Deliverable 3.3: Results of a parametric study of flow devices, guidelines for design

Transcription

1 Deliverable 3.3: Results of a parametric study of flow devices, guidelines for design María Aparicio, Raquel Martín, Arturo Muñoz, Alvaro González CENER April 1 st, 2016 Agreement n.: FP7-ENERGY / n Duration: November 2013 to November 2017 Coordinator: ECN Wind Energy, Petten, The Netherlands Supported by: This project has received funding from the European Union s Seventh Programme for research, technological development and demonstration under grant agreement No FP7-ENERGY / n

2 Document information Document Name: Confidentiality Class Results of a parametric study of flow devices, guidelines for design PU Document Number: D3.3 Editor: Contributing authors: CENER María Aparicio, Arturo Muñoz, Raquel Martín, Alvaro Gonzalez CENER Méndez, B., Gómez-Iradi, S., Munduate, X. USTUTT Jost, E., Lutz, T. UoG Barakos, G., Colonia, S. TU DELFT Florentie, L. ECN Caboni, M. DTU WIND Ramos-García, N., Troldborg, N., Sørensen, N. NTUA Prospathopoulos, J., Chassapoyiannis, P., Manolesos, M., Diakakis, K., Voutsinas, S. GE Stettner, M. Review: ECN Gerard Schepers Date: 25/04/16 WP: WP3: Models for flow devices and flow control Task: Task 3.3 Page 2 of 141 WP no.: 3.3

3 Table of contents 1. Introduction 5 2. Scope and content 7 3. Selection of parameters 8 4. Test cases: Matrices TE flap study Study of flap length and shape Study of flap angle Study of unsteady aerodynamic response Study of flap influence in the 3D rotor LE flap study Study of unsteady aerodynamic response VGs study Study of VGs height and position in the chord direction Study of VGs length, angle wrt flow and internal and external distance Study of VGs influence in the 3D rotor Details about codes and simulations WMB (CENER) Code description Simulation details AdaptFoil2D (CENER) Code description Simulation details FAST (CENER) Code description Simulation details MaPFlow (NTUA) Code description Simulation details Foil2w (NTUA) Code description Simulation details GENUVP (NTUA) Code description Simulation details FLOWer (USTUTT) Code description Simulation details OpenFOAM (TU Delft) Code description Simulation details HMB (University of Glasgow) Code description Simulation details MIRAS (DTU) 25 Page 3 of 141 WP no.: 3.3

4 Code description Simulation details Aero-Module (ECN) Code description Simulation details Results of the parametric study TE flap Study of flap length and shape Study of flap angle Study of unsteady aerodynamic response Study of influence in the 3D rotor LE flap Study of unsteady aerodynamic response VGs Study of VGs height and position in the chord direction Study of VGs length, angle wrt flow and internal and external distance Study of VGs influence in the 3D rotor Conclusions and guidelines for design 125 Nomenclature 131 References 132 Acknowledgements Appendix Appendix Page 4 of 141 WP no.: 3.3

5 1. Introduction The present document is related to the activities carried out within WP3 under Task 3.3. WP3 as a whole aims at the development and assessment of models for flow devices and flow control on large wind turbine blades. As a first step to this aim, the creation of a CFD and experimental database was defined and completed in Task 3.1. The information related to this database was included in deliverable 3.1 (Ref. [1]). The second task of the work package was the development of aerodynamic codes for modelling of flow devices on aerofoils and rotors, reported on deliverable 3.2 (Ref. [4]). The present document corresponds to the third deliverable of the work package and the main objective is the parametric study of flow devices and the impact on the performance of aerofoils and blades. As the rest of the work package, the study is focused on vortex generators and LE / TE flaps. Participants in task 3.3 include CENER, ECN, DTU WIND, NTUA, TU DELFT, UoG, USTUTT, GE and LM. Using the codes validated in Task 3.2, it is possible to perform a parametric study of the LE / TE flaps and vortex generators from an aerodynamic point of view, in order to identify the trends and limitations on the aerofoil and blade of the AVATAR and INNWIND rotor. The Task is structured as follows: 1. First, a detailed matrix of cases is defined for the evaluation of different concepts of LE / TE flaps and VGs on 2D airfoils and blade sections. Different parameters are evaluated related to dimensions of the flow devices, distribution and operating conditions. 2. The configurations and working conditions selected for the flaps and VGs equipped in aerofoils and blades are computed using the tools developed in Task 3.2. A crosscomparison is performed in order to validate the trends and analyse the differences between the codes. 3. In the analysis of the results, some performance indicators are defined to quantitatively evaluate the potential of each concept. The KPI s are related to loads, efficiency or production. 4. After the study, some guidelines for the design of flow control devices are established. First, a brief description of the scope and content of the task is mentioned in chapter 2. In chapter 3, the selection of parameters for the parametric study is explained. Chapter 4 deals with the description of test matrices for the simulations and for the detailed parametric study. In chapter 5, the different tools used by the participants are mentioned, including some details about the simulations. Chapter 6 presents the results of the different parametric studies performed. Finally, chapter 7 mentions the main conclusions extracted from the work and establishes some guidelines for design. In Appendix 1, the output specifications are included, while in Appendix 2 a FORTRAN code is included for the curved flap shape. The TE flap study includes the evaluation of the effect of different design parameters (flap length, shape and angle). In addition, the influence of the unsteady conditions is analysed with Page 5 of 141 WP no.: 3.3

6 flap oscillations and ramps. Finally, the location and extension of the TE flap on the blade is studied. The objective is to understand the potential of the TE flap for load reduction purposes (and power control). For the LE flap, the unsteady aerodynamic behaviour of a pitching airfoil with a LE flap in different static positions is analysed. The objective is the evaluation of the use of a LE flap for the modification of the unsteady separated flow progress and the dynamic stall phenomenon. Different design parameters of the VGs are considered (height, length, angle with respect to the flow and internal and external distances) in the study of the airfoil performance with the aerodynamic devices. In addition, the influence of the location and extension of the VGs in the rotor performance is evaluated. The objective is the study of the potential of the VGs configurations for improving the wind turbine performance. Page 6 of 141 WP no.: 3.3

7 2. Scope and content This report presents the parametric study performed within the Task 3.3: Parametric study of the impact of flow devices on the performance of aerofoils and blades, in accordance to the DoW. The main scope is the investigation of the impact of the flow control devices (design aspects) in the airfoil and rotor performance. However, as a follow-up of T3.2, there are still modelling activities going on, and there is a secondary scope related to the characterization of the quality of the results obtained with the different models used. Deliverable 3.3 is a report of the parametric evaluation of geometry, distribution and actuation of LE/TE flaps and vortex generators using different levels of modelling (CFD, vortex methods and BEM). Results and analysis are included in this document. The key points included in the deliverable are: 1. Selection of parameters for the analysis of flow control devices. In terms of geometry, at least the size of the flaps and vortex generators will be evaluated. In terms of distribution, the extension of the devices in the blade and the location will be considered. In terms of flap actuation, the reduced frequencies and amplitudes will be taken into account. The operating conditions will be included in terms of the reference rotors context. 2. Description of the test matrix. At least 3 different values for each of the above mentioned parameters are considered, for the same 6 aerofoils applied in deliverable 3.1 related to the AVATAR and INNWIND rotor (LE and TE flaps are applied at the middle and tip sections and VGs at the root region). In the case of CFD codes, only a selection of the full list of cases will be computed due to computational limitations. The working team has been split to minimize the work load. 3. Cross comparison between the results of different models and codes. 4. Conclusions of the parametric study. It is important to estimate the influence of the variation of the above given parameters on the aerodynamic performance, in particular in loads, efficieny and production. 5. Guidelines for design. The optimum ranges are estimated for the practical application of flow devices in terms of the parameters and computations above given. The study parameters are selected based on different categories: dimensions, operating conditions, distribution or configuration in the blade. The idea behind the selection is to include design parameters and operational conditions of the flow control devices which have a major impact in the aerodynamic performance of rotors and blades. The effect of those parameters in the aerodynamic performance, evaluated through some selected KPIs, is used in order to establish the design guidelines and for a comprehensive evaluation of the practical use of the flow devices in the reference wind turbines. Page 7 of 141 WP no.: 3.3

8 3. Selection of parameters The parameters are selected from different categories: Dimensions of the control device Operating conditions: External conditions and actuation Distribution / Configuration in the blade For each category, one or several parameters are selected to be included in the parametric study. The parameters are selected attending to the relevance in the design of the device and the importance in the aerodynamic influence in terms of the rotor and airfoil performance. Table 1 shows the parameters selected in the present study for VGs and LE and TE flaps. Category VGs (triangular concept) Flaps Dimensions Height, h [1] Length Length, l [1] Shape [2] Angle wrt flow, γ [1] Position, x/c [1] Internal distance, d [1] External distance, D [1] Angle Actuation conditions β (sinusoidal oscillation) [3] External conditions Operating α Re M Δβ (sinusoidal oscillation) [3] k (sinusoidal oscillation) [3] β ini (ramp) [3] β range (ramp) [3] β (ramp) [3] α Distribution Location (mid-point) Location (mid-point) Radial extension Table 1: Selection of study parameters Re M Radial extension [1] Dimensions are defined for the triangular concept of VG. In that case, the shape and dimensions are described in Figure 1, for the downwash lay-out. Page 8 of 141 WP no.: 3.3

9 x/c γ Figure 1: Selection of study parameters for the downwash lay-out of VGs [2] It is possible to evaluate the effect of the flap shape. Traditional solutions are rigid straight flaps or deformable flaps based on different curved shapes. Apart from the straight rigid flap, the same approach of a curved flap used in tasks 3.1 and 3.2 is evaluated. The proposed formulation and a FORTRAN code are included in Appendix 2. [3] These parameters are used to evaluate the influence of the operating conditions in the unsteady aerodynamic response of the control device. β, Δβ, and k are respectively the mean flap angle, the amplitude of the flap angle and the reduced frequency of a sinusoidal oscillation of the flap. β ini, β range, and β are the initial flap angle, the range of the flap motion and the flap angle change rate during the ramp. Page 9 of 141 WP no.: 3.3

10 4. Test cases: Matrices Considering the different parameters proposed in Table 2, the final test matrix for the parametric study can be obtained combining different values. In order to limit the work, the analysis is carried out step by step where decision in the next step are based on the results of the previous steps e.g. decisions on 3D rotor simulations are based on analysis from 2D cases. The study is based on the application of the flap devices for load reduction and the VGs for increased power production. TE flaps are better suited for load reduction in attached flow conditions with respect to the LE flap because the potential variation of Cl is higher with the same flap deployment. However, LE flaps are expected to have a good potential at operating conditions with unsteady separated flow / dynamic stall because LE separation is influenced by the specific flow conditions at the LE region. The airfoils proposed for the 2D study are selected consistently with previous tasks 3.1 and 3.2. The list of airfoils and flow control devices are: FFA_W3_333_VG: FFA W3 333 (INNWIND) + VGs, Chord =6.06 m (r/r=0.35) DU_331_VG: DU-331 (AVATAR) + VGs, Chord =5.84 m (r/r=0.35) FFA_W3_248_LEFlap: FFA W3 248 (INNWIND) + LE flap, Chord =4.43 m (r/r=0.6) DU_240_LEFlap: DU 240 (AVATAR) + LE flap, Chord =4.36 m (r/r=0.6) FFA_W3_241_TEFlap: FFA W3 241 (INNWIND) + TE flap, Chord =3.31 m (r/r=0.75) DU_240_TEFlap: DU 240 (AVATAR) + TE flap, Chord =3.45 m (r/r=0.75) In the 3D part, the AVATAR and INNWIND RWTs are used, including different configurations of VGs and TE flaps in the blades. 4.1 TE flap study Study of flap length and shape Based on the results of Cl, Cd, Cm and Cl/Cd, the study aims at the selection of the flap length and shape of the TE flap. The objective would be to have a good compromise of Cl/Cd, with a significant potential of variation of the loading using the flap. In order to evaluate the influence of the flap length and shape in the aerodynamic response of the flap in terms of force and moment coefficients, different static cases are defined (Table 3). The static polars of different combinations of shape and length of the flap are obtained with the flap fixed in different positions. The conditions of the simulations for the reference airfoils are defined in Table 2. FFA_W3_241_TEFlap Wind speed = m/s Angles of attack = -16:2:30º Flap angle = -10, 0, 10º Chord =3.31 m DU_240_TEFlap Wind speed = m/s Angles of attack = -16:2:30º Flap angle = -10, 0, 10º Chord =3.45 m Page 10 of 141 WP no.: 3.3

11 M=0.2 Re=15M Fully turbulent M=0.22 Re=18M Fully turbulent Table 2: Airfoils and specifications for the analysis of flap length and shape Flap shape TE flap length (% x/c) No flap Straight X X X X X X X X Curv 1 X X X X X X X Table 3: Matrix of cases for the analysis of flap length and shape (highlighted: CFD control points) After the study, an optimum shape and length of the TE flap are selected according to the potential effect of the flap in the aerodynamic coefficients Cl, Cd, Cl/Cd and Cm. The selection will be used in the rest of the study Study of flap angle Based on the variation of Cl, Cd, Cm and Cl/Cd, the study aims at the selection of the maximum flap angle (positive and negative). The objective would be to have a good compromise of Cl/Cd, with a significant potential of variation of the loads using the flap. The configuration of TE flap selected in the previous section is used (flap length and shape), and the specifications are included in Table 4. In order to evaluate the effect of the flap angle, different static cases are defined (Table 5). FFA_W3_241_TEFlap Wind speed = m/s Angles of attack = -16:2:30º Flap length = 10%c Flap shape = Curv 1 Chord =3.31 m M=0.2 Re=15.57M Fully turbulent DU_240_TEFlap Wind speed = m/s Angles of attack = -16:2:30º Flap length = 10%c Flap shape = Curv 1 Chord =3.45 m M=0.22 Re=18.59M Fully turbulent Table 4: Airfoils and specifications for the analysis of flap angle TE flap angle (º) X X X X X X X X Table 5: Matrix of cases for the analysis of flap angle (highlighted: CFD control points) After the study, the influence of positive and negative flap angles is characterized according to the potential effect on Cl, Cd, Cl/Cd and Cm. The selection of limits should be based mainly on load reduction purposes Study of unsteady aerodynamic response In the previous sections, the static airfoil configuration with a TE flap is finished, in terms of dimensions and geometry of the flap and actuation conditions in terms of flap angles (negative and positive). In this section, the unsteady aerodynamic response to the flap angle is evaluated in order to characterize the effect of the flap under unsteady conditions. Several unsteady cases for TE flap oscillations and ramps have been defined, in different angle of attack conditions. Page 11 of 141 WP no.: 3.3

12 The configuration of flap selected in the previous section is used (flap length and shape) and all the information required is included in Table 6. Table 7 shows the matrix of cases with oscillations of the flap angle in different unsteady conditions at different angles of attack. In addition, Table 8 presents the matrix of cases with different flap ramps at different unsteady conditions, at different values of the angle of attack. The value of the ramp rate has also been normalized by wind speed and semi-chord. The highlighted cases are recommended for CFD simulations. FFA_W3_241_TEFlap Wind speed = m/s Flap length = 10%c Flap shape = Curv 1 Chord =3.31 m M=0.2 Re=15.57M Fully turbulent DU_240_TEFlap Wind speed = m/s Flap length = 10%c Flap shape = Curv 1 Chord =3.45 m M=0.22 Re=18.59M Fully turbulent Table 6: Airfoils and specifications for the analysis of unsteady flap angles α (º) 0 k β (º) Δβ (º) X X 0.05 X X X X 0.1 X X X X 0.15 X X X X X X 0.05 X X X X X X 0.1 X X X X X X 0.15 X X X X X X 0.2 X X X X 0.05 X X X X 0.1 X X X X 0.15 X X X X X X X 0.1 X X X 0.15 X X X Table 7: Matrix of cases for the analysis of unsteady flap oscillations (highlighted: CFD control points) α (º) 0 (2*V/c)* (º/s) βini (º) βrange (º) X X 0.2 X X X X 0.4 X X X X Page 12 of 141 WP no.: 3.3

13 X X X X 0.1 X X 0.2 X X X X X X 0.4 X X X X X X 0.6 X X X X X X 1.0 X X 0.2 X X X X 0.4 X X X X 0.6 X X X X 0.2 X X X 0.4 X X X 0.6 X X X Table 8: Matrix of cases for the analysis of unsteady flap ramps (highlighted: CFD control points) After the study, with the comparison of the unsteady cycles of Cl, Cd and Cm, some conclusions about the effect of the unsteady conditions on the TE flap aerodynamic response are obtained. This could be useful for practical design applications to evaluate the effect of the unsteady motion of the flap in the aerodynamic force and moment coefficients Study of flap influence in the 3D rotor In the previous sections, a complete 2D analysis of the two airfoils with TE flap has been finished. The selected flap solution is now used for a 3D study of the flap influence in the wind turbine performance. Several static flap cases in the INNWIND and AVATAR RWTs are defined using combinations of the flap location (mid-point) in the blade and the flap extension in the radial direction. The objective of the study is the evaluation of the effect of the TE flap distribution on the blade in wind turbine loads and power. The blade root flapwise and edgewise moments, torsion and wind turbine power and thrust are obtained for the analysis of different flap configurations. The different specifications for the AVATAR and INNWIND rotor are included in Table 9. The different flap positions (mid-point) in the blade and the extensions of the flap region have been combined in the test matrix of Table 10. INNWIND Wind speed = m/s Ω = 9.6 rpm No tower, no precone, no tilt Pitch=1.8º FFA_W3_241_TEFlap: Flap angle = -10, 0, 10º Flap length = 10%c Flap shape = Curv 1 Fully turbulent AVATAR Wind speed = m/s Ω = 9.6 rpm No tower, no precone, no tilt Pitch=2.06º DU_240_TEFlap: Flap angle = -10, 0, 10º Flap length = 10%c Flap shape = Curv 1 Fully turbulent Table 9: Wind turbine and specifications for the analysis of the TE flap configuration on the blade Page 13 of 141 WP no.: 3.3

