Master of Science Thesis
Particle System Modelling and Dynamic Simulation of a Tethered Rigid Wing Kite for Power Generation
Mustafa Can Karadayı
September 21, 2016
Faculty of Aerospace Engineering Delft University of Technology ·
Particle System Modelling and Dynamic Simulation of a Tethered Rigid Wing Kite for Power Generation
Master of Science Thesis
For obtaining the degree of Master of Science in Aerospace Engineering at Delft University of Technology
Mustafa Can Karadayı
September 21, 2016
Thesis Registration Number: 088#16#MT#FPP
An electronic version of this thesis is available at http://repository.tudelft.nl
Faculty of Aerospace Engineering Delft University of Technology · Copyright c Mustafa Can Karadayı
All rights reserved. Delft University Of Technology Department Of Flight Performance and Propulsion
The undersigned hereby certify that they have read and recommend to the Faculty of Aerospace Engineering for acceptance a thesis entitled “Particle System Modelling and Dynamic Simulation of a Tethered Rigid Wing Kite for Power Gener- ation” by Mustafa Can Karadayı in partial fulfillment of the requirements for the degree of Master of Science.
Dated: September 21, 2016
Chairperson of Committee: Dr.-Ing. Roland Schmehl
Reader: Dr. Axelle C. Vir´e
Reader: Dr.ir. Mark Voskuijl
Daily Supervisor: Uwe Fechner, MSc
Abstract
Kite Power research group at TU Delft developed the KiteSim framework for dynamic simulation of crosswind kite power systems. Currently, KiteSim has the capability to simulate leading edge inflatable kites. The use of rigid wing kites in crosswind kite power systems is becoming widespread. Accordingly, KiteSim framework is planned to be augmented to simulate the rigid wing kite power systems as well. Therefore, objective of this thesis is set to enhance the KiteSim framework in order to include the capability of the dynamic simulation of a rigid wing pumping crosswind kite power system by developing a particle system model to represent the tethered rigid wing. The particle system model of rigid wing consists of 6 point masses and 13 spring-damper elements. Positions and masses of the discrete particles are calculated in order to represent the rigid wing properties accurately. The spring-damper elements are interconnecting the particles. Lifting line theory and Kirke’s post stall correlation methods are used to create the full angle of attack aerodynamic model for wings and tails. Available atmospheric, tether and winch models of the KiteSim framework are also used for rigid wing dynamic simulations. Equation of the motion of the particle system model is formulated as an implicit problem, which is simulated by implicit Runge-Kutta method of fifth-order. Validation simulations are conducted with the particle system model of NASA SGS 1-36 flight test sailplane. Validation cases show satisfactory results with the flight test data. For the remaining simulations, AP2 PowerPlane of Ampyx Power is modelled as particle system and manually flown in KiteSim. Mass properties of the model is found to be highly accurate throughout the simulations. Power generation capabilities of the model is checked by flying figure-of-eight trajectories. Reel-out phases are investigated, a peak mechanical power of 45 kW and a mean power of 10 kW are obtained. Further comparison of the reel-out dynamic simulations with the quasi-steady theoretical calculations is done. Moreover, gliding and stalling manoeuvres are simulated for a plausibility check study. The developed particle system model for rigid wings shows satisfactory results and a potential for further development.
v vi Abstract Acknowledgements
My enthusiasm for airborne wind energy was born when I started taking kite power lectures back in 2013 from Dr.-Ing. Roland Schmehl. I want to thank him for introducing me to this fascinating topic and leading me to choose this thesis project. I would like to thank my daily supervisor Uwe Fechner, MSc. who has been always ready to help me with my countless questions. I want to thank Ampyx Power for sharing the data of their previous prototype with me. I am also thankful to Ted Spinders for the help and feedback on piloting the simulation. I am grateful to all past and present members of Kite Power team and Room 6:08 who have been great mates not only in office but also outside. Thank you Tassos, Lukas, Jo, Felix, Chris, Pietro, Lorenzo, Julius, Mike, Apurva, Rachel, Andres, Bas, Viktor, Pranav. I want to thank my friends from FPP; Ashwin, Phillip, Martin, Akshay, Malcom and Roderick for being such a great company. In particular, I feel so happy to meet my companion Dino; our endless talks from science to politics shall never end. Thanks to all Happy Hour People, for sharing the life in Delft since the first day. Special thanks to the great panpas of G¨ulizAbla who became my family in Delft: Sercan, for pre-/peri-/post-fitness adventures; Emre, for interesting topics on the mops; Yasemin, for great midnight dessert sessions; Argun, for good coffee and hazelnut taste; Onursal, for interesting philosophical conversations and even more interesting laptop wars; Alper, for random tricky dj skills; Cansın, for tasty sushi workshops; Burak, for famous one- hand-capable-water-skiing tricks; Ekin, for overnights and broodjes memories; Tilbe, for hilarious countryside trips; and Cansel, for good cheer and ’But, I!’ moments. Last but not the least, I am sending my greatest thanks to my parents for their trust and support during my whole life.
