large eddy simulation (LES) model, and the three-dimensional ow model were

used. Figure 11 shows the history of vorticity in the area of the three-dimensio-

nal Darrieus-type rotor. The gure shows vortex iso-surfaces with the same value

(0.15 1/s) but opposite turns. The blue color suggests the clockwise vorticity di-

rection and the red color the opposite direction.

A ow instability appears for the azimuth of –. This resulting ow

instability evolves and moves with the main stream. Since the velocity of the

rotor blade is larger than the main stream velocity, the blade-wake interaction is

visible for the azimuth of around –. The second but smaller blade-wake

interaction is visible near the azimuth of .

5. Conclusions

The purpose of this work was to estimate the velocity eld around a one-bla-

ded vertical-axis wind turbine using full CFD models and the author’s ACM me-

thod. The results presented in this article have shown that:

The advantage of testing a single-bladed rotor is the ability to analyze the

interaction of the aerodynamic wake with a rotor blade. The analyses carried

Velocity Field around Darrieus Wind Turbine Rotor207

Figure 10. Contour maps of vorticity magnitude for dierent numerical approaches

out by the authors of this article have shown that the aerodynamic wake behind

a vertical-axis wind turbine is very complicated compared to rotors of classic

wind turbines with a horizontal axis of rotation.

Rotors of the Darrieus-type, due to large uctuations in aerodynamic forces,

cannot be tested using the RANS approach.

The use of standard turbulence models, such as e.g. the -family, is not

sucient to study the aerodynamic phenomena of the Darrieus rotor. The SAS

model shows many more details of the ow.

The ACM model gives satisfactory results in the ow eld. They are comparable

with other CFD methods. The ACM model is further studied in order to be able

to estimate the aerodynamic forces on the basis of given prole characteristics.

The intention of the authors of the paper is also to study three-dimensional

rotors using the developed method.

In its present form, the ACM method is suitable for testing mixers of some

types.

The laminar model surprisingly well predicts the aerodynamic blade loads and

the velocity proles downstream behind the rotor. Rogowski [16] has also proved

208 K. Rogowski and K. Rogowska

Figure 11. Iso-surfaces of vorticity for SAS simulations

this in the case of another single-bladed rotor. This provides the basis for testing

the transitional laminar-turbulent model.

Acknowledgements

The presented numerical computations were performed in the Interdiscipli-

nary Centre for Mathematical and Computational Modeling of the Warsaw Uni-

versity. The current work was prepared as part of the computing grant GB73–5.

References

[1] Szuster J T 2000 Wind generators with vertical axis of rotation, National Forum of

Renewable Energy, Łódź (in Polish)

[2] Darrieus G J M 1931 U. S. Patent 1834018

[3] Madsen H A, Paulsen U S and Vitae L 2012 J. Phys.: Conf. Ser. 555 12065

[4] Paraschivoiu I 2002 Wind Turbine Design With Emphasis on Darrieus Concept, Inter-

national Press, Canada

[5] Castelin D 2015 Dynamic stall on vertical Axis Wind Turbines, MSc Thesis, TU Delft

[6] Strickland J H, Smith T and Sun K 1981 A Vortex Model of the Darrieus Turbine: An

Analytical and Experimental Study, Technical Report SAND’81–7017

[7] Rogowski K, Hansen M O L, Maroński R and Lichota P 2016 J. Phys.: Conf. Ser.

753 22050

[8] Rogowski K, Maroński R and Hansen M O L 2018 J. Theor. App. Mech. 51 (1)203

[9] Rogowski K, Hansen M O L, Hansen R, Piechna J and Lichota P 2018 J. Phys.: Conf.

Ser. 1037 22019

Velocity Field around Darrieus Wind Turbine Rotor209

[10] Rogowski K 2018 J. Phys.: Conf. Ser. 1101 12028

[11] Strickland J H, Webster B T and Nguyen T 1979 A Vortex Model of the Darrieus Turbine:

An Analytical and Experimental Study, Technical Report SAND’79–7058

[12] Strickland J H, Webster B T and Nguyen 1979 J. Fluid Eng. 101 500

[13] Rogowski K 2014 Analysis of Performance of the Darrieus Wind Turbines, Ph.D. Thesis,

Warsaw

[14] Rogowski K and Maroński R 2015 J. Theor. App. Mech. 53 (1)37

[15] Bangga G, Hutomo G, Wiranegara R and Sasongko H 2017 J. Mech. Sc. and Techn. 31

(1)261

[16] Rogowski K 2018 J. Mech. Sc. and Techn. 32 (5)2079