14 β Flap mid point in blade radius (y/r %) X X X X X X 10 X X X X X 15 X X X X 20 X X X 10 5 X X X X X X 10 X X X X X 15 X X X X 20 X X X Table 10: Matrix of cases for the analysis of the TE flap configuration in the blade (highlighted: CFD control points) Rad extension (y/r %) The different configurations are evaluated in terms of the effect in the loads and power. The effect of the different TE flap configurations in power and thrust is analyzed, for a suitable evaluation of the potential for practical applications. 4.2 LE flap study Study of unsteady aerodynamic response In this section, the influence of the LE flap in the unsteady aerodynamics of the 2 reference airfoils is investigated. The potential of the LE flap for dynamic stall control is evaluated. In this section, the unsteady aerodynamic response of a pitching airfoil with a static LE flap of different lengths and angles is evaluated. The different conditions for the simulations are included in Table 11, and several unsteady cases for angle of attack oscillations and ramps have been defined in Table 12, using different LE flap configurations (only for the curved flap). FFA_W3_248_LEFlap DU_240_LEFlap Wind speed = m/s Flap length = 10, 20, 30 (% x/c) Flap angle = 5, 10º Flap shape = Curv 1 Chord =4.43 m Wind speed = m/s Flap length = 10, 20, 30 (% x/c) Flap angle = 5, 10º Flap shape = Curv 1 Chord =4.36 m M=0.16 Re=16.57M Fully turbulent M=0.18 Re=18.59M Fully turbulent Table 11: Airfoils and specifications for the analysis of unsteady flap angles α (º) 15 Δα (º) Flap length (% x/c) Flap angle (º) X X X 0.1 X X X X X X k 0.05 X X X X X X X 0.1 X X X X X X X 0.15 X X X X X X X X X X 0.1 X X X Page 14 of 141 WP no.: 3.3

15 X X X 0.05 X X X X X X X 0.1 X X X X X X X 0.15 X X X X X X X Table 12: Matrix of cases for the analysis of unsteady flap oscillations (highlighted: CFD control points) After the study, with the comparison of the unsteady cycles of Cl and Cd, some conclusions about the effect of the LE flap in the aerodynamic response are obtained. The evaluation of the effect of the flap in the maximum and peak to peak loads is useful for control purposes. 4.3 VGs study Study of VGs height and position in the chord direction Based on the results of Cl, Cd, Cm and Cl/Cd, the study aims at the selection of the height and position in the chord direction of the VGs. The value of length, angle with respect to flow and internal and external distances are fixed for this investigation and varied in the next step. The objective would be to have a good compromise of Cl/Cd, with a useful modification of the aerodynamic characteristics with respect to the airfoil without VGs. In order to evaluate the influence of the VGs height and location in the aerodynamic response in terms of force and moment coefficients, different cases are defined with a quasi-2d extension of the airfoil. The proposed values of height have been extracted by ECN based on a study of the boundary layer thickness. The static polars of different combinations of the parameters will be obtained and compared with the case without VGs. Table 13 shows the specifications for the analysis, and Table 14 present the matrix of cases. FFA_W3_333_VG DU_331_VG Re = x10^6 Re = x10^6 M = M = Chord =6.06 m Chord =5.84 m Angles of attack = 0,4,8,10,12,14,16,18,20º Angles of attack = 0,4,8,10,12,14,16,18,20º l=3*h l=3*h Angle wrt flow = 20º Angle wrt flow = 20º d=2*h d=2*h D= 5*h D= 5*h CFD simulation: Fully turbulent CFD simulation: Fully turbulent Table 13: Airfoils and specifications for the analysis of VGs dimensions and orientation x/c (%) FFA W3 333 DU 331 No VG X X h [mm] 12 X 12 X X 18 X 36 X 36 X 60 X 60 X X 18 X 36 X 36 X Page 15 of 141 WP no.: 3.3

16 40 60 X 60 X 18 X 18 X 36 X 36 X 60 X 60 X 90 X 90 X Table 14: Matrix of cases for the analysis of VGs dimensions and orientation length (highlighted: CFD control points) At the end of the parametric study, it is important to understand the relation between the VGs characteristics and the aerodynamic improvement (or influence in the aerodynamic response of the airfoil). The ratio between the VGs height and the physical boundary layer thickness can be strongly related to the modifications of the polars. Consequently, it would be very useful to quantify the boundary layer thickness with high fidelity codes (such as CFD), although it requires previous simulations. After the study, the height and location are selected for the VGs according to the potential effect in the aerodynamic coefficients Cl, Cd, Cl/Cd and Cm. The selection will be used in the rest of the study Study of VGs length, angle wrt flow and internal and external distance Based on the results of Cl, Cd, Cm and Cl/Cd, the study aims at the selection of the length, angle with respect to flow and the internal and external distance of an array of VGs. The objective would be to have a good compromise of Cl/Cd, with a useful modification of the aerodynamic characteristics with respect to the airfoil without VGs. The conditions for the simulations are included in Table 15. Starting from the height and location in the chord selected in the previous step, the rest of the parameters are varied in Table 16 (once per time, keeping the other parameter constant and equal to those of the case from previous step). This leads to 8 more cases. It would be interesting to have CFD calculations for all of these 8 cases. FFA_W3_333_VG DU_331_VG Wind speed = m/s Wind speed = m/s Angles of attack = 0:2:20º Angles of attack = 0:2:20º h=30mm h= 30mm Position (x/c) = 40%c Position (x/c) = 40%c Chord =6.06 m Chord =5.84 m Re = 13.95x10^6 Re = 15.24x10^6 M = M = Fully turbulent Fully turbulent Table 15: Airfoils and specifications for the analysis of VGs external and internal distance Cases l 2*h 4*h 3*h 3*h 3*h 3*h 3*h 3*h Angle wrt 20º 20º 15º 25º 20º 20º 20º 20º flow d 2*h 2*h 2*h 2*h 1.5*h 2.5*h 2*h 2*h D 5*h 5*h 5*h 5*h 5*h 5*h 4.5*h 5.5*h Page 16 of 141 WP no.: 3.3

17 Table 16: Matrix of cases for the analysis of the VGs external and internal distance (highlighted: CFD control points) After the study, the sensitivity of the parameters is analyzed according to the effect in the aerodynamic coefficients Study of VGs influence in the 3D rotor In the previous sections, a complete 2D analysis of the two airfoils with VGs has been finished. The selected VGs configuration is now used for a 3D study of the influence in the wind turbine performance. Several cases of the INNWIND and AVATAR RWTs are defined for combinations of the VGs location (mid-point) in the blade and the extension in the radial direction. The specifications of the simulations are mentioned in Table 17. The objective of the study is the evaluation of the effect of the VGs distribution on the blade in wind turbine loads and power. The blade root flapwise and edgewise moments, torsion and wind turbine power are used for the analysis of different VGs configurations. The matrix of cases is shown in Table 18. INNWIND AVATAR Wind speed = m/s Wind speed = m/s Ω = 9.6 rpm Ω = 9.6 rpm No tower, no precone, no tilt No tower, no precone, no tilt Pitch=1.8º Pitch=2.06º VGs info: VGs info: d=60mm d=60mm D=150mm D=150mm h=30mm h=30mm l=90mm l=90mm Angle wrt flow =20º Angle wrt flow =20º Position=40%c Position=40%c Fully turbulent Fully turbulent Table 17: Wind turbine and specifications for the analysis of the VGs configuration on the blade VGs mid point in blade radius (y/r %) X X X Radial extension (y/r 10 X X X %) 15 X X X Table 18: Matrix of cases for the analysis of the VGs configuration in the blade (highlighted: CFD control points) The different configurations are evaluated in terms of the effect in the thrust and power. Page 17 of 141 WP no.: 3.3

18 5. Details about codes and simulations 5.1 WMB (CENER) Code description The main features of WMB were described in Ref. [1]. WMB (Wind Multi-Block, Ref. [2]) is a CFD code developed at Liverpool University and validated by CENER and the University of Liverpool for wind turbine aerodynamics analysis (2D and 3D). It is capable of solving the compressible Unsteady Reynolds Averaged Navier-Stokes (URANS) flow equations on multi-block structured grids using a cell-centred finite-volume method for the spatial discretization. CENER has also used HMB2 for the study of the effect of unsteady TE flaps (see reference of HMB by UoG in section 5.9) Simulation details For the 2D cases WMB CFD code was used for the simulation of the static flap cases. CENER 2D calculations were performed with an ANSYS ICEMCFD mesh with a total number of nodes of and 434 over the airfoil surface. The domain size is 25c both upstream and downstream. Fully turbulent k-ω baseline calculations were performed. These calculations are steady state calculations. For the 3D cases WMB CFD code was used for the simulation of the static flap cases (18 in total proposed flap cases at rated and above rated speeds). Just a third of the rotor was meshed (a single blade) and computed, assuming the periodicity in space and time. For that, fully turbulent steady computations in a non-inertial frame of reference with the Menter s k-ω SST (Ref. [3]) turbulence model were computed. All the computations converged in less than 20 hours (114 cores). The meshes were generated using ICEM CFD v16.2 with non-dimensional distance of the first node from the blade surface equal to c. 310 nodes covered the blade chord and depending on the case, the span-wise cell distribution varied from 334 to 381 cells. The total size of the meshes ranged between 16.1 and 17.9 million cells. The computational boundaries were fixed at 3R towards inflow, 6R towards outflow and 4R towards the far-field. 5.2 AdaptFoil2D (CENER) Code description The main features of AdaptFoil2D were described in deliverable 1, Ref. [1], while some specific developments for deformable geometries were mentioned in deliverable 2, Ref. [4]. Page 18 of 141 WP no.: 3.3

19 Basically, AdaptFoil2D is a code developed in CENER for 2D airfoil aeroelastic modelling based on a potential solution. In attached flow, in comparison with the viscous-inviscid codes, the potential solution gives a larger increase of Cl with the angle of attack or flap angle, and no drag. The aerodynamic part is based on panel and vortex methods. The code was developed taking into account a good balance between accuracy and computational effort for aerodynamic modelling. Flow separation is considered in AdaptFoil2D using a double wake approach. The steady separation location is assumed to be provided in a tabulated form, from experiments, CFD or viscous-inviscid calculations. Once the steady separation location is known, the unsteady separation is calculated by the panel code taking advantage of some concepts used in engineering models. The unsteady motion of the separation location in the airfoil surface is obtained by a first-order lag to the steady value, as in the Beddoes-Leishman model (Ref. [5]). In addition, the development of the dynamic stall vortex and the effect in the aerodynamic response can be calculated by the panel code. AdaptFoil2D applies the critical condition in the leading edge pressure mentioned by Evans and Mort (Ref. [6]) in order to predict the dynamic stall onset. AdaptFoil2D is suitable for the modelling of active or passive deformations of the airfoil geometry. For active deformations (for example flaps), the code is able to interpolate between different geometries and the corresponding flow separation information Simulation details AdaptFoil2D was used to obtain two-dimensional static polars of TE flap cases (study of length, shape and angle) and for two-dimensional unsteady simulations (LE and TE flaps unsteady aerodynamics). Concretely, the tip and mid sections of the INNWIND and the AVATAR rotor blades have been modelled including TE and LE flaps. Using the original blunt TE geometry of the corresponding airfoils, a sharp profile is created with XFoil and divided into 400 panels. The simulation time step was approximately (FFA_W3_248), (FFA_W3_241) and (DU240) seconds. For steady cases with flow separation, at least 2 seconds are needed for convergence, and the results are averaged from the 2 nd to the 10 th second. The unsteady cases including conditions of separated flow, present similar time to converge and the final results have been averaged from the 2 nd to the 15 th seconds. 5.3 FAST (CENER) Code description The main features of FAST with some specific developments for deformable geometries were mentioned in deliverable 2, Ref. [4]. FAST is an aeroelastic simulation code developed by NREL which can model the dynamic response of 2 or 3 bladed horizontal axis wind turbines. It is coupled with the aerodynamic code Page 19 of 141 WP no.: 3.3

20 AeroDyn. In terms of computational time, the engineering codes are very fast in comparison to higher level codes as vortex methods or CFD. In the calculation of the aerodynamic loads on the blades, AeroDyn uses different models and corrections for induced velocities on the blades, tip and hub effects, tower, etc. The unsteady aerodynamics in the blade sections are included using unsteady models based on the Beddoes- Leishman model. In our version of the code, apart from the models included by NREL, CENER has included DYSTOOL, a specific dynamic stall tool for an accurate representation of the unsteady aerodynamics of different airfoils in different operating conditions. AeroDyn has been modified for the modelling of the effect of dynamic flaps in the blades. First, the unsteady aerodynamic model has been extended in order to consider the unsteady aerodynamic response of the airfoil to a change of the geometry, for example the angle of a flap. In addition, a model based on the Kutta-Joukowski theorem and the Biot-Savart law has been implemented in order to include the 3D effects of the distributed devices in the blade Simulation details FAST was used for the three-dimensional simulations. The sections of the blade were taking into account the required widths of the flap. The static polars for the different flap angles have been included as an input for these calculations. The time step used for these cases is 0.02 s, and the time required to reach convergence is approximately 1 second. All the cases of the AVATAR and the INNWIND rotors have been simulated and the final results presented have been obtained for the 1000 th time step (from the point of view of the blade sections, the conditions are steady due to the definition of the cases). 5.4 MaPFlow (NTUA) Code description MaPFlow (Papadakis G. and Voutsinas S.G. 2014) is an in-house multi-block MPI enabled compressible URANS solver equipped with preconditioning in regions of low Mach flow. Unstructured meshes generated by the ICEMCFD ANSYS software are used. The models implemented in the code are: The Spalart-Allmaras (SA) and k-ω SST eddy viscosity turbulence models The correlation γ-reθ model of Menter, the Granville/Schlichting transition method and the en transition models. The BAY and the phenomenological VG models Description of the main code characteristics and the implemented models can be found in deliverables 3.1 and 3.2 (Refs. [1] and [4]) Simulation details Page 20 of 141 WP no.: 3.3

21 MaPFlow was used in the TE flap simulations of the DU-240 airfoil. O-type meshes of approximately nodes (520x200) were utilized, consisted of quadrilateral cells. The far field boundary of the computational domain was set 50 chords far from the airfoil. Flow was considered as fully turbulent. For the static flap simulations, a number of time steps were proven to be sufficient for full convergence of the momentum equation residuals. For the oscillating flap simulations, a number of 720 time steps per cycle were used. For the highest AoA (15 o ) convergence was achieved after 10 cycles, whereas 5-6 cycles were sufficient for the lower AoAs. A mean dual step error had of 10-9 was necessary in order to achieve convergence. For the ramp cases, a number of 8000 time steps were used for the discretization of the ramp duration. This resulted in a dimensional time step of s, s and s for the reduced ramp rates of 0.1, 0.4 and 1 respectively. A number of 8000 steady time steps were used to simulate the steady state before and after the ramp duration. For the VG simulations two different models were used: BAY model and phenomenological VG model. The first is by means of the BAY model (Bender, E.E., Anderson, B.H., Yagle, 1999) which assumes that the presence of a zero thickness vane VG can be represented as a source term in the momentum and energy equations. The source term simulates the lift force introduced by the VG in the flowfield. This term aligns the flow with the VG direction. The model is applied in its jbay variation (Jirasek, 2005) in which the VG is replaced by a surface with zero thickness. It should be noted that BAY model works only in 3-dimensional simulations. The phenomenological VG model utilized 2d grids of approximately cells (500x200) consisted of quadrilateral cells. The farfield boundary of the computational domain was set 50 chords far from the airfoil. Flow was considered fully turbulent. Steady simulations were carried out up to 12 degrees AoA. For higher angles, the simulations were unsteady with a dimensionless timestep of For all these VG simulations, timesteps were sufficient for the code to converge sufficiently. 5.5 Foil2w (NTUA) Code description Foil2w (Riziotis V.A. and Voutsinas S.G. 2008) is an in-house viscous-inviscid interaction code. The e N transition model is implemented. The potential flow part is simulated by singularity distributions along the airfoil geometry and the wake. The wake is represented by vortex particles which are allowed to freely move with the local flow velocity. The viscous flow solution is obtained by solving the unsteady integral boundary layer equations defined by Drela. Description of the code characteristics is included in Deliverables 3.1 and 3.2 (Refs. [1] and [4]) Simulation details Foil2w was used in the TE flap simulations of both DU-240 and FFA-W3-241 airfoils. The geometry of each airfoil was discretized using 140 panels. Fully turbulent flow conditions were simulated by setting fixed transition at a distance of 1% chord on the suction side of the airfoil. In the static TE flap simulations the non-dimensional time step was set to For the Page 21 of 141 WP no.: 3.3