Delft, The Netherlands Mustafa Can Karadayı September 21, 2016
vii viii Acknowledgements Contents
Abstract v
Acknowledgements vii
List of Figures xv
List of Tables xvii
Nomenclature xix
1 Introduction1
2 Crosswind Kite Power Fundamentals3 2.1 Physics of Crosswind Kite Power Systems...... 3 2.1.1 Power formula...... 3 2.1.2 Power harvesting factor...... 4 2.2 Classification of Crosswind Kite Power Systems...... 4 2.2.1 Power extraction mode classification...... 4 2.2.2 Flexible wing vs. rigid wing classification...... 5 2.2.3 Other classifications...... 6 2.3 Working Principles of Crosswind Kite Power Systems...... 6
3 Literature Review9 3.1 Modelling and Simulation...... 9 3.2 Crosswind Kite Modelling...... 10 3.3 TU Delft Dynamic Kite Power System Simulator...... 12 3.3.1 Atmospheric model...... 12 3.3.2 Tether model...... 13 3.3.3 Kite models...... 13
ix x Contents
3.3.4 Winch model...... 14 3.3.5 Implementation...... 14 3.4 Particle System Modelling...... 15 3.5 Component Buildup Method...... 16 3.6 Aerodynamic Models for Dynamic Simulation...... 18
4 Research Objective 21
5 Methodology of the Simulation Modelling 23 5.1 Rigid Wing Particle System Model Structure...... 23 5.1.1 Discretization of the particles...... 24 5.1.2 Formation of the spring-damper elements...... 29 5.2 Rigid Wing Aerodynamic Model...... 29 5.2.1 XFLR5 software...... 31 5.2.2 Post-stall correlation method...... 32 5.2.3 Aerodynamic data approximation script...... 34 5.3 Rigid Wing Controller Implementation...... 34 5.4 Atmospheric Model...... 35 5.5 Tether Particle System Model...... 36 5.6 Rigid Wing Particle System Model...... 39 5.6.1 Calculation of the gravitational forces...... 40 5.6.2 Calculation of the spring-damper forces...... 40 5.6.3 Calculation of the aerodynamic forces...... 41 5.7 Equation of Motion of the Particle System Model...... 46 5.8 Formulation of the Implicit Problem...... 46 5.9 Simulation Implementation...... 47 5.9.1 Numeric solver...... 47 5.9.2 Initial and simulation parameters...... 48
6 Model Validation 49 6.1 Validation Aircraft...... 50 6.2 Validation Methodology...... 50 6.2.1 Simulation of non-tethered flight in KiteSim...... 50 6.2.2 Fuselage drag estimation...... 50 6.3 Validation Results and Discussions...... 51 Contents xi
7 Simulation Results and Discussions 57 7.1 Mass Properties Results...... 57 7.1.1 Case 1: Tethered, non-maneuvering kite at very high atmospheric wind speed...... 57 7.1.2 Case 2: Tethered, maneuvering kite at high atmospheric wind speed 60 7.2 Reel-Out Phase Results...... 63 7.2.1 Case 3: Tethered, reel-out, figure-of-eight trajectory flight at strong wind...... 63 7.2.2 Case 4: Tethered, reel-out, figure-of-eight trajectory flight at mod- erate wind...... 67 7.2.3 Comparison of reel-out cases results and theoretical analysis.... 70 7.3 Plausibility Checking Results...... 71 7.3.1 Case 5: Non-tethered, gliding flight...... 71 7.3.2 Case 6: Non-tethered, stalling maneuver...... 73 7.4 Discussions...... 75 7.4.1 Discussion on the developed particle system model...... 75 7.4.2 Discussion on the controlling and piloting...... 76 7.4.3 Discussion on the mass properties...... 76 7.4.4 Discussion on the power generation...... 77
8 Conclusions and Recommendations 79
References 81
A Input File for the Initialization Script of AP2 PowerPlane 85
B XFLR5 Models for Wing, HT and VT of AP2 PowerPlane 87
C Correction Factor Plots for Post-Stall Aerodynamic Coefficients 89
D Approximate AeroData RW Script 91
E Wing Axes System Creation for Wing Dihedral Effects 95
F Schweizer SGS 1-36 Sailplane 97 F.1 Input file for SGS 1-36 Sailplane...... 97 F.2 General Data for SGS 1-36 Sailplane...... 99 F.3 XFLR5 models for wing, HT and VT of SGS 1-36 Sailplane...... 103
G Winch Model and Winch Controller Parameters 105 G.1 Winch Model Parameters...... 105 G.2 Winch Controller Parameters...... 106
H Additional Figures and Data for Quasi-Steady Analysis 107
I Additional Simulation Results 111 xii Contents List of Figures
2.1 Examples of flexible, hybrid and rigid wing kite designs...... 5 2.2 Different ground based CWKP steering and control concepts...... 6 2.3 Traction and retraction phases of the pumping cycle operation...... 7 2.4 3D representation of the figure-of-eight flight pattern...... 7