22 oscillating TE flap simulations a number of 500 or 1000 time steps per cycle were used, and 4-5 cycles were sufficient to obtain a periodic solution. For the ramp simulations, an initial time step, dt s, was used to simulate the state before and after the ramp. At the start of ramp, the time step was reduced to 0.1dt 0, and then followed a successive linear increase and decrease, reaching again the value of 0.1 dt 0 at the end of the ramp. 5.6 GENUVP (NTUA) Code description GENUVP (Voutsinas S.G. 2006) is an unsteady potential flow solver in which the effect of solid boundaries is represented by means of surface source and/or dipole distributions while the wakes are modelled by means of freely moving vortex blobs. Viscous corrections can be introduced using C L -C D data. Description of the code characteristics is included in Deliverables 3.1 and 3.2 (Refs. [1] and [4]) Simulation details GENUVP was used in the 3D static TE flap simulations of both AVATAR and INNWIND rotor blades. The aerodynamic part of the AVATAR blade geometry (starting from the 30% of its span) was described using a surface grid of 51x39 points (39 radial stations with 51 points per airfoil section). In the case of INNWIND rotor the root section was taken into account and a surface grid of 59x51 points was used. A full rotation was simulated using 90 steps and viscous corrections were taken into account for both rotors. For this purpose, 2D C L -C D polars calculated by Foil2w (section 5.5.2, static TE simulations) were used for the TE flap simulations of both DU-240 and FFA-W3-241 airfoils. 5.7 FLOWer (USTUTT) Code description At the Institute of Aerodynamics and Gas Dynamics (IAG), University of Stuttgart, a process chain for the simulation of wind turbines has been developed in the last decade (Ref. [7]). The basic part constitutes the CFD code FLOWer, which was originally developed by the German Aerospace Center (DLR) (Ref. [8]). FLOWer is a compressible code that solves the 3D, Reynolds-averaged Navier-Stokes equations in integral form. The numerical scheme is based on a finite-volume formulation for block-structured grids. To determine the convective fluxes a second order central discretisation with artificial damping is used, also called the Jameson-Schmidt-Turkel (JST) method. Time accurate simulations make use of the Dual-time-stepping method as implicit scheme. To close the Navier-Stokes equation system several state of the art turbulence models can be applied as for example the SST model by Menter used in this study. There are two main code features for the simulation of wind turbines. The ROT module for moving and rotating reference frames in combination with the CHIMERA technique for overlapping meshes allow body motions relative Page 22 of 141 WP no.: 3.3

23 to each other in time accurate simulations. FLOWer is optimized for parallel computing and uses Message-Passing Interface (MPI). Trailing edge flaps are realised using grid deformation based on radial basis functions (Ref. [9]). For this the flap part of the airfoil/blade is deflected or deformed while the rest is kept rigid. A detailed description of the approach can be found in deliverable 3.2 (Ref. [4]) Simulation details For the 2D airfoil simulations, grids were generated with an in-house script for the commercial grid generator Pointwise. In the background of the results of deliverable 3.2, 418 nodes were used on the airfoil surface. In wall normal direction, 64 boundary layer nodes were chosen with a growth rate of 1.1 for a y+-value corresponding to 0.5. Another 101 nodes extent the grid to the far field boundary. For the polar calculations the farfield boundary was chosen to be ~ 45c, no corrections were applied. The unsteady simulations of the oscillating flaps and flap ramps were simulated with ~90c, also no corrections have been applied. Since the trailing edge flap for arbituary deflections is realized within the CFD code only two grids, one for each airfoil, had to be generated. The simulations of the airfoil polars have been performed steady and if needed unsteady to achieve a converged solution. In the unsteady cases the temporal resolution was chosen to be 100 steps per convective time unit with 30 to 50 inner iterations in the Dual-Time-Stepping scheme. For the unsteady aerodynamic response study with moving flap the temporal resolution was chosen in the same scale except for the cases with very low reduced frequency of In these cases, the time step had to be increased for reasonable computation time. But since this frequency was only computed at an angle of attack of 5, no major influence of separation is expected, for which the small time step is mainly needed. In all cases at least 1800 time steps where performed per oscillation period or 500 per ramp. 3D rotor simulations with flap are performed in the same manner as in deliverable 3.2, where the procedure is described in detail. Due to the radial refinements needed at the flap edges, four different blade grids had to be generated per turbine. These imply the specified flap location and extension. Grid resolutions and spacings orient on previous results (273 chord nodes, at least 157 along blade span). All 3D computations are initiated steady state with 8000 iterations on 2nd and 1st Multi-Grid level respectively. Afterwards an unsteady restart was conducted with a timestep corresponding to 2 azimuth. In case of the INNWIND turbine, 1 revolution was simulated and the second half was averaged for evaluation. For the AVATAR turbine, 1.5 revolutions had to be conducted and again the last half revolution was averaged. 5.8 OpenFOAM (TU Delft) Code description Page 23 of 141 WP no.: 3.3

24 OpenFOAM is an open source CFD software distributed under the General Public Licence (GPL), thereby giving users the freedom to modify the source code and develop their own additional pieces of code. It is a segregated finite volume code able to solve compressible and incompressible flows on either structured or unstructured grids. For the current work, the steady, incompressible Reynolds averaged Navier Stokes (RANS) equations are solved using the SIMPLE algorithm and the governing equations are solved on structured grids using second order upwind discretization schemes for the convective terms. The linear systems arising from the equation discretization are solved using the preconditioned (bi-) conjugate gradient method with diagonal incomplete Cholesky and diagonal incomplete LU preconditioners for the symmetric and asymmetric systems respectively. Closure for the RANS equations is provided by Menters two-equation k ω shear stress transport model (k ω SST), which is based on the Boussinesq approximation. The effect of VGs is accounted for by using the BAY model. A source term that consists of an approximation for the force introduced by the VG on the flow field is locally added to the momentum equation. This force causes the flow to align with the VG direction. The cells where this source term is applied are chosen such as to correspond to the physical location of the VG Simulation details OpenFOAM with the BAY model was used to perform 3D simulations of the DU331 airfoil (extended in the 3rd dimension) equipped with counterrotating VGs. For the simulations a small slice containing half a VG pair is considered, using symmetry boundary conditions to take into account the influence of neighboring VGs. An O-grid mesh is constructed around the blunt TE airfoil geometry with the farfield boundary located 75 chord lengths from the airfoils LE. The numerical mesh consists of 2.3 million cells, with 508 x 26 x 185 points in chordwise, normal and spanwise direction respectively. Local mesh refinements are applied near the wall (yielding y+ < 5 on the airfoil surface) and in the neighborhood of the VG. 5.9 HMB (University of Glasgow) Code description The Helicopter Multi-Block (HMB) solver has been under constant development for the past 14 years at the Universities of Glasgow and earlier at Glasgow and Liverpool. The code was also used as the basis for the WMB solver that is also used in this project by CENER and Glasgow University (Ref. [10] and section 5.1). HMB is capable of solving Euler and Navier Stokes equations on any type of mesh (structured, unstructured, Chimera, or sliding). It is a compressible flow solver with options for low-mach (Ref. [11]) simulation and all-mach numerical schemes (Ref. [12]). The solver is second order accurate in time and can be 1 st, 3 rd or 4 th order accurate in space. Page 24 of 141 WP no.: 3.3

25 5.9.2 Simulation details For the 2D cases HMB was used for static and dynamic flap configurations. CFD grids were generated using the ICEM-HEXA tool of ANSYS. The size of the 2D grids varied from 75 to 100 thousands cells and the far-field boundary was placed at 40 aerofoil chords. For AVATAR mainly CH-type (C topology at the leading edge and H topology at the trailing edge) multiblock structured grids were used for aerofoils and computations in steady and unsteady fashion were performed. For the 3D cases HMB was used for rigid and aeroelastic blade simulations. Static and dynamic flaps were simulated mainly for the AVATAR blade. A third of the rotor was meshed (a single blade) and computed, assuming the periodicity in space and, for steady computations, in time too. For that, fully turbulent steady computations in a non-inertial frame of reference with the Menter s k-ω SST (Ref. [3]) turbulence model were computed. All the computations converged in less than 96 hours on 64 cores. The meshes were generated using ICEM CFD v16.2 with non-dimensional distance of the first node from the blade surface equal to c max (where c max is the maximum chord of the blade) and 251 to 501 nodes around the blade chord and in the span-wise direction depending on the case. The total size of the meshes ranged between 8.5 to 30 million cells. The computational boundaries were fixed at 3R towards inflow, 6R towards outflow and 4R towards the far-field MIRAS (DTU) Code description MIRAS (Method for Interactive Rotor Aerodynamic Simulations) is a 3D viscous-inviscid interactive solver for rotor simulations (Refs. [13] and [14]). The solver predicts the aerodynamic behaviour of wind turbine wakes and blades for steady and unsteady conditions. A newly developed version of MIRAS is used in the present WP3, its novelty is based on the use of a free-wake/particle-mesh hybrid method in order to accurately simulate the wake downstream the wind turbine and reduce the required CPU time. Figure 2 and Figure 3 depict MIRAS simulations using the hybrid method. The proposed wake model uses a hybrid approach, where the near wake is simulated by using vortex filaments that carry the vorticity shed by the trailing edge of the blades, while further downstream the wake is modelled with a particle-mesh approach. In the particle-mesh method, the vorticity carried out by the filaments is interpolated into a Cartesian mesh, where the Poisson equation is solved using FFTs. The computed velocity field is employed to update the filament marker locations and re-calculate the wake induction in the blade cell centres. Page 25 of 141 WP no.: 3.3

26 Figure 2: MIRAS hybrid wake simulations: vorticity isocontours Figure 3: MIRAS hybrid wake simulations: mesh velocity The MIRAS code consists of inviscid and viscous parts. The inviscid part is a 3D panel method using a surface distribution of quadrilateral sources and doublets. The inviscid part is coupled to the viscous part through the integral viscous boundary layer solver, Q3UIC (Ref. [15]). Viscous and rotational effects inside the boundary layer are taken into account via the transpiration velocity concept, which is applied using a strip theory approach with the cross sectional angle of attack as coupling parameter. The transpiration velocity is obtained from the solution of the integral boundary layer equations, which in the present version of the code is externally calculated by using the viscous-inviscid solver Q3UIC. The required input data for MIRAS simulations are: blade geometry, incoming wind speed, tip speed ratio, type of transition (free or forced transition) and the transition location in case of forced transition. In the case of flap simulations, the flap geometry as well as the prescribed motion can be set in MIRAS input file Simulation details A surface mesh consisting of 40 span-wise cells and 150 chord-wise cells has been employed in the present computations, with 32 wake revolutions simulated with an azimuthal discretization of 5.4 degrees. The boundary layer is tripped at 5% of the chord in both the pressure and suction sides of the airfoil. All the cases of the AVATAR have been simulated and the final results presented have been obtained for the last time step Aero-Module (ECN) Code description The in-house developed ECN Aero-Module in Ref. [16] is used for the presented aerodynamic computations related to the investigation of the VG influence on the 3D rotor. The ECN Aero- Module includes both blade element momentum (BEM) as well as a lifting line free vortex wake formulation, allowing the same external input (e.g. wind, tower, airfoil data) to be used for both models. The BEM formulation is based on PHATAS (Ref. [17]), including state of the art Page 26 of 141 WP no.: 3.3

27 engineering extensions which have matured over decades of research in wind turbine rotor aerodynamics. The free vortex wake method is based on the AWSM code (Ref. [18]) Simulation details BEM and AWSM simulations were performed for both the INNWIND and AVATAR turbines at a wind speed of 11.4 m/s and a rotational speed of 9.6 rpm. For both of these turbines 2 cases have been considered, namely a) a clean case, using the reference polars taken from the AVATAR website (Ref. [19]), and b) a VG case, correcting the reference polars to account for the effect of VGs. A detailed description of the cases performed is given in Table 19. clean case VG case Radial extension (y/r %) X VGs mid point in blade radius (y/r %) 35 5 X 10 X 15 X Table 19: Investigation of the VGs influence on the 3D rotor: matrix of cases simulated by means of Aero-module The polar correction was based on the 2D CFD simulations performed by DTU for clean airfoil and airfoil implementing 30 mm high VGs installed at a chordwise location of 40%. For these conditions, Table 20 shows the lift and drag coefficient percent differences, C L% and C D%, due to VGs, as a function of the angle of attack, for the FFA W3 333 and DU 331 airfoils. C L% is defined as: C L% = C L,VG C L,clean C L,clean where C L,clean is the lift coefficient in clean case and C L,VG is the lift coefficient of the airfoil implementing VGs. C D% is defined as: C D% = C D,VG C D,clean C D,clean where C D,clean is the drag coefficient in clean case and C D,VG is the drag coefficient of the airfoil implementing VGs. Angle of attack FFA W3 ΔC L% ΔC D% DU 331 ΔC L% ΔC D% Table 20: Lift and drag coefficient percent differences due to VGs as a function of the angle of attack, for the FFA W3 333 and DU 331 airfoils. These difference were found for a VG height of 30 mm and a chordwise location of 40% Page 27 of 141 WP no.: 3.3

28 For the free vortex wake simulation with AWSM, the number of wake points was chosen to make sure that the wake length was developed over at least 3 rotor diameters downstream of the rotor plane. For both clean and VG cases, a 3D correction is applied based on the model of Snel (Ref. [20]) as modified in PHATAS, dependent on chord over radius and tip speed ratio. As such it is embedded in the overall code, applied during the calculation and restricted to the inboard region below 50 deg angle of attack. Page 28 of 141 WP no.: 3.3

29 6. Results of the parametric study In the next sections, the results obtained in the parametric study are included, through comparison and analysis of the different simulations carried out by the participants. The study is performed assuming the flaps are used mainly for load reduction. When possible, a cross-check between different codes has been performed in order to confirm the trends observed. Those trends are analyzed based on different metrics or KPI s. The definition of measurable KPI s is strongly recommended and applied in order to extract useful information from the parametric study. The following KPI s are proposed for the study (mean and / or statistics): 2D parametric study (airfoil): o Cl max o Cl at rated angle of attack o ΔCl/Δβ o ΔCd/Δβ o Cl/Cd o Δ(Cl/Cd) o Cm hinge o 3D parametric study (rotor): o Blade root flapwise moment o Blade root edgewise moment o Torque o Thrust o Aerodynamic power o 6.1 TE flap Study of flap length and shape The first step for the parametric study of the TE flap is the evaluation of the effect of flap length and shape. The length of the flap is varied between 0% (without flap) and 30% chord, while for the shape only two options have been evaluated, straight and curved flap. The polars obtained with different codes have been compared in order to cross-check the results and assure similar trends for the subsequent analysis. Figure 4, Figure 5 and Figure 6 show the results of Cl, Cd and Cm respectively, obtained with different codes for the DU240 airfoil without flap (straight lines) and with a 10% chord TE curved flap at an angle of +10º (dashed lines). For the Cl, in the case without flap, all codes agree in the linear region except AdaptFoil2D that is based on an inviscid calculation. For the case with flap, the agreement in the linear region is a bit worse. This could be due to the definition of the flap geometry. Increasing the angle of attack either with or without flap, the dispersion in the results increases significantly and important differences between the codes can be observed. The deviations observed between the codes seem to be similar with and without flap. This problem was also observed in WP2 in the Page 29 of 141 WP no.: 3.3

30 simulations of airfoil polars using different tools. However, attending to the differences between the results with and without flap, the effect of the flap is similar for the different codes. The Cl increases when the flap is deployed in the positive direction. In addition, the increment is less pronounced at high angles of attack once there is flow separation. Then, the differences between the codes modelling the clean airfoil may not affect the ability of modelling the effect of the flap. Figure 4: Cl vs α for the DU240 airfoil, without flap and with a 10%c TE flap at +10º The analysis is similar for the Cd. There are significant differences between the codes, but the effect of the flap is similar in all the cases. First, the Cd increases with the increase of the flap angle (positive flap). In addition, the sharp increase of Cd when the flow separation is progressing is observed at lower angles of attack when the flap is deployed in the positive side. Figure 5: Cd vs α for the DU240 airfoil, without flap and with a 10%c TE flap at +10º Page 30 of 141 WP no.: 3.3

31 For the Cm, in spite of the significant dispersion between the codes, the trends in the comparison between the cases with and without flap are captured by all the codes. The negative Cm is increased when the flap is deployed to a positive angle. In addition, the increment is lower at high angles of attack once there is flow separation. Figure 6: Cm vs α for the DU240 airfoil, without flap and with a 10%c TE flap at +10º Figure 7, Figure 8 and Figure 9 show the same results of Cl, Cd and Cm but for the FFA_W3_241. The analysis is the same as for the DU240. Figure 7: Cl vs α for the FFA_W3_241 airfoil, without flap and with a 10%c TE flap at +10º Page 31 of 141 WP no.: 3.3

32 Figure 8: Cd vs α for the FFA_W3_241 airfoil, without flap and with a 10%c TE flap at +10º Figure 9: Cm vs α for the FFA_W3_241 airfoil, without flap and with a 10%c TE flap at +10º Similar comparisons have been obtained for all the flap lengths, for a straight flap and for negative deployments. The conclusion is that, in spite of the differences observed in the polars obtained with different codes, all of them show a similar effect in the aerodynamic response when the polars without flap are compared with the polars with flap. Then, although the reference baseline could be different, the codes should be valid for the parametric study, to evaluate the effect of the variation of the each parameter. WP2 is focusing on differences in default conditions, but this task should focus purely on relative differences attending to the different configurations of flow devices. Different KPI s are evaluated to measure the effect of the flap and the parameters. Page 32 of 141 WP no.: 3.3

33 First, the maximum Cl obtained is studied depending on the flap length. Figure 10 and Figure 11 show the results for the DU240 and FFA_W3_241 with a curved TE flap deployed 10º in the positive direction. The maximum Cl increases with the length of the flap. However, the rate of this increment decreases progressively. The increment is bigger for the FFA_W3_241. This behaviour has also been observed in the straight flap. The result of FOIL2W without flap shows a slightly different trend with respect to the rest of the curves. Figure 10: Maximum Cl at different flap lengths for the DU240 airfoil with a curved TE flap at β=+10º Figure 11: Maximum Cl at different flap lengths for the FFA_W3_241 airfoil with a curved TE flap at β=+10º The trend of the angle of attack for max efficiency (Cl/Cd max) is similar for all codes. Figure 12 and Figure 13 show the angle of attack of maximum efficiency for the same TE flap angle at different flap lenghts. For both airfoils, the angle of attack of maximum efficiency decreases with the length of the flap. For the DU240, the rated angle of attack is 1º in the 75% blade section considered in the AVATAR rotor, far from the optimum of the airfoil without flap. With positive flap angle, it would be closer to the maximum efficiency but the contrary occur for negative angles. For the FFA_W3_241, the rated angle of attack is 8º for the 75% blade section in the INNWIND rotor, corresponding to the optimum for the airfoil without flap but not for the rest of configurations with flap. Figure 12: AoA of max efficiency at different flap lengths for the DU240 airfoil with a curved TE flap at β=+10º Figure 13: AoA of max efficiency at different flap lengths for the FFA_W3_241 airfoil with a curved TE flap at β=+10º Page 33 of 141 WP no.: 3.3

34 In Figure 14 and Figure 15, the maximum efficiency is presented depending on the flap length for the DU240 and FFA_W3_241 airfoils respectively. For both airfoils, it seems that the maximum efficiency can be observed between the airfoil without flap and flap lengths up to 10-15%c, although for the DU240, the maximum values are closer for the airfoil without flap while for the FFA_W3_241 are closer for flap lengths around 5-10% chord (except for WMB). In both cases, the efficiency significantly decreases with lengths above 15% chord. Figure 14: Max efficiency at different flap lengths for the DU240 airfoil with a curved TE flap at β=+10º Figure 15: Max efficiency at different flap lengths for the FFA_W3_241 airfoil with a curved TE flap at β=+10º It is also interesting to evaluate the increase of Cl per unit flap length (in % chord) for the different flap lengths at rated angle of attack with respect to the case without flap. This characterizes the effectiveness of the flap deflection for load control. The trends can be observed in Figure 16 and Figure 17 for the reference airfoils at rated angle of attack (α=1.1º for the DU240 and α=8.1º for the FFA_W3_241) with flap angle 10º. For both airfoils, in spite of the continuous Cl increase, the increment per unit length is lower for longer flaps. Figure 16: Increase of Cl per unit flap length at different flap lengths for the DU240 airfoil at rated AoA with a curved TE flap at β=+10º Figure 17: Increase of Cl per unit flap length at different flap lengths for the FFA_W3_241 airfoil at rated AoA with a curved TE flap at β=+10º Page 34 of 141 WP no.: 3.3

35 Similarly to the Cl, Figure 18 and Figure 19 show the increase of Cd per unit flap length (in % chord) for the different flap lengths of the reference airfoils. For the DU240, there is no clear trend and the value of Cd is low because the rated angle of attack is low. For the FFA_W3_241, the Cd per unit flap length presents a minimum between 8 and 15% chord and increases with longer flaps. Figure 18: Increase of Cd per unit flap length at different flap lengths for the DU240 airfoil at rated AoA with a curved TE flap at β=+10º Figure 19: Increase of Cd per unit flap length at different flap lengths for the FFA_W3_241 airfoil at rated AoA with a curved TE flap at β=+10º Figure 20 and Figure 21 show another interesting performance indicator, the efficiency for both airfoils at rated angle of attack (α=1.1º for the DU240 and α=8.1º for the FFA_W3_241). It is important to mention that the rated angle of attack is the angle of attack obtained in a specific blade section in rated operating conditions. For the DU240, it is not clear what happens for lengths above 15% chord, but below that the efficiency decreases. Attending to the different simulations, lengths between 10-15% chord seems to be reliable for high efficiencies. For the FFA_W3_241, the efficiency drops significantly at flap lengths above 10% - 15%c. Some results at angles of attack above and below the rated value have been analysed, showing different values of the efficiency and different trends depending on the flap length. Figure 20: Efficiency at different flap lengths for the DU240 airfoil at rated AoA with a curved TE flap at β=+10º Figure 21: Efficienty at different flap lengths for the FFA_W3_241 airfoil at rated AoA with a curved TE flap at β=+10º Page 35 of 141 WP no.: 3.3

36 The same study is performed considering a negative flap angle of -10º, but the most important limitations can be observed for positive flaps, where the ΔCl and ΔCd can be affected by flow separation at lower angles of attack. As an example, Figure 22 and Figure 23 show the increase of Cl per unit length (in % chord) for increasing flap lengths at flap angle -10º. For both airfoils, the increase per unit flap length is lower for longer flaps, but the effect is more significant than for the positive angle. For negative flaps, the trend of ΔCl compared to the change obtained for positive flaps is very smooth, because in that case there is no important effect of the separation. Figure 22: Increase of Cl per unit flap length at different flap lengths for the DU240 airfoil at rated AoA with a curved TE flap at β=-10º Figure 23: Increase of Cl per unit flap length at different flap lengths for the FFA_W3_241 airfoil at rated AoA with a curved TE flap at β=-10º In the case of the INNWIND rotor, the operating conditions lead to more difficulties and uncertainties in the evaluation of the flap performance because the rated angle of attack is high and close to flow separation. The rated condition for the 75% section of the INNWIND rotor is close to the optimum angle of attack without flap (maximum efficiency). Then, the effect of the flap could cause a loss of efficiency due to the balance between the change of Cl and Cd contributions. Apart from the curved flap, a straight flap has been analysed. The conclusions have been the same but the absolute values obtained slightly differ. In the following figures, a brief comparison of the results with curved and straight flaps is included. Figure 24 and Figure 25 show the comparison of the efficiency at rated angle of attack for both airfoils, with curved or straight flap. In general, the curved flap gives higher efficiency except for the FFA_W3_241, in the simulations of FOIL2W, and WMB at length 10%c. Page 36 of 141 WP no.: 3.3

37 Figure 24: Comparison of efficiency between curved and straight flap at different flap lengths for the DU240 airfoil at rated AoA with a curved TE flap at β=+10º Figure 25: Comparison of efficiency between curved and straight flap at different flap lengths for the FFA_W3_241 airfoil at rated AoA with a curved TE flap at β=+10º Figure 26 and Figure 27 show the comparison of the ΔCl at rated angle of attack for both airfoils, with curved or straight flap. In general, the curved flap gives a higher ΔCl when the flap is deployed to a certain angle. Figure 26: Comparison of ΔCl between curved and straight flap at different flap lengths for the DU240 airfoil at rated AoA with a curved TE flap at β=+10º Figure 27: Comparison of ΔCl between curved and straight flap at different flap lengths for the FFA_W3_241 airfoil at rated AoA with a curved TE flap at β=+10º Finally, it is interesting to evaluate the effect of the flap length in the flap hinge moment because it is a relevant size/property for the actuator needed to control an active flap. Figure 28 and Figure 29 show the variation of the moment coefficient at different flap lengths with a negative and positive angle of 10º for the DU240 and FFA_W3_241 airfoils. The negative flap angle is not critical, but for the positive deployment, the moment in the actuator is approximately 10 times bigger extending the flap from 10%c to 30%c. Page 37 of 141 WP no.: 3.3

38 Figure 28: Moment in the flap hinge for the DU240 depending on the flap length at flap angles +10º and -10º Figure 29: Moment in the flap hinge for the FFA_W3_241 depending on the flap length at flap angles +10º and -10º Conclusions The following conclusions can be extracted from the study of flap length and shape: Small flaps are relatively more effective to increase / decrease the Cl. The rotor design, in terms of the rated angles of attack, has significant influence in the flap performance and practical application. Page 38 of 141 WP no.: 3.3

39 The positive deployment of the flap shows more limitations due to flow separation. The curvilinear shape outperforms the straight shape in terms of Cl change and efficiency. The effect of the flap in the hinge moment increases significantly with the flap length above 15% chord length. Based on the study of flap length and shape, the following configurations have been suggested for the rest of the study: FFA_W3_241, flap length 10% chord, curvilinear shape DU240, flap length 10% chord, curvilinear shape (15% chord would be slightly better but for the sake of simplicity it has been decided to use the same flap length in both airfoils) Study of flap angle For the configurations selected in the previous section of flap length and shape, the effect of the flap angle is investigated. The flap angle has a clear effect in Cl and Cd, and the potential benefits for load control can be limited due to flow separation or a loss of efficiency. In this section, some comparisons and metrics are included to illustrate the analysis of the flap angle influence in the aerodynamic response of the airfoil. First, the different polars obtained have been analyzed in order to confirm the same trends with the different codes. Figure 30 to Figure 33 show the Cl curves of the FFA_W3_241 airfoil, obtained for different flap angles using different codes. Figure 34 to Figure 38 show the same information for the DU240. In general, the different codes show similar trends with the variation of the flap angle. Some differences can be mentioned using the panel codes. The results of FOIL2W show values of maximum Cl a bit underestimated with positive flap angles. This is related to the fact that the positive flap increases the effective AoA and Foil2w predicts separation earlier than the other codes. AdaptFoil2D shows deviations in the attached flow region increasing with the flap angle due to the inviscid calculation. Figure 30: Cl vs angle of attack for different flap angles using FLOWer, FFA_W3_241 airfoil Page 39 of 141 WP no.: 3.3

40 Figure 31: Cl vs angle of attack for different flap angles using WMB, FFA_W3_241 airfoil Figure 32: Cl vs angle of attack for different flap angles using FOIL2W, FFA_W3_241 airfoil Page 40 of 141 WP no.: 3.3

41 Figure 33: Cl vs angle of attack for different flap angles using AdaptFoil2D, FFA_W3_241 airfoil Figure 34: Cl vs angle of attack for different flap angles using MaPFlow, DU240 airfoil Page 41 of 141 WP no.: 3.3

42 Figure 35: Cl vs angle of attack for different flap angles using FLOWer, DU240 airfoil Figure 36: Cl vs angle of attack for different flap angles using HMB2, DU240 airfoil Page 42 of 141 WP no.: 3.3

43 Figure 37: Cl vs angle of attack for different flap angles using FOIL2W, DU240 airfoil Figure 38: Cl vs angle of attack for different flap angles using AdaptFoil2D, DU240 airfoil Figure 39 to Figure 41 show the Cd curves of the FFA_W3_241 airfoil, obtained for different flap angles using different codes. Figure 42 to Figure 45 show the same information for the DU240. In general, in spite of some deviations between the different codes (observed for the airfoil without flap), the trend of the flap angle effect is similar with the different codes. In the region of positive angles of attack, increasing the flap angle, the Cd is generally increased and the sharp increase related to flow separation is found in lower angles of attack. Page 43 of 141 WP no.: 3.3

44 Figure 39: Cd vs angle of attack for different flap angles using FLOWer, FFA_W3_241 airfoil Figure 40: Cd vs angle of attack for different flap angles using WMB, FFA_W3_241 airfoil Page 44 of 141 WP no.: 3.3

45 Figure 41: Cd vs angle of attack for different flap angles using FOIL2W, FFA_W3_241 airfoil Figure 42: Cd vs angle of attack for different flap angles using MaPFlow, DU240 airfoil Page 45 of 141 WP no.: 3.3

46 Figure 43: Cd vs angle of attack for different flap angles using FLOWer, DU240 airfoil Figure 44: Cd vs angle of attack for different flap angles using HMB2, DU240 airfoil Page 46 of 141 WP no.: 3.3

47 Figure 45: Cd vs angle of attack for different flap angles using FOIL2W, DU240 airfoil Apart from Cl and Cd, Cm and Cl/Cd have also been analysed, with similar conclusions. In general, the different codes are able to estimate the effect of the flap angle in the aerodynamic response with respect to the airfoil without flap. In the following part of the analysis, the specific influence of the flap is investigated with the different tools. In order to investigate the effect of the flap angle in the aerodynamic response of the airfoil, different metrics have been used, as in the previous section for the study of flap length and shape. First, the ΔCl with respect to the airfoil without flap have been evaluated at rated angle of attack. Figure 46 and Figure 47 show the ΔCl obtained with different codes for the DU240 and FFA_W3_241 respectively, while Figure 48 and Figure 49 show the ΔCl divided by the flap angle. When the flap angle is between -20 and 5º degrees, there is no flow separation and the increase of ΔCl is linear (the increase of Cl per unit flap angle is in general slightly decreasing). Above 5º, the slope of the Cl increase is reduced, with a clear loss in the potential of variation of Cl. In that case, the increase of Cl per unit flap angle is significantly reduced. AdaptFoil2D shows an abrupt effect at the onset of separation, between 5 and 10º, while the rate of decrease before and after the jump is lower than for the rest of codes. This behaviour is because AdaptFoil2D performs an inviscid calculation as mentioned in section 5.2. The rapid decrease of the value of the results of FOIL2W at the higher flap deployments is because this code predicts separation earlier than the other codes. Page 47 of 141 WP no.: 3.3

48 Figure 46: Increase of Cl at different flap angles for the DU240 airfoil at rated AoA Figure 47: Increase of Cl at different flap angles for the FFA_W3_241 airfoil at rated AoA Figure 48: Increase of Cl per degree of flap angle at different flap angles for the DU240 airfoil at rated AoA Figure 49: Increase of Cl per degree of flap angle at different flap angles for the FFA_W3_241 airfoil at rated AoA It is also important to study the change of Cd with the flap. Figure 50 and Figure 51 show the ΔCd due to the flap using different codes for the DU240 and FFA_W3_241 respectively. Figure 52 and Figure 53 show the ΔCd divided by the flap angle. The increase of Cd with positive flap angles is less critical for the DU240. The reason is that the rated angle of attack is lower for the DU240. For both airfoils, the ΔCd per unit flap angle is increasing with the flap angle and the increase is more significant at positive flap angles above 0-10º, probably due to the onset of flow separation. Page 48 of 141 WP no.: 3.3

49 Figure 50: Increase of Cd at different flap angles for the DU240 airfoil at rated AoA Figure 51: Increase of Cd at different flap angles for the FFA_W3_241 airfoil at rated AoA Figure 52: Increase of Cd per degree of flap angle at different flap angles for the DU240 airfoil at rated AoA Figure 53: Increase of Cd per degree of flap angle at different flap angles for the FFA_W3_241 airfoil at rated AoA The relation between the efficiency and the flap angle is also analysed. Figure 54 and Figure 55 show the efficiency (Cl/Cd) at rated angle of attack. The objective of a negative deployment of the flap is usually a reduction of loads (reduction of Cl). In consequence, the efficiency is reduced at negative flap angles, and it is not a suitable measurement to evaluate the potential for load reduction. At positive flap angles, for the DU240, the efficiency is increased with respect to the case without flap up to 10-15º, while for the FFA_W3_241, the efficiency is always decreased above 5º. The reason is the value of the rated angle of attack selected for each reference rotor. In the case of the INNWIND rotor, the angle of attack is near the optimum efficiency for the FFA_W3_241 airfoil, and the margin for further improvements is drastically limited. This is also affecting the possibility to control the torque. Page 49 of 141 WP no.: 3.3

50 Figure 54: Efficienty (Cl/Cd) vs flap angles for the DU240 airfoil at rated AoA Figure 55: Efficienty (Cl/Cd) vs flap angles for the FFA_W3_241 airfoil at rated AoA Finally, it is interesting to evaluate the effect of the flap angle in the flap hinge moment. Figure 56 and Figure 57 show the effect of the flap angle in the hinge moment at rated angle of attack. In this case, the effect is more or less linear until 10º. Above 10º, the rate of increase of the hinge moment with the flap angle is reduced probably due to flow separation. Figure 56: Moment in the flap hinge for the DU240 depending on the flap angle (flap length 10%c) Page 50 of 141 WP no.: 3.3

51 Conclusions Figure 57: Moment in the flap hinge for the FFA_W3_241 depending on the flap angle (flap length 10%c) The following conclusions can be extracted from the study of flap angle: Positive flap angles above 5º 10º can lead to flow separation with a significant change of the aerodynamic effect which is not recommended. The rotor design, in terms of the rated angles of attack, has significant influence in the flap performance and practical application. The conditions of the AVATAR rotor give some margin for efficiency improvements with the flap angle, with slightly larger variations of Cl but mainly with lower increases of Cd. This allows a better potential for control purposes of loads and torque in comparison to the INNWIND rotor. At the lower angles of attack studied, the effect of the flap angle is linear in the hinge moment coefficient Study of unsteady aerodynamic response In steady conditions, the flap has a specific influence in the aerodynamic coefficients Cl, Cd and Cm. The variation of the coefficients between the case without flap and the cases with positive and negative flap angle give an idea of the potential of the flap for control purposes. However, in unsteady conditions, the values of Cl, Cd and Cm differ from the steady ones depending on the reduced frequency. The evaluation of the effect of the unsteady conditions in the aerodynamic response of the airfoil is important to understand the real potential of the flap in practical applications. In this section, unsteady oscillations and ramps of the flap deployment have been investigated. Oscillation cases Page 51 of 141 WP no.: 3.3

52 The oscillatory cases presented in this document are all with a mean value of flap angle of 0º and amplitude of 5º and 10º, at different angles of attack and for different reduced frequencies. The simulations using HMB2 for the FFA_W3_241 airfoil have been performed by CENER, while the simulations using HMB3 for the DU240 airfoil have been performed by UoG. Figure 58 to Figure 65 show the Cl and Cd cycles for the FFA_W3_241 airfoil computed with HMB2, FLOWer, FOIL2W and AdaptFoil2D at angle of attack 5º and a TE flap oscillating at different reduced frequencies. Higher frequency leads to wider Cl loops (the delay of the lift coefficient is clearly higher). For cases with higher frequency, the slope is lower, the minimum Cl increases and the maximum Cl decreases. Then, the potential for control purposes is reduced. For the Cd, the unsteady loops are wider for higher frequencies, but due to an overshoot instead of a delay, increasing the maximum and minimum Cd far away from the steady values. The trends in Cl and Cd are similar for all the codes. In the case of AdaptFoil2D, as it is an inviscid calculation, the Cd cycles are displaced to lower values because the steady value of Cd is approximately 0. In FOIL2W, the maximum values of Cd are slightly underestimated. Figure 58: Cl vs flap angle computed with HMB2, FFA_W3_241 airfoil at AoA=5º Figure 59: Cd vs flap angle computed with HMB2, FFA_W3_241 airfoil at AoA=5º Figure 60: Cl vs flap angle computed with FLOWer, FFA_W3_241 airfoil at AoA=5º Figure 61: Cd vs flap angle computed with FLOWer, FFA_W3_241 airfoil at AoA=5º Page 52 of 141 WP no.: 3.3

53 Figure 62: Cl vs flap angle computed with FOIL2W, FFA_W3_241 airfoil at AoA=5º Figure 63: Cd vs flap angle computed with FOIL2W, FFA_W3_241 airfoil at AoA=5º Figure 64: Cl vs flap angle computed with AdaptFoil2D, FFA_W3_241 airfoil at AoA=5º Figure 65: Cd vs flap angle computed with AdaptFoil2D, FFA_W3_241 airfoil at AoA=5º Figure 66 to Figure 75 show the Cl and Cd cycles for the DU240 airfoil computed with HMB3, FLOWer, MaPFlow, FOIL2W and AdaptFoil2D at angle of attack 5º and TE flap oscillations of different reduced frequencies. The influence of the reduced frequency in the Cl and Cd loops is the same observed for the FFA_W3_241 airfoil. In this case, the results show flow separation at the highest positive deployment of the flap, with a strong influence in AdaptFoil2D, although the trends of maximum and minimum values of Cl are still similar. For the Cd, the results of the positive deployment are a bit different for FOIL2W, and the reason could be a drop in the steady value of Cd. Page 53 of 141 WP no.: 3.3

54 Figure 66: Cl vs flap angle computed with HMB3, DU240 airfoil at AoA=5º Figure 67: Cd vs flap angle computed with HMB3, DU240 airfoil at AoA=5º Figure 68: Cl vs flap angle computed with FLOWer, DU240 airfoil at AoA=5º Figure 69: Cd vs flap angle computed with FLOWer, DU240 airfoil at AoA=5º Page 54 of 141 WP no.: 3.3

55 Figure 70: Cl vs flap angle computed with MaPFlow, DU240 airfoil at AoA=5º Figure 71: Cd vs flap angle computed with MaPFlow, DU240 airfoil at AoA=5º Figure 72: Cl vs flap angle computed with FOIL2W, DU240 airfoil at AoA=5º Figure 73: Cd vs flap angle computed with FOIL2W, DU240 airfoil at AoA=5º Figure 74: Cl vs flap angle computed with AdaptFoil2D, DU240 airfoil at AoA=5º Figure 75: Cd vs flap angle computed with AdaptFoil2D, DU240 airfoil at AoA=5º Page 55 of 141 WP no.: 3.3

56 Figure 76 and Figure 77 represent the cycles of lift and drag coefficient for the DU240 airfoil, in the case with reduced frequency of 0.1. All codes present similar trends. The different slope of the lift coefficient and the different mean value of the drag coefficient, reported by AdaptFoil2D in comparison with the rest of the codes are due to the inviscid calculation. This code presents also a peak on both coefficients due to onset of separation. FOIL2W shows again underestimated maximum values of Cd, explained by a drop in the steady value of Cd. Figure 76: Cl vs flap angle for flap angle oscillation of k=0.1, DU240 airfoil at AoA=5º Figure 77: Cd vs flap angle for flap angle oscillation of k=0.1, DU240 airfoil at AoA=5º Figure 78 and Figure 79 represent cycles of lift and drag coefficient in the FFA_W3_241 airfoil, for a flap oscillation with reduced frequency of As for the DU240 airfoil, all codes present similar trends. Apart from the differences of AdaptFoil2D mentioned previously, FOIL2W shows a different behaviour of Cd during the positive deployment of the flap, explained again by a drop in the steady value of Cd. Figure 78: Cl vs flap angle for flap angle oscillation of k=0.1, FFA_W3_241 airfoil at AoA=5º Figure 79: Cd vs flap angle for flap angle oscillation of k=0.1, FFA_W3_241 airfoil at AoA=5º Page 56 of 141 WP no.: 3.3

57 For an angle of attack of 10º, part of the flow can be separated. In the simulations of the DU240 airfoil, the effect of the flap angle in the separation seems to be more significant for FOIL2W. Figure 80 to Figure 83 show the Cl and Cd cycles for the DU240 airfoil computed with AdaptFoil2D and FOIL2W at angle of attack 10º and different reduced frequencies. The results of AdaptFoil2D are similar to the case at angle of attack 5º, but the range of variation of Cl is slightly lower. The results of FOIL2W present the same trends on slopes but different maximum values of lift coefficient (higher as frequency increases). This is related to the progress of separation when the flap angle is increasing and frequency is decreasing. Figure 80: Cl vs flap angle computed with AdaptFoil2D, DU240 airfoil at AoA=10º and different values of k Figure 81: Cd vs flap angle computed with AdaptFoil2D, DU240 airfoil at AoA=10º and different values of k Figure 82: Cl vs flap angle computed with FOIL2W, DU240 airfoil at AoA=10º and different values of k Figure 83: Cd vs flap angle computed with FOIL2W, DU240 airfoil at AoA=10º and different values of k Figure 78 and Figure 79 represent cycles of lift and drag coefficient in the DU240 airfoil, for a flap oscillation with reduced frequency of 0.1. The codes present similar trends, but there are more differences than in the case at angle of attack 5º due to the flow separation. Those differences can be related to the steady values. Page 57 of 141 WP no.: 3.3

58 Figure 84: Cl vs flap angle for flap angle oscillation of k=0.1, DU240 airfoil at AoA=10º Figure 85: Cd vs flap angle for flap angle oscillation of k=0.1, DU240 airfoil at AoA=10º The main conclusion of the analysis of Cl-beta and Cd-beta cycles is that the unsteady effect observed depends clearly on the reduced frequency of the motion. The selection of the KPIs for the study has been focused on the maximum and minimum values of the coefficients obtained in the unsteady cases in comparison with the steady value without flap. ΔCl min = Cl min_unsteady -Cl _NoFlap ΔCl max = Cl max_unsteady -Cl _NoFlap ΔCd min = Cd min_unsteady -Cd _NoFlap ΔCd max = Cd max_unsteady -Cd _NoFlap In the following figures, the corresponding steady values with the flap deployed have been included at zero frequency, in order to have a baseline. Figure 86 to Figure 89 show the KPI s obtained with the different codes for the DU240 airfoil at angle of attack 0º. In general, the trends of the ΔCl max and ΔCl min with the reduced frequency are similar for all the codes. For the ΔCl max of the panel codes and the trends of the CFD codes are almost the same, while for the ΔCl min all the codes show an approximately linear behaviour. For the Cd, all the codes show an increasing ΔCd max and decreasing ΔCd min, but the slopes are a bit different for the panel codes. Page 58 of 141 WP no.: 3.3

59 Figure 86: ΔClmax vs frequency at AoA=0º, Δβ=10º, DU240 airfoil Figure 87: ΔClmin vs frequency at AoA=0º, Δβ=10º, DU240 airfoil Figure 88 ΔCdmax vs frequency at AoA=0º, Δβ=10º, DU240 airfoil Figure 89: ΔCdmin vs frequency at AoA=0º, Δβ=10º, DU240 airfoil Figure 90 to Figure 93 present the same case but at angle of attack 5º. The conclusions are similar, although in this case the general trends for the change of maximum and minimum Cd show a good agreement. Page 59 of 141 WP no.: 3.3

60 Figure 90: ΔClmax vs frequency at AoA=5º, Δβ=10º, DU240 airfoil Figure 91: ΔClmin vs frequency at AoA=5º, Δβ=10º, DU240 airfoil Figure 92: ΔCdmax vs frequency at AoA=5º, Δβ=10º, DU240 airfoil Figure 93: ΔCdmin vs frequency at AoA=5º, Δβ=10º, DU240 airfoil Figure 94 to Figure 97 include the same case but at angle of attack 10º. For the Cl, in general, the conclusions are similar to angles of attack 0º and 5º, but the slope is lower. The reason is that the higher angle of attack leads to flow separation at lower frequency. For example FOIL2W presents a ΔCl max increasing with the reduced frequency at low frequencies. In the case of Cd, the analysis is similar to angle of attack 5º. Page 60 of 141 WP no.: 3.3

61 Figure 94: ΔClmax vs frequency at AoA=10º, Δβ=10º, DU240 airfoil Figure 95: ΔClmin vs frequency at AoA=10º, Δβ=10º, DU240 airfoil Figure 96: ΔCdmax vs frequency at AoA=10º, Δβ=10º, DU240 airfoil Figure 97: ΔCdmin vs frequency at AoA=10º, Δβ=10º, DU240 airfoil Figure 98 to Figure 101 show the case at angle of attack 15º. For the Cl, depending on the influence of the flap in the flow separation, the ΔCl max show increasing or decreasing values. For the ΔCl min the values are slightly increasing with the reduced frequency. Due to the separation, the trends in Cd are not as clear as for lower angles of attack, but in general all the codes show an increasing ΔCd max and decreasing ΔCd min with k. Page 61 of 141 WP no.: 3.3

62 Figure 98: ΔClmax vs frequency at AoA=15º, Δβ=10º, DU240 airfoil Figure 99: ΔClmin vs frequency at AoA=15º, Δβ=10º, DU240 airfoil Figure 100: ΔCdmax vs frequency at AoA=15º, Δβ=10º, DU240 airfoil Figure 101: ΔCdmin vs frequency at AoA=15º, Δβ=10º, DU240 airfoil For the FFA_W3_241 airfoil, the conclusions are the same. As an example, Figure 102 to Figure 105 show the KPI s obtained with the different codes for the FFA_W3_241 airfoil at angle of attack 0º and the conclusions are similar to the same case for the DU240. Page 62 of 141 WP no.: 3.3

63 Figure 102: ΔClmax vs frequency at AoA=5º, Δβ=10º, FFA_W3_241 airfoil Figure 103: ΔClmin vs frequency at AoA=5º, Δβ=10º, FFA_W3_241 airfoil Figure 104: ΔCdmax vs frequency at AoA=5º, Δβ=10º, FFA_W3_241 airfoil Figure 105: ΔCdmin vs frequency at AoA=5º, Δβ=10º, FFA_W3_241 airfoil Conclusions The previous analysis provides some information that can be useful for the design and use of flaps: It is not possible to achieve in unsteady conditions the potential increase and decrease of the Cl obtained in the steady simulations, because of the delays between excitation (flap deflection) and response (change in lift). The differences observed between a steady case and the unsteady results are really important and the design phase of a blade with flaps must take into account this unsteady effect. The ΔClmax and ΔClmin obtained present differences up to 30% between the steady case and the oscillating result with the highest frequency (k=0.2). The unsteady effects depend on the state of the flow (separated or attached). The effect of the frequency of the flap motion in the linear region presents high variations on the lift coefficient. The relation between the maximum and minimum Cl and the reduced frequency is almost linear in most of the cases. Page 63 of 141 WP no.: 3.3

64 The maximum unsteady value of Cd, with respect to the steady value, is increased and the minimum unsteady value is decreased when the reduced frequency is increased. Then, the maximum loads are increased and also the peak to peak values. At high angles of attack the variation of the Cl is reduced (separated flow) and the peak to peak of the drag coefficient is higher. Consequently, the efficiency of the use of the flaps is higher for attached flow. In general, the KPIs of this section present good agreement (especially up to 10º) and the differences observed are related to the differences in the steady polars. For the cases which does not present separated flow, all codes presents very close values of the ΔCl. The agreement between codes is poor in regions where the separated flow starts. This disagreement is in part related to the differences observed in the steady polars (presented even in the cases without flap). Ramp cases In real operating conditions the speed associated with the flap deployment will be determined by the actuator dynamics, and this influences the aerodynamic response. Then, it is important to analyse the influence of a flap ramp for different ramp rates on the aerodynamic response of the airfoil. Different flap ramps have been analysed for this purpose at different angles of attack for the two reference airfoils, DU240 and FFA_W3_241. In this document, only ramps with low angles of attack, close to rated condition on the AVATAR rotor (0º and 5º), will be presented. Ramps presented starts at 0º of flap angle and ends at 10º and -10º. The value at the end of the ramp of the different coefficients will be compared with the equivalent steady value obtained in the flap angle analysis, in order to measure the delay on the aerodynamic response and the influence of the dynamic response in comparison to the steady-state values. First, different ramps obtained for each code have been analysed for several ramp rates. Figure 106 to Figure 108 show the results of Cl and Cd with different codes for the DU240 airfoil, in angle of attack 0º and flap angle ramp between 0º and 10º. The same case but for the FFA_W3_241 airfoil is shown in Error! No se encuentra el origen de la referencia. to Error! No se encuentra el origen de la referencia.. Page 64 of 141 WP no.: 3.3

65 Figure 106: Cl vs flap angle at different ramp rates, DU240 airfoil computed with Adaptfoil2D, α=0º Figure 107: Cd vs flap angle at different ramp rates, DU240 airfoil computed with Adaptfoil2D, α=0º Figure 108: Cl vs flap angle at different ramp rates, DU240 airfoil computed with FOIL2W, α=0º Page 65 of 141 WP no.: 3.3

66 Figure 109: Cl vs flap angle at different ramp rates, FFA_W3_241 airfoil computed with Adaptfoil2D, α=0º Figure 110: Cd vs flap angle at different ramp rates, FFA_W3_241 airfoil computed with Adaptfoil2D, α=0º Figure 111: Cl vs flap angle at different ramp rates, FFA_W3_241 airfoil computed with FOIL2W, α=0º Figure 112 to Figure 116 show the results of different codes for the DU240 airfoil and Error! No se encuentra el origen de la referencia. to Error! No se encuentra el origen de la referencia. for the FFA_W3_241 airfoil for the same flap ramp (β = [0, 10 ]) at 5º angle of attack. Page 66 of 141 WP no.: 3.3

67 Figure 112: Cl vs flap angle at different ramp rates, DU240 airfoil, computed with Adaptfoil2D, α=5º Figure 113: Cd vs flap angle at different ramp rates, DU240 airfoil computed with Adaptfoil2D, α=5º Figure 114: Cl vs flap angle at different ramp rates, DU240 airfoil computed with FOIL2W, α=5º Figure 115: Cl vs flap angle at different ramp rates, DU240 airfoil computed with MaPFlow, α=5º Figure 116: Cd vs flap angle at different ramp rates, DU240 airfoil computed with MaPFlow, α=5º Page 67 of 141 WP no.: 3.3

68 Figure 117: Cl vs flap angle at different ramp rates, FFA_W3_241 airfoil computed with Adaptfoil2D, α=5º Figure 118: Cd vs flap angle at different ramp rates, FFA_W3_241 airfoil computed with Adaptfoil2D, α=5º Figure 119: Cl vs flap angle at different ramp rates, FFA_W3_241 airfoil computed with FOIL2W, α=5º Figure 120: Cl vs flap angle at different ramp rates, FFA_W3_241 airfoil computed with FLOWer, α=5º Figure 121: Cd vs flap angle at different ramp rates, FFA_W3_241 airfoil computed with FLOWer, α=5º Page 68 of 141 WP no.: 3.3

69 The differences mentioned in the previous sections are also clear in this analysis, such as the higher mean value of Cl reported by AdapFoil2D due to the inviscid calculation. However, the trends are very similar between the different codes. Peaks on AdapFoil2D results for the DU240 airfoil are due to flow separation and are not present in the FFA_W3_241 results. As it could be expected the lift coefficient increases for lower ramp rates due to a lower delay. In general, the progress of the drag coefficient with the flap angle is almost parabolic with higher values as ramp rate increases, so an increasing overshoot is observed in the dynamic response. The negative values observed in the results of AdaptFoil2D for the DU40 airfoil could be related with flow separation. In order to analyse the effect of the unsteady response of the flap deployment, the difference between lift and drag coefficients at the end of the ramp and the corresponding steady values from the steady flap angle study have been calculated for each code and ramp rate. Figure 122 and Figure 124 represent the shift of the lift coefficient with the flap ramp rate at 5º angle of attack at the end of the ramp (when flap angle 10º is reached), for the DU240 and FFA_W3_241 respectively. Figure 123 and Figure 125 represent the shift of drag coefficient at the same case. As it has been mentioned earlier, the DU240 results from AdaptFoil2D are influenced by flow separation. Then, although the trends match quite well for both lift and drag coefficient with the rest of the codes, showing that the delay in Cl and the overshot in Cd between steady and unsteady flap deployment increases with the ramp rate, mean values of these differences could vary. For the FFA_W3_241 airfoil, AdaptFoil2D predicts a larger delay of lift coefficient with the same rate of flap deployment than FOIL2W and FLOWer. These differences are not so pronounced in the drag coefficient, where the differences between codes stay closer with the rate of flap deployment. Page 69 of 141 WP no.: 3.3

70 Figure 122: Shift of Cl vs ramp rate at the end of the ramp, DU240 airfoil, at β=10º, α=5º Figure 123: Shift of Cd vs ramp rate at the end of the ramp, DU240 airfoil, at β=10º, α=5º Figure 124: Shift of Cl vs ramp rate at the end of the ramp, FFA_W3_241 airfoil, at β=10º, α=5º Figure 125: Shift of Cd vs ramp rate at the end of the ramp, FFA_W3_241 airfoil, at β=10º, α=5º Ramps with negative flap deployment have also been analyzed. Figure 126 and Figure 127 represent the lift and drag coefficient with flap deployment for the DU240 airfoil, at 5º angle of attack in a flap ramp between 0º and -10º, computed with AdaptFoil2D. Figure 128 represent the Cl computed with FOIL2W. In this case, the trends between codes agree better than those for positive flap ramp because flow remains attached. The shift between AdaptFoil2D and FOIL2W is because the calculation of AdaptFoil2D is potential. Peaks on FOIL2W results for this case are a numerical consequence of the refinement of time step during the ramp, with no physical meaning. Page 70 of 141 WP no.: 3.3

71 Figure 126: Cl vs flap angle at different ramp rates, DU240 airfoil computed with AdaptFoil2D, α=5º Figure 127: Cd vs flap angle at different ramp rates, DU240 airfoil computed with AdaptFoil2D, α=5º Figure 128: Cl vs flap angle at different ramp rates, DU240 airfoil computed with FOIL2W, α=5º Comparisons of the unsteady lift and drag referenced to the steady value of the deployed flap are shown in Figure 129 and Figure 130 for the DU240 airfoil and in Figure 131 and Figure 132 for the FFA_W3_241 airfoil. Figure 129 to Figure 132 show the difference between the dynamic response and quasi-steady Cl and Cd values at the end of the ramp in the case of negative flap deployment. The delay of Cl increases with the ramp rate. For the Cd, an increase in the negative overshoot is observed with the increase of the ramp rate. In general, the trends obtained with the different codes are similar. Page 71 of 141 WP no.: 3.3

72 Figure 129: Shift of Cl vs ramp rate at the end of the ramp, DU240 airfoil, at β=-10º, α=5º Figure 130: Shift of Cd vs ramp rate at the end of the ramp, DU240 airfoil, at β=-10º, α=5º Figure 131: Shift of Cl vs ramp rate at the end of the ramp, FFA_W3_241 airfoil, at β=-10º, α=5º Figure 132: Shift of Cd vs ramp rate at the end of the ramp, FFA_W3_241 airfoil, at β=-10º, α=5º Conclusions With this analysis of unsteady flap deployment at different flap rates it has been established that: The unsteady deployment of the flap implies a delay of the lift signal and an overshoot of the drag signal with respect to the steady values. Differences observed between codes remains similar to those observed in the analysis of the steady flap deployment, including the differences according to prediction of flow separation. When considering the suitable flap deployment in specific operating conditions for control purposes, it should be taken into account that: Page 72 of 141 WP no.: 3.3

73 The difference between the unsteady and the steady value (converged value) of the lift coefficient seems to be linear at the lower ramp rates but the slope decreases at higher ramp rates. The difference between the unsteady and the steady value of the drag coefficient follows a parabolic trend with the rate of flap deployment, resulting in an overshoot for positive and negative flap ramps Study of influence in the 3D rotor The last step in the parametric study of TE flaps is the evaluation of the technology in the AVATAR and INNWIND reference wind turbines. As it is mentioned in the description of the test cases, different mid-point locations and radial extensions for the TE flap in the blade have been tested. The objective is to evaluate the effect in the rotor performance of the TE flap distribution in the blade. First, the results of the different codes have been compared in terms of the load coefficients in the blade, in order to analyze the effect of the flap and cross-check the trends obtained with different tools. Figure 133 and Figure 134 show the Cn and Ct (force coefficients normal and tangential to the chord direction) along the blade radius for the AVATAR rotor at rated wind speed conditions without flap. The different results are similar but there are some deviations that can be related to differences between the codes or differences in the modelling (consideration of trailing vortices, blade discretization, etc). Figure 135 to Figure 138 show the same cases but with a flap deployed at 10º. Figure 135 and Figure 136 show a flap centered in the 80% radial station with 10% extension and Figure 137 and Figure 138 show a flap centered in the 90% radial station with 10% extension. In general, the effect of the flap on Cn follows a similar trend, with an increase of the Cn in the flap region also visible in a small area around the flaps slightly extended outside the flap. The effect inside the flap is slightly less for FLOWer (this is caused by flow separation in the fillet at the pressure side of the blade) and bigger for GENUVP. The 3D effects outside the flap are extended in the results of MIRAS and GENUVP, due to the progressive mesh transition between the regions with and without flap. For the Ct, the differences are more significant although the same trends are captured by the codes, with a decrease of the Ct. Results of FAST show a bigger drop in the Ct, while GENUVP show a smaller drop, than the other three results. MIRAS show again the extension of the effect due to the longer transition flap no flap. Page 73 of 141 WP no.: 3.3

74 Figure 133: Cn along the blade for the AVATAR rotor, in rated wind speed conditions without flap Figure 134: Ct along the blade for the AVATAR rotor, in rated wind speed conditions without flap Figure 135: Cn along the blade for the AVATAR rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 136: Cn along the blade for the AVATAR rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 137: Cn along the blade for the AVATAR rotor, in rated wind speed conditions with flap in the 90% and extension 10% Figure 138: Ct along the blade for the AVATAR rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 139 and Figure 140 show the Cn and Ct along the blade radius for the INNWIND rotor at rated wind speed conditions without flap. The different results are similar except a slight overestimation of the results of WMB. Figure 141 a Figure 142 show the Cn and Ct with a flap with angle of 10º centered in the 80% radial station with 10% extension, and Figure 143 and Figure 144 show the same flap centered in the 90% radial station. In general, the effect of the flap in the Cn follows a similar trend, with an increase of the Cn in the flap region also slightly Page 74 of 141 WP no.: 3.3

75 extended outside the flap. The effect inside the flap is a bit different, the major effect is observed again for GENUVP and FAST, while the effect for FLOWer and WMB is lower. For the Ct, the differences are large between the codes, and the integrated contribution to the global loads can be very different. FAST shows an overestimation of the drop in Ct, which is related to the steady polars used for this rotor, and GENUVP show an increase of Ct in contrast to the others. Figure 139: Cn along the blade for the INNWIND rotor, in rated wind speed conditions without flap Figure 140: Ct along the blade for the INNWIND rotor, in rated wind speed conditions without flap Figure 141: Cn along the blade for the INNWIND rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 142: Ct along the blade for the INNWIND rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 143: Cn along the blade for the INNWIND rotor, in rated wind speed conditions with flap in the 90% and extension 10% Figure 144: Ct along the blade for the INNWIND rotor, in rated wind speed conditions with flap in the 80% and extension 10% Page 75 of 141 WP no.: 3.3

76 The previous analysis of the coefficients is useful to evaluate the effect of the flap in the different blade sections and cross-check the trends obtained from the different codes. However, from the point of view of the rotor performance, the forces in-plane and out-of-plane are more interesting. Figure 145 and Figure 146 show the aerodynamic forces Fx (out-of-plane) and Fy (in-plane) without flap. The relative differences are more significant in the case of Fy, but the values and trends are similar in all the codes. The differences between HMB and FLOWer are still being investigated, and it seems a different consideration of friction forces is causing the differences in driving force. The differences in thrust are still to clarify. Figure 147 to Figure 150 show the same forces but with a flap deployed at 10º. Figure 147 and Figure 148 show a flap centered in the 80% radial station with 10% extension and Figure 149 and Figure 150 show the same flap centered in the 90% radial station. In general, the effect of the flap in the Fx is similar with the different codes, with an increase of the in the flap region slightly extended outside the flap. The analysis is similar to the one performed for the Cn. For the Fy, the differences between the codes are more significant, but in general an increase with respect to the case without flap is observed in the flap region leading to a higher torque. Figure 145: Fx along the blade for the AVATAR rotor, in rated wind speed conditions without flap Figure 146: Fy along the blade for the AVATAR rotor, in rated wind speed conditions without flap Figure 147: Fx along the blade for the AVATAR rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 148: Fy along the blade for the AVATAR rotor, in rated wind speed conditions with flap in the 80% and extension 10% Page 76 of 141 WP no.: 3.3

77 Figure 149: Fx along the blade for the AVATAR rotor, in rated wind speed conditions with flap in the 90% and extension 10% Figure 150: Fy along the blade for the AVATAR rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 151 and Figure 152 show the Fx and Fy along the blade radius for the INNWIND rotor at rated wind speed conditions without flap. In this case, the results show a bigger dispersion than for the AVATAR rotor. Figure 153 and Figure 154 show the Fx and Fy with a flap with angle 10º centered in the 80% radial station with 10% extension, and Figure 155 and Figure 156 show the same flap centered in the 90% radial station. In general, the effect of the flap in the Fx is again similar with the different codes, with an increase of the force in the flap region slightly extended outside the flap. As for the Cn, the effect inside the flap is not exactly the same for all the codes. For the Fy, there are significant deviations between the different codes, and the global effect in the integrated Fy is not very clear and depends on the specific result. Figure 151: Fx along the blade for the INNWIND rotor, in rated wind speed conditions without flap Figure 152: Fy along the blade for the INNWIND rotor, in rated wind speed conditions without flap Page 77 of 141 WP no.: 3.3

78 Figure 153: Fx along the blade for the INNWIND rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 154: Fy along the blade for the INNWIND rotor, in rated wind speed conditions with flap in the 80% and extension 10% Figure 155: Fx along the blade for the INNWIND rotor, in rated wind speed conditions with flap in the 90% and extension 10% Figure 156: Fy along the blade for the INNWIND rotor, in rated wind speed conditions with flap in the 80% and extension 10% The final step of the analysis is the evaluation of the effect of the TE flap in the performance of the wind turbine and the comparison between the different proposed configurations. The increase or decrease of power and thrust with respect to the case without flap using a flap at 10º and -10º are evaluated. Figure 157 to Figure 160 show the percentage of variation of thrust and power for the AVATAR rotor at rated wind speed with a positive and negative static flap in all the proposed configurations. For the positive flap, there are a few configurations with results from different codes and in general the trends are the same although the values can be a bit different. Taking into account the trends and all the configurations obtained with FAST, the effect of the flap increases with the flap extension. In general, the highest potential is with the flap in the 85% for extensions of 15% and 20% and in the 90% for extensions of 10% and 5%, except for the power with positive flap when the highest potential is always with the flap in the 85% independently of the extension. Page 78 of 141 WP no.: 3.3

79 Figure 157: % variation of thrust with a static TE flap at angle 10º in different configurations, AVATAR Figure 158: % variation of aero power with a static TE flap at angle 10º in different configurations, AVATAR Figure 159: % variation of thrust with a static TE flap at angle -10º in different configurations, AVATAR Page 79 of 141 WP no.: 3.3

80 Figure 160: % variation of aero power with a static TE flap at angle -10º in different configurations, AVATAR Figure 161 to Figure 164 show the same study of flap configuration in the blade but for the INNWIND rotor. In this case, the analysis of the thrust is similar to the AVATAR rotor. For positive flap angle, the best configurations with the highest potential are with the flap in the 90% for all extensions (except obviously 20% with a maximum in 85%). For negative flap angle, the best configuration is always with the flap centred in the 85%. The potential for thrust variation in this case is lower than in the AVATAR rotor because the thrust level is higher for this rotor. For the power, the different codes show different trends, but in all the cases the potential of variation of the power have been significantly reduced with respect to the AVATAR rotor. The reason is that the configuration of the INNWIND is close to the maximum Cl/Cd of the different airfoils in the blade sections and it is not possible to have a significant increase of the torque acting with the flap. Consequently, the results of the different codes are very sensitive to the estimation of Cd or to the calculation of the induction. Figure 161: % variation of thrust with a static TE flap at angle 10º in different configurations, INNWIND Page 80 of 141 WP no.: 3.3

81 Figure 162: % variation of aero power with a static TE flap at angle 10º in different configurations, INNWIND Figure 163: % variation of thrust with a static TE flap at angle -10º in different configurations, INNWIND Figure 164: % variation of aero power with a static TE flap at angle -10º in different configurations, INNWIND Page 81 of 141 WP no.: 3.3

82 Apart from the study of rated wind speed, the rated wind speeds +2 m/s and -2 m/s have been included in the study with similar results. Conclusions The main conclusions of the analysis of the influence of TE flap over 3D rotor are: The force coefficients present evident differences between codes for the cases with and without flap. This difference is also present in forces, torque and thrust. In general, the trends observed in force and coefficients in the comparison between the cases with and without flap are similar for all codes, especially for the AVATAR rotor. The AVATAR rotor is better suited for load reduction or power control purposes with a TE flap in comparison to the INNWIND rotor because the potential of modification of the aerodynamic forces with the flap is larger in operating conditions near rated. The effect of the TE flap is higher in terms of thrust than in power so a TE flap would fit better with a load reduction purpose than with power control. The configuration with highest potential of variation, in both turbines is a flap centered in the 85% of the blade with 20% of extension. For a reduced extension of 15%, the best configuration is again with the flap centered in the 85%, but with shorter extensions the optimum position is displaced to the 90% in the AVATAR rotor. 6.2 LE flap Study of unsteady aerodynamic response The TE flap technology is very suitable for load reduction in operating conditions mainly involving attached flow in the blade sections. In addition, it can be used for torque control in certain rotor configurations, as it has been observed in the study of the previous sections. Taking into account the effect in the aerodynamic response, it is clear that TE flap is better suited than LE flap for those applications. When the operating conditions lead to high angles of attack and critical unsteady cycles (extreme yaw, extreme shear, gusts, etc), the TE flap is not capable to deal with the cyclic unsteady loads, and the effect of the flap deployment in the dynamic stall is very limited. In the present study, the objective is to evaluate the effect of the LE flap in the unsteady cycles and dynamic stall due to an oscillation of the angle of attack in the range of light and deep stall (this could be caused by extreme yaw or extreme shear for example). Figure 165 to Figure 170 show an example of angle of attack oscillation for the DU240 airfoil with LE flap of 20% chord at angle 10º, 0º and -10º. The figures show the Cl, Cd and Cm versus time or angle of attack. Those results have been obtained with the code HMB3 (UoG). In principle, in steady simulations and attached flow, a negative LE flap leads to a slight decrease of Cl. In these unsteady conditions, it is clear than the negative flap angle (acting in the nose Page 82 of 141 WP no.: 3.3

83 down direction) leads to higher values of Cl and Cm (the Cm is obtained in the 35% chord, pitch axis) due to a delay in the flow separation progress, and lower values of Cd. Apart from the unsteady cycles, there is no clear indication of the dynamic stall LE vortex shedding. Figure 165: Cl vs time for the DU240, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 166: Cl vs AoA for the DU240, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 167: Cd vs time for the DU240, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 168: Cd vs AoA for the DU240, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 169: Cm vs time for the DU240, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 170: Cm vs AoA for the DU240, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 171 to Figure 176 show the same angle of attack oscillation for the FFA_W3_248 airfoil with LE flap of 20% chord at angle 10º, 5º and -10º. The results have been obtained with the code AdaptFoil2D (CENER). The main difference with respect to the DU240 is for flap angle 10º, with an abrupt dynamic stall, clearly indicated by the peaks of Cl, Cd and Cm. The dynamic stall also leads to a sharp drop of Cl after the peak with lower minimum values of Cl during the reattachment. Page 83 of 141 WP no.: 3.3

84 Figure 171: Cl vs time for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 172: Cl vs AoA for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 173: Cd vs time for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 174: Cd vs AoA for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 175: Cm vs time for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 176: Cm vs AoA for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 The main reason why an abrupt dynamic stall is observed in the FFA_W3_248 and not in the DU240 can be related to the steady aerodynamic characteristics of each airfoil. Figure 177 shows the comparison between the Cl for both airfoils. The maximum Cl and subsequent stall is at least 5º delayed in the case of the DU240, and this can also delay the dynamic stall process. Page 84 of 141 WP no.: 3.3

85 Figure 177: Comparison of Cl for the FFA W3 248 and the DU240 In addition to the loads and moments, different KPI s have been analysed in order to study the effect of the LE flap angle in the unsteady aerodynamic response of the airfoil. In general, the maximum Cl, Cd and Cm (or minimum Cm), and the peak to peak Cl (Cl max -Cl min ) have been considered. For the DU240, the KPI s have been obtained from results of HMB2 (UoG), while for the FFA_W3_248, results of AdaptFoil2D (CENER) have been used. For the DU240, Figure 178 shows the maximum Cl for an angle of attack oscillation at different frequencies with flap angles 10º, 0º and -10º. The maximum Cl is always higher for lower flap angles due to a delay in the separation which increases the maximum load obtained even for the steady polars. Figure 178: Maximum Cl for the DU240, angle of attack oscillation: α=15±10º, k=0.05, 0.1, 0.15, Re=18.59M, M=0.18 Figure 179 shows the peak to peak Cl (Cl max -Cl min ). The value for flap angle -10º is in general higher due to the higher value of the maximum Cl. The minimum Cl in general converges to the attached value except for the highest frequency where the unsteady loop in the case of flap angle 10º lies below due to a delayed reattachment. Page 85 of 141 WP no.: 3.3

86 Figure 179: Peak to peak Cl for the DU240, angle of attack oscillation: α=15±10º, k=0.05, 0.1, 0.15, Re=18.59M, M=0.18 Figure 180 show the comparison for the maximum Cd. The values of flap 10º are above the values of flap -10º and 0º because the extension of the separation is bigger. In addition, the increase of the frequency increases in general de overshoot of Cd with respect to the steady value. Figure 180: Maximum Cd for the DU240, angle of attack oscillation: α=15±10º, k=0.05, 0.1, 0.15, Re=18.59M, M=0.18 For the FFA_W3_248, Figure 181 shows the maximum Cl for the same angle of attack oscillation at different frequencies with flap angles 10º, 5º and -10º. In this case, the maximum Cl is significantly higher for flap angle 10º at frequencies 0.1 and 0.15 due to the occurrence of dynamic stall. The reason is probably that the positive deployment of the LE leads to conditions of Cp in the LE region more favourable for dynamic stall than in the case of a motion in the nose down direction. At low frequencies of 0.05 and below, dynamic stall is supressed and the case of flap angle -10º shows again higher maximum Cl as for the DU240. Page 86 of 141 WP no.: 3.3

87 Figure 181: Maximum Cl for the FFA_W3_248, angle of attack oscillation: α=15±10º, Re=18.59M, M=0.18 Figure 182 shows the peak to peak Cl (Cl max -Cl min ). The value for flap angle 10º is in general significantly higher due to the higher value of the maximum Cl and a lower value of Cl in the reattachment due to the abrupt stall and reattachment delay. The suppression of dynamic stall at lower frequencies leads to similar values of the peak to peak Cl. Figure 182: Peak to peak Cl for the FFA_W3_248, angle of attack oscillation: α=15±10º, Re=18.59M, M=0.18 Figure 183 show the comparison for the maximum Cd. At low frequencies, the values of flap 10º are slightly above the values of flap 5 and -10º due to a less delayed separation. At higher frequencies, the values of flap 10º are significantly above the values of flap 5 and -10º due to the dynamic stall effect. Page 87 of 141 WP no.: 3.3

88 Figure 183: Maximum Cd for the FFA_W3_248, angle of attack oscillation: α=15±10º, Re=18.59M, M=0.18 For the FFA_W3_248, different simulations have been performed in order to evaluate the effect of the flap angle depending on the flap length and mean and amplitude of the angle of attack oscillations. Figure 184 to Figure 187 show the differences between oscillations with flap angle - 10º and -10º with different flap lengths. At low frequency, the possibility to have dynamic stall is reduced, as it has been observed previously, and the results with positive and negative flaps are similar. This is observed in the Figures for frequency 0.05 except for flap length 20%c and flap angle 10º where dynamic stall is present. At higher frequencies, dynamic stall increases the unsteady cycle for flap angle 10º, and the effect seems to be more significant for flap lengths 20% and 25% chord. Figure 184: Maximum Cl for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.05, Re=18.59M, M=0.18 Figure 185: Maximum Cl for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Page 88 of 141 WP no.: 3.3

89 Figure 186: Peak to peak Cl for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.05, Re=18.59M, M=0.18 Figure 187: Peak to peak Cl for the FFA_W3_248, angle of attack oscillation: α=15±10º, k=0.1, Re=18.59M, M=0.18 Figure 188 to Figure 191 show the differences of flap angle -10º and 10º with different oscillations of the flap angle. The 4 configurations are: º º º º At frequency 0.05, the differences are not very significant. At frequency 0.1, the oscillation 1 and 2 show a similar peak to peak Cl and the maximum Cl is higher in the case of flap angle -10º. The reason is that dynamic stall is not present, in configuration 1 because the maximum angle of attack of 20º is too small and in configuration 2 because the minimum angle of attack of 20º is not enough for the flow reattachment. However, in configurations 3 and 4, the angle of attack vary between 5 and 25º and 10 and 30º respectively, a range suitable for the dynamic stall occurrence in this airfoil and the results of Cl show a significant increase of maximum Cl and peak to peak Cl, mainly for flap angle 10º. Figure 188: Maximum Cl for the FFA_W3_248 at different angle of attack oscillations, with LE flap of 20%c Figure 189: Maximum Cl for the FFA_W3_248 at different angle of attack oscillations, with LE flap of 20%c Page 89 of 141 WP no.: 3.3

90 Figure 190: Peak to peak Cl for the FFA_W3_248 at different angle of attack oscillations, with LE flap of 20%c Figure 191: Peak to peak Cl for the FFA_W3_248 at different angle of attack oscillations, with LE flap of 20%c Conclusions In general, the main conclusions of the unsteady aerodynamics of the airfoil with a LE flap are: The static LE flap angle has a strong influence in the unsteady aerodynamic response associated with angle of attack motions (oscillations). The effect of the flap depends also on its length. The specific effect of flap angle is directly influenced by the occurrence (or not) of the dynamic stall phenomenon. Consequently, the effect of the flap angle in the unsteady aerodynamic response depends on the parameters which have an influence in the dynamic stall, as for example the specific aerodynamic characteristics of the airfoil, the reduced frequency, the mean angle of attack and the amplitude of the oscillation. The use of a LE flap for applications with the objective of controlling the unsteady aerodynamic response should be designed considering the steady and unsteady characteristics of the airfoil and operating conditions of the practical case. 6.3 VGs Study of VGs height and position in the chord direction For the parametric study of VGs, the first step is the evaluation of height and chord location. The height is varied between 0 (without VGs) and 90 mm, while the devices are placed in 3 different chord positions, 25, 30 and 40%c. The polars of different combinations of chord location and heights have been compared in order to cross-check the different simulations and assure similar trends for the subsequent analysis. In the next Figures, some comparisons are presented for the FFA_W3_333 airfoil. The analysis of the DU331 airfoil is similar. Figure 192 and Figure 193 show the Cl obtained by DTU WIND and NTUA with VGs of different heights in the 30% chord. In general, the VGs re-energize the boundary layer leading to a delay of flow separation and increase the value of Cl. This is not true in the results of NTUA in the linear region for the lowest heights (this has also been observed in results of DTU WIND with Page 90 of 141 WP no.: 3.3

91 the VGs in other chord locations), and for height 60mm when the flow separation seems to appear earlier. For intermediate values of the VGs height, the trend is always an increase of Cl with the height, mainly at higher angles of attack. Figure 194 and Figure 195 show the Cd obtained by DTU WIND and NTUA with VGs of different heights in the 30% chord. In the results of DTU WIND, there is a slight increase of Cd in the linear region for the cases with VGs, but the value is very similar between the different heights. In the results of NTUA, the trend is not so obvious. Increasing the angle of attack, the Cd without VGs show a sharp increase at lower angles than the rest of the configurations (except for the case of height 60mm calculated by NTUA). In general, the sharp increase of Cd is delayed with the height of the VGs. It seems that there is a clear trend for the VGs in terms of the effect on Cl and Cd at intermediate heights, and in some of the simulations those trends are cancelled when the VGs are too small or too large. The tools have been considered useful in the analysis, and the objective is to understand if there is an optimum value for the VGs height and chord location. Figure 192: Cl vs α for the DU331 airfoil, VGs at 30% chord location with different heights Page 91 of 141 WP no.: 3.3

92 Figure 193: Cd vs α for the FFA_W3_333 airfoil, VGs at 30% chord location with different heights Figure 194: Cd vs α for the DU331 airfoil, VGs at 30% chord location with different heights Page 92 of 141 WP no.: 3.3

93 Figure 195: Cd vs α for the FFA_W3_333 airfoil, VGs at 30% chord location with different heights For the parametric study, in order to evaluate the effect of the different configurations of VGs height and chord location in the aerodynamic performance of the airfoils or blade section, some KPI s have been used. The different configurations presented in the comparisons for the DU331 airfoil are: 0: No VGs 1: 25% chord, height 10mm 2: 25% chord, height 12mm 3: 25% chord, height 15mm 4: 25% chord, height 18mm 5: 25% chord, height 30mm 6: 25% chord, height 36mm 7: 25% chord, height 60mm 8: 30% chord, height 15mm 9: 30% chord, height 18mm 10: 30% chord, height 30mm 11: 30% chord, height 36mm 12: 30% chord, height 60mm 13: 40% chord, height 15mm 14: 40% chord, height 18mm 15: 40% chord, height 30mm 16: 40% chord, height 36mm 17: 40% chord, height 60mm 18: 40% chord, height 90mm The different configurations presented in the comparisons for the FFA_W3_333 airfoil are: 0: No VGs 1: 25% chord, height 6mm Page 93 of 141 WP no.: 3.3

94 2: 25% chord, height 10mm 3: 25% chord, height 12mm 4: 25% chord, height 15mm 5: 25% chord, height 18mm 6: 25% chord, height 24mm 7: 25% chord, height 30mm 8: 25% chord, height 36mm 9: 25% chord, height 60mm 10: 30% chord, height 9mm 11: 30% chord, height 15mm 12: 30% chord, height 18mm 13: 30% chord, height 30mm 14: 30% chord, height 36mm 15: 30% chord, height 60mm 16: 40% chord, height 15mm 17: 40% chord, height 17mm 18: 40% chord, height 18mm 19: 40% chord, height 30mm 20: 40% chord, height 34mm 21: 40% chord, height 36mm 22: 40% chord, height 60mm 23: 40% chord, height 68mm 24: 40% chord, height 90mm First the maximum Cl obtained is studied depending on the VGs configuration in terms of height and chord location. Figure 196 and Figure 197 show the results for different configurations of the VGs in the DU331 and FFA_W3_333 airfoils respectively. For the DU331, the maximum lift coefficient increases with the height of the VGs at 25% and 30% chord, and in general at 40% except for the highest heights where abrupt drops can be observed. It seems that VGs closer to the LE restrict the drop of Cl probably delaying the separation at higher AoAs. For the FFA_W3_333, the behaviour is similar but the drop is observed also in the 30% chord location for the simulations performed by DTU. Page 94 of 141 WP no.: 3.3

95 Figure 196: Maximum Cl at different configurations for the DU331 airfoil Figure 197: Maximum Cl at different configurations for the FFA_W3_333 airfoil The results of the angle of attack for max efficiency are observed in Figure 198 and Figure 199. The trends obtained by NTUA and DTU are different. In the simulations of NTUA, the AoA of maximum efficiency is mainly at 12º for the DU331 and 10º for the FFA_W3_333, while for the simulations of DTU that angle increases with the VGs height from 8 to 14º. Page 95 of 141 WP no.: 3.3

96 Figure 198: Angle of attack for maximum efficiency at different configurations for the DU331 airfoil Figure 199: Angle of attack for maximum efficiency at different configurations for the FFA_W3_333 airfoil In Figure 200 and Figure 201, the maximum efficiency is presented depending on the VG height and chord location for the DU240 and FFA_W3_241 airfoils respectively. Efficiency increases with the height of the VGs, except a drop observed in the simulations with the BAY model at the biggest values of height. The trend of the BAY models at 30% chord for the FFA_W3_333 airfoil is similar (the simulations of NTUA reach lower VGs height). In general, the results of NTUA show lower efficiencies, mainly due to the higher Cd values. According to the results obtained Page 96 of 141 WP no.: 3.3

97 with the BAY model, the maximum efficiency for the FFA_W3_333 would be close to the following heights: 30mm at 25% chord, 30-36mm at 30% chord and 60mm at 40% chord, and for the DU331: 30mm at 25% chord, 36mm at 30% chord and 36mm at 40% chord. Figure 200: Maximum efficiency at different configurations for the DU331 airfoil Figure 201: Maximum efficiency at different configurations for the FFA_W3_333 airfoil It is also interesting to evaluate the increase of Cl with respect to the case without VGs in specific angles of attack depending on the configuration. Figure 202 and Figure 203 show the increase of Cl for the reference airfoils at rated angle of attack (α=3º for the DU331 and α=8.8º Page 97 of 141 WP no.: 3.3

98 for the FFA_W3_333). The trend is similar with all the codes with an increase of the Cl with the VGs height. It seems that the increase is reduced (or even the trend is inverted) at the highest values of the VGs height. The sensitivity to the height seems bigger with the VGs at the 25%c. The maximum values are reached in the 30% and 40% chord location attending to the simulations with the BAY model. Figure 202: Increase of Cl with respect to the case without VGs at rated angle of attack for the DU331 airfoil Figure 203: Increase of Cl with respect to the case without VGs at rated angle of attack for the FFA_W3_333 airfoil Page 98 of 141 WP no.: 3.3

99 Figure 204 and Figure 205 show the change of Cd for the reference airfoils at rated angle of attack. With the available information, the trend is similar using the DTU and NTUA BAY models for the FFA_W3_333 with the VGs in the 30%c (the lowest VGs height is different for the NTUA simulations: 9mm and DTU simulations: 15mm). In general, the variation of Cd is not very significant except in some cases for the lowest and highest values of the VGs height. The results obtained with VGFlow show a trend completely different to the results of the BAY model and significantly larger Cd. It would be possible that VGFlow is overestimating the Cd. Figure 204: Increase of Cd with respect to the case without VGs at rated angle of attack for the DU331 airfoil Figure 205: Increase of Cd with respect to the case without VGs at rated angle of attack for the FFA_W3_333 airfoil Page 99 of 141 WP no.: 3.3

100 Figure 206 and Figure 207 show another interesting performance indicator, the efficiency increase with respect to the case without VGs for both airfoils at rated angle of attack (α=3º for the DU240 and α=8.8º for the FFA_W3_241). For the FFA_W3_333, with the VGs in the 30% chord location, the trends observed using the BAY models are similar (taking into account that the lowest VGs height is 9mm for NTUA and 15mm for DTU). The VGFlow results shows in general an increasing efficiency, and lower than the other codes. Attending to the BAY results for the DU331, the efficiency increase presents maximum values at 18mm for 25% chord, at 30 mm for 30% chord and at mm at 40% chord. The only configurations with positive ΔEff are between mm at 40% chord. For the FFA_W3_333, the efficiency increase presents maximum values at 30mm for 25% chord, at mm for 30% chord and at 30 mm for 40% chord (although at 60 mm there is a similar value). In this case, most of the configurations obtained with the BAY model show a positive increase of the efficiency. The reason is that the rated angle of attack is lower for the INNWIND reference rotor (DU331 airfoil). Figure 208 show the same comparison for the DU331 but at a value of the angle of attack of rated + 2º. The analysis is similar to Figure 15 but most of the values obtained with the BAY model are positive in this case. Figure 206: Increase of efficiency with respect to the case without VGs at rated angle of attack for the DU331 airfoil Page 100 of 141 WP no.: 3.3

101 Figure 207: Increase of efficiency with respect to the case without VGs at rated angle of attack for the FFA_W3_333 airfoil Figure 208: Increase of efficiency with respect to the case without VGs at rated +2º angle of attack for the DU331 airfoil The comparisons have showed some interesting trends. Some conclusions from the analysis are: The maximum efficiency with VGs occurs at higher angles of attack in comparison to the case without VGs. Page 101 of 141 WP no.: 3.3

102 In general, the maximum efficiency increases with the VGs height. In some cases, the cases of the highest VGs do not follow this tendency. In general, the ΔCl increases with the VGs height. In some cases, the cases of the highest VGs do not follow this tendency. Depending on the value of the angle of attack (with respect to the angle for maximum efficiency), the VGs can increase or decrease the efficiency in comparison to the airfoil without VGs. Some deviations observed in the estimated efficiency can be associated to deviations in the Cd results. The prediction of the drag coefficient is as well critical to realistically quantify the efficiency. The effect of the VGs in the Cl show similar trends with different tools. The results of Cd are not clear, and VGFlow shows different trends with respect to the BAY model. Those differences were not observed in Task 3.2. The VGs closer to the LE (25% chord) show slightly higher Cd and lower Cl than VGs placed towards the middle of the chord (30% and 40%), but they also seem to be more effective to delay separation. For the highest VGs, the codes do not show clear tendencies with the parameters. Some codes predict a behavior degradation that might be related with heights bigger than the boundary layer thickness. For that reason, it is safer to keep VGs height below 60 mm. Too small VGs originate big drag and low Cl improvement leading to low efficiencies. The best configuration seems to be with the VGs in the 40% of the chord with height close to 36mm. The best configuration with the VGs in the 30% of the chord is height 30mm Study of VGs length, angle wrt flow and internal and external distance In the previous section the optimum height and position along the chord has been studied. This section constitutes a step forward defining the best geometrical configuration of these flow control devices. The objective is to study the optimal geometrical parameters of the VGs and their sensitivity on the VGs behavior. This analysis focuses on the behavior in 2D steady inflow. However, VGs could provide behavioral improvements if used with unsteady flows or at certain points outside the rated conditions but it is not in the scope of the present study. The first consideration would be how the operation point influences the effect of the VGs. Importance of the operating conditions (angle of attack) The main effect of the VGs on the airfoil performance is a delay of the flow separation. At low angles of attack, i.e. pure linear Cl-α region, they slightly modify the behavior of the airfoils, mainly through an increase in the drag coefficient. However, as soon as the Cl slope of the clean airfoil starts to decrease, the VGs return the Cl to the potential linear behavior again. At high angles of attack, this effect increases the lift coefficient, and even decreases the drag with respect to the clean airfoil. This leads to a delay of the maximum efficiency point to a larger angle of attack. Page 102 of 141 WP no.: 3.3

103 The first step of the study is the comparison of the effect of the VGs in a wide angle of attack range (0º to 20º). As the investigated trends turn out to be independent of the design parameters, only results referring to the angle wrt flow (γ) are shown (note that TUDelft calculations are performed only for the DU331 airfoil). The fact that TUDelft calculation does not follow the tendency might be related to the small γ used. Figure 209 to Figure 220 show the comparison of Cl, Cd and efficiency between the case without VGs and the case with VGs with different angles with respect to the flow, for both airfoils and with different codes. Figure 209: Cl of the FFA_W3_333 airfoil calculated by DTU BAY model for different γ Figure 210: Cl of the FFA_W3_333 airfoil calculated by NTUA VGflow model for different γ Figure 211: Cd of the FFA_W3_333 airfoil calculated by DTU BAY model for different γ Figure 212: Cd of the FFA_W3_333 airfoil calculated by NTUA VGflow model for different γ Page 103 of 141 WP no.: 3.3

104 Figure 213: Efficiency of the FFA_W3_333 airfoil calculated by DTU BAY model for different γ Figure 214: Efficiency of the FFA_W3_333 airfoil calculated by NTUA VGflow model for different γ Figure 215: Cl of the FFA_W3_333 airfoil calculated by NTUA BAY model for different γ Figure 216: Cl of the DU331 airfoil calculated by TUDelft BAY for different γ Figure 217: Cd of the FFA_W3_333 airfoil calculated by NTUA BAY model for different γ Figure 218: Cd of the DU331 airfoil calculated by TUDelft BAY for different γ Page 104 of 141 WP no.: 3.3

105 Figure 219: Efficiency of the FFA_W3_333 airfoil calculated by NTUA BAY model for different γ Figure 220: Efficiency of the DU331 airfoil calculated by TUDelft BAY for different γ For the NTUA code VGflow, the angle at which maximum efficiency happens does not change with the parameter values (but it does change with respect to the NO-VGs configuration). On the other hand, the DTU software based on BAY model shows that the maximum efficiency could be delayed up to a certain point varying the angle wrt the flow. This also happens with the parameters D or d, but this delay has a limit that shows that further changes in the parameters do not keep on delaying the maximum efficiency point (except with the parameter l which increases the angle of maximum efficiency in all the range. In the airfoil without VGs, maximum efficiency is reached at AOA=8º. On the other hand, configurations with VGs show their maximum efficiency around AOA=10-12º. This study shows the need to include VGs already in the initial design. It will allow the blade to work at higher angles of attack making use of the full capabilities of this device. On later design phases, the final shape and size of the VGs could be refined. DU331 airfoil The DU331 airfoil is located at the 35 % span of the AVATAR rotor. At rated conditions, the airfoil operates at an angle of attack equal to 3º according to NTUA simulations. This region of operation is characterized by attached flow. According to the working mechanism of the VGs they are not expected to be fully effective in this region, however, some dependencies on design parameters can be studied at this angle of attack already. Figure 221 to Figure 224 show the variation of the Cl with the different parameters for the DU331. It is recalled that the investigated design parameters are given in Figure 1. For all the studied cases, all the computational methods show a small increase of lift coefficient independently of the parameter. Moreover, the differences of lift among the studied values of each parameter are very small, especially for the VGflow code. Page 105 of 141 WP no.: 3.3

106 Figure 221: ΔCl at α=3º of the DU331 airfoil for different γ Figure 222: ΔCl at α=3º of the DU331 airfoil for different D values Figure 223: ΔCl at α=3º of the DU331 airfoil for different d values Figure 224: ΔCl at α=3º of the DU331 airfoil for different l values Figure 225 to Figure 228 show the variation for the Cd. The drag coefficient always shows an increase when the vortex generators are placed. The DTU s BAY model shows a maximum drag increase for all the studied parameters, except from the l value. However, the NTUA s VGflow shows larger drag variations with respect to the NO-VGs case that decrease monotonously with increasing D and it shows a maximum value when the effect of the angle wrt flow is studied. Page 106 of 141 WP no.: 3.3

107 Figure 225: ΔCd at α=3º of the DU331 airfoil for different γ Figure 226: ΔCd at α=3º of the DU331 airfoil for different D values Figure 227: ΔCd at α=3º of the DU331 airfoil for different d values Figure 228: ΔCd at α=3º of the DU331 airfoil for different l values Figure 229 to Figure 232 show the variation of the efficiency. DTU s BAY model show an almost constant efficiency value in the linear zone, however, the efficiency can be bigger or smaller than the original airfoil efficiency depending on the value of the parameters. The VGflow software of NTUA predicts an important drop of the efficiency with respect to the clean case (which is related to the big Cd increase). Both of them estimate small differences between the different parameters. Page 107 of 141 WP no.: 3.3

108 Figure 229: ΔEff at α=3º of the DU331 airfoil for different γ Figure 230: ΔEff at α=3º of the DU331 airfoil for different D values Figure 231: ΔEff at α=3º of the DU331 airfoil for different d values Figure 232: ΔEff at α=3º of the DU331 airfoil for different l values The FFA_W3_333 working at an angle of attack in the linear zone show the same tendencies as the DU331 airfoil. FFA_W3_333 airfoil The FFA_W3_333 airfoil is located near 35% span of the INNWIND rotor. At rated conditions, the airfoil operates at an angle of attack equal to 8.8º according to NTUA simulations. The following analysis studies the behavior of this airfoil at this angle of attack, i.e. an operating condition that is very likely affected by flow separation in the original 2D configuration, where potential of the VGs could be expected. Figure 233 to Figure 236 show the variation of the Cl with the different parameters for the FFA_W3_333. The lift coefficient increases in all the studied cases. According to the DTU model, the point at which the lift drops happens at lower AOA in configuration with VGs, however, up to that point the configuration with VGs shows bigger lift coefficients. On the other hand the NTUA code shows a delay of the stall point which also happens at higher lift coefficients. DTU s BAY model show a limit for the maximum lift coefficient as a function of the Page 108 of 141 WP no.: 3.3

109 values of the parameters (for γ=20, D=150 and d=60). This tendency is only confirmed by the VGflow code for the FFA airfoil for the γ parameter. Figure 233: ΔCl at α=8.8º of the FFA_W3_333 airfoil for different γ Figure 234: ΔCl at α=8.8º of the FFA_W3_333 airfoil for different D values Figure 235: ΔCl at α=8.8º of the FFA_W3_333 airfoil for different d values Figure 236: ΔCl at α=8.8º of the FFA_W3_333 airfoil for different l values Figure 237 to Figure 240 show the variation of the Cd. The drag coefficient shows two different tendencies at this point for each code. While the DTU code shows a decrease in the drag coefficient at this point, the NTUA code predicts a drag increase. However, the NTUA code also shows a decrease for higher AOAs (bigger than 10º). Changing the parameters that define the VGs imply small variations with no clear tendencies. Page 109 of 141 WP no.: 3.3

110 Figure 237: ΔCd at α=8.8º of the FFA_W3_333 airfoil for different γ Figure 238: ΔCd at α=8.8º of the FFA_W3_333 airfoil for different D values Figure 239: ΔCd at α=8.8º of the FFA_W3_333 airfoil for different d values Figure 240: ΔCd at α=8.8º of the FFA_W3_333 airfoil for different l values Figure 241 to Figure 244 show the variation of efficiency. Following the same tendency previously observed, DTU predicts an efficiency increase while NTUA still show an efficiency decrease that can be turned to an increase if the AOA keeps growing. Figure 241: ΔEff at α=8.8º of the FFA_W3_333 airfoil for different γ Figure 242: ΔEff at α=8.8º of the FFA_W3_333 airfoil for different D values Page 110 of 141 WP no.: 3.3

111 Figure 243: ΔEff at α=8.8º of the FFA_W3_333 airfoil for different d values Figure 244: ΔEff at α=8.8º of the FFA_W3_333 airfoil for different l values If the DU331 airfoil operated in this region it would behave in the same way as the FFA airfoil. Conclusions The previous analysis provides some information that can be useful for the design and use of vortex generators: Based on the 2D analysis, the use of VGs should be included in the initial design phase already since the optimal operating point changes between the configuration with and without VGs In the linear zone, VGs increase the lift and the drag coefficient. These opposing effects can make the airfoil efficiency either higher or lower than the efficiency of the clean airfoil. VGs are a useful flow control device to delay the separation At high AOAs (above 10º) VGs increase the lift coefficient and decrease the drag values leading to higher airfoil efficiencies. The independency of airfoil properties on geometrical parameters are small compared to the VGs effect, i.e. the change of the airfoil with and without VGs. This happens for the ranges of values for each parameter considered in this study. Wind tunnel experiments would be interesting to check if the available codes are able to confirm the small independency of VGs behavior due to geometrical changes. The scope of this study is limited to 2D inflow conditions and 2D or quasi-3d simulations. VGs might provide an interesting behavior under other conditions such as the 3D flow studied in the following section. Page 111 of 141 WP no.: 3.3

112 KPIs l [mm] angle wrt flow [deg] d [mm] D [mm] Clmax [-] alpha Clmax slope alpha=0 Cdmin [-] alpha Cdmin Cl/Cd max [-] alpha Cl/Cd max ΔCl alpha=3º ΔCd alpha=3º DTU_BAY Δslope alpha=3º ΔCl/Cd alpha=3º NTUA_VGflow TUDelft_BAY Table 21: KPIs for the DU331 airfoil Page 112 of 141 WP no.: 3.3

113 l [mm] angle wrt flow [deg] d [mm] D [mm] Clmax [-] alpha Clmax slope alpha=0 Cdmin [-] alpha Cdmin Cl/Cd max [-] alpha Cl/Cd max ΔCl alpha=8.8º ΔCd alpha=8.8º DTU_BAY Δslope alpha=8.8º ΔCl/Cd alpha=8.8º NTUA_BAY NTUA_VGflow Table 22: KPIs for the FFA_W3_333 airfoil Page 113 of 141 WP no.: 3.3

114 6.3.3 Study of VGs influence in the 3D rotor The last step in the parametric study of VGs is the evaluation of the technology in the AVATAR and INNWIND reference wind turbines. Only different radial extensions for the VGs in the blade have been tested. The objective is to evaluate the effect in the rotor performance of the inclusion of VGs in the blade, and analyses if the trends confirm the conclusions of the 2D study. ECN has performed simulations at rated wind speed with the VGs in the 35% radius with different radial extensions. Distributed loads study Figure 245 to Figure 248 show the comparison of Cn and Ct in the root blade region without VGs and with different VGs extensions, with the Aero-Module code (BEM and AWSM models) for the AVATAR rotor. Cn varies around 3% in the VGs region, but the results of the BEM and AWSM models are similar and show an increase of Cn. The effect in the Ct is around 4% with both codes in the region where the VGs are located. Figure 245: Cn along the blade without VGs and with different VGs extensions centred in the 35%, AVATAR rotor simulation with Aero-Module (BEM) Figure 246: Cn along the blade without VGs and with different VGs extensions centred in the 35%, AVATAR rotor simulation with Aero-Module (AWSM) Page 114 of 141 WP no.: 3.3

115 Figure 247: Ct along the blade without VGs and with different VGs extensions centred in the 35%, AVATAR rotor simulation with Aero-Module (BEM) Figure 248: Ct along the blade without VGs and with different VGs extensions centred in the 35%, AVATAR rotor simulation with Aero-Module (AWSM) Figure 249 to Figure 252 show the comparison of Cn and Ct in the root blade region without VGs and with different VGs extensions, with the Aero-Module code (BEM and AWSM models) for the INNWIND rotor. The effect in the Cn is smaller than in the case of the AVATAR rotor, but again the results of the BEM and AWSM models are similar and show an increase of Cn. The effect in the Ct is around -2% in the VGs region, but in this case a slight decrease of Ct can be noticed. Figure 249: Cn along the blade without VGs and with different VGs extensions centred in the 35%, INNWIND rotor simulation with Aero-Module (BEM) Figure 250: Cn along the blade without VGs and with different VGs extensions centred in the 35%, INNWIND rotor simulation with Aero-Module (AWSM) Page 115 of 141 WP no.: 3.3

116 Figure 251: Ct along the blade without VGs and with different VGs extensions centred in the 35%, INNWIND rotor simulation with Aero-Module (BEM) Figure 252: Ct along the blade without VGs and with different VGs extensions centred in the 35%, INNWIND rotor simulation with Aero-Module (AWSM) Figure 253 to Figure 256 show the comparison of Fx and Fy in the root blade region without VGs and with different VGs extensions, with the Aero-Module code (BEM and AWSM models) for the AVATAR rotor. Fx increases around 2% in the VGs region. In this case, the effect in Fy is an increase around 3%. Figure 253: Fx along the blade without VGs and with different VGs extensions centred in the 35%, AVATAR rotor simulation with Aero-Module (BEM) Figure 254: Fx along the blade without VGs and with different VGs extensions centred in the 35%, AVATAR rotor simulation with Aero-Module (AWSM) Page 116 of 141 WP no.: 3.3

117 Figure 255: Fy along the blade without VGs and with different VGs extensions centred in the 35%, AVATAR rotor simulation with Aero-Module (BEM) Figure 256: Fy along the blade without VGs and with different VGs extensions centred in the 35%, AVATAR rotor simulation with Aero-Module (AWSM) Figure 257 to Figure 260 show the comparison of Fx and Fy in the root blade region without VGs and with different VGs extensions, with the Aero-Module code (BEM and AWSM models) for the INNWIND rotor. The effect in the Fx is around 0.4% (smaller than in the case of the AVATAR rotor), as it was also observed for the Cn. The effect in the Fy is around 2% in the VGs region which is smaller than in the AVATAR rotor, but still a slight increase is observed. Figure 257: Fx along the blade without VGs and with different VGs extensions centred in the 35%, INNWIND rotor simulation with Aero-Module (BEM) Figure 258: Fx along the blade without VGs and with different VGs extensions centred in the 35%, INNWIND rotor simulation with Aero-Module (AWSM) Page 117 of 141 WP no.: 3.3

118 Figure 259: Fy along the blade without VGs and with different VGs extensions centred in the 35%, INNWIND rotor simulation with Aero-Module (BEM) Figure 260: Fy along the blade without VGs and with different VGs extensions centred in the 35%, INNWIND rotor simulation with Aero-Module (AWSM) Table 23 and Table 24 show a summary of the loads present in the blade at the middle of the VGs position (r/r=0.35). They group the differences with respect to the case without VGs for all the studied cases. Code position (r/r) extension ΔCn[%] ΔCt[%] ΔFx[%] ΔFy[%] AWSM BEM Table 23: Differences with respect to the case without VGs for the ECN contribution. AVATAR rotor Code position (r/r) extension ΔCn[%] ΔCt[%] ΔFx[%] ΔFy[%] AWSM BEM Table 24: Differences with respect to the case without VGs for the ECN contribution. INNWIND rotor Page 118 of 141 WP no.: 3.3

119 Power and thrust study Figure 261 to Figure 264 show the % of variation of thrust and power including VGs in the AVATAR and INNWIND blade. For the AVATAR case, the torque and thrust are increased in all the cases and extending the VGs region the thrust and power and increased. Both models give similar trends with a very good agreement. The maximum increase in power is under the 0.35% with almost a 0.5% increase in thrust. For the INNWIND rotor, the increase in thrust and power obtained with the AWSM model is significantly lower than for the AVATAR rotor. For the BEM model, the values are negative, but the reason is an increase of the loads in a region outside the VGs area towards the root. In the region of the VGs the contribution to power and thrust is similar to the AWSM simulation. Figure 261: % Change of thrust including VGs, AVATAR rotor simulation with Aero-Module Figure 262: % Change of power including VGs, AVATAR rotor simulation with Aero-Module Figure 263: % Change of thrust including VGs, INNWIND rotor simulation with Aero-Module Figure 264: % Change of power including VGs, INNWIND rotor simulation with Aero-Module Influence of rotational effects Placing VGs at inboard sections makes it necessary to also include rotational effects. Rotation will enhance the performance of the section in case separation is onset. This implies that Page 119 of 141 WP no.: 3.3

120 rotational effects will compete with VGs in supressing the effects of separation and therefore the design of VGs may be different than that based on 2D simulations In this section a limited number of simulations is presented in an attempt to see how rotation may affect the behaviour of VGS. First results are presented for a section placed at r/r=0.35. Four simulations are considered: a pure 2D simulation without VGs, a simulation with VGs (BAY model), a quasi 3D (2.5D) simulation without VGs and a quasi 3D simulations with VGs (2.5D+vg(BAY). All simulations have been performed with MaPFlow over a slice with spanwise width equal to one half period of the VG layout. Quasi 3D simulations refer to the set-up where the section is placed at its correct position with respect to the blade geometry and the effect of rotation is included. On the side boundaries of the slice modified periodic conditions are applied which differ from the ordinary ones by considering that the total pressure is conserved. Figure 265 and Figure 266 show the pressure and friction coefficient distribution along the chord for all four cases. It is observed that the VG effect is much smaller for the 2.5D case. Due to the spanwise component of the flow, the VG faces the local flow at a reduced angle of attack. This raises issues about the optimal positioning of VGs as well as about the counter or co-rotating set up. The effect of rotation is also noted as the separation on the suction side is already suppressed. Finally, separation for this very thick section is observed also on the pressure side, which may suggest that flow control devices could be also considered for this part of the airfoil. Figure 265: Cp distribution for the four cases, MaPFlow 2D, MaPFlow + BAY model, MaPFlow 2.5D, MaPFlow 2.5D + BAY model Page 120 of 141 WP no.: 3.3

121 Figure 266: Cf distribution for the four cases, MaPFlow 2D, MaPFlow + BAY model, MaPFlow 2.5D, MaPFlow 2.5D + BAY model Figure 267 shows contours of the radial velocity for the 2.5D cases, while in Figure 268 contours of streamwise vorticity component are added on slices normal to the x axis. In both cases there is radial flow developed. The radial velocity is increased when the VG is used, especially close to the TE. This implies that the local flow conditions, at which the VGs are operating, correspond to a different local flow angle and therefore their performance is different from that seen in the 2D case. The effect of VGs on the vorticity contours is also clearly seen. Figure 267: 2.5D cases at 35% and at an angle of attack of 3.7º. (Left) 2.5D case, (Right) 2.5D+VG(BAY). Radial velocity contours on a plane normal to the Y axis Page 121 of 141 WP no.: 3.3

122 Figure 268: 2.5D cases at 35% and at an angle of attack of 3.7º. (Left) 2.5D case, (Right) 2.5D+VG(BAY). Radial velocity contours on a plane normal to the Y axis and X-vorticity contours on a plane normal to the X axis In an attempt to also see how rotation affects the aerodynamic performance at higher angles of attack than the normal setting of 3.7 o, simulations with VGFlow have been carried out for the same section. The higher angles of attack were obtained by pitching the section while maintaining the rotor speed constant. In Figure 269 pressure and friction distributions are compared assuming purely 2D without the presence of VGs and 2D and 2.5D flow conditions with the presence of VGs. The presence of the VGs gets important at higher angles. The VG effect appears to be smaller when rotational effects are included. Cp at 3.7 o angle of attack Cf at 3.7 o angle of attack Page 122 of 141 WP no.: 3.3