3.1 Complexity vs computation time for different kite modelling approaches. 10 3.2 Description of point mass, rigid body, particle system (lumped mass) and multi plate models...... 11 3.3 Representation of the rigid body model of the kite and the particle system model of the tether...... 12 3.4 Particle system representation of the tether (points P0 to PKCU) and the four point kite (points A to D)...... 13 3.5 Representation of the point mass and four-point model...... 14 3.6 Representation of a particle and the systems of particles formed by the particles along with local axis frames...... 15 3.7 Representation of the external forces that cause the translation (left) and the rotation (right) on a rigid body...... 16 3.8 The discretization of the kiteplane aerodynamic surfaces...... 17 3.9 Illustration of lift and drag forces on the discrete elements of the kiteplane 17 3.10 Lift and drag coefficients versus angle of attack curves for LEI kites in KiteSim...... 18
5.1 Representation of the PS model on a conventional RW kite, with 6 discrete point masses and 13 spring-damper elements...... 24 5.2 RW kite particles of the PS model for the AP2 RW kite...... 28 5.3 The particles and the spring-damper elements of the PS model, a generic RW kite is drawn for comparison...... 30 5.4 3D aerodynamic coefficients versus AoA obtained from XFLR5 software. 33 5.5 Aerodynamic coefficients as a function of AoA...... 35
xiii xiv List of Figures
5.6 Power law, logarithmic law and fitted wind profiles along with three mea- sured wind speeds at different altitudes...... 36
5.7 Discrete tether particles (points tp0 to tpn) of the tether model...... 37 5.8 Representation of the position and velocity vectors along with the segment vector and segment relative velocity for a segment...... 38 5.9 Representation of the position and velocity vectors along with the segment vector and segment relative velocity for a segment...... 40 5.10 Representation of the RW particles with the lift and drag forces acting on the particles m2, m3, m4 and m5 ...... 41 5.11 Apparent wind velocity and its components for the calculation of angle of attack and sideslip angle...... 43 5.12 Equivalent flat plate method used in the calculation of the control surface deflection effect on the AoAs...... 43 5.13 The aerodynamic moments and aerodynamic moment induced forces ex- plained on the PS model of the RW kite...... 45
6.1 Modified Schweizer SGS 1-36 sailplane used by NASA...... 49 6.2 L/D versus airspeed plot obtained from the flight tests and the KiteSim. 53 6.3 Sink rate versus horizontal velocity plot obtained from the flight tests and the KiteSim...... 53 6.4 Rudder input and directional time history...... 55 6.5 Aileron input and lateral time history...... 55 6.6 Elevator input and longitudinal time history...... 56
7.1 Flight trajectory; case 1...... 58 7.2 Apparent wind speed versus time; case 1...... 58 7.3 Reel-out speed versus time; case 1...... 58 7.4 PS model mass properties versus time; case 1...... 59 7.5 Flight trajectory; case 2...... 61 7.6 Apparent wind speed versus time; case 2...... 61 7.7 Reel-out speed versus time; case 2...... 61 7.8 PS model mass properties versus time; case 2...... 62 7.9 Flight trajectory; case 3...... 63 7.10 Time history; case 3...... 65 7.11 Simulation results for the power related parameters; case 3...... 66 7.12 Flight trajectory; case 4...... 67 7.13 Time history; case 4...... 68 7.14 Simulation results for the power related parameters; case 4...... 69 7.15 Flight trajectory; case 5...... 71 7.16 Time history for gliding phase; case 5...... 72 7.17 Flight trajectory; case 6...... 73 7.18 Time history for stall maneuver; case 6...... 74 List of Figures xv
B.1 AP2 PowerPlane Wing...... 87 B.2 AP2 PowerPlane HT...... 88 B.3 AP2 PowerPlane VT...... 88
F.1 SGS 1-36 general data, from [33]...... 99 F.2 SGS 1-36 mass properties, from [33]...... 99 F.3 SGS 1-36 general data, from [19]...... 100 F.4 SGS 1-36 general data (continued), from [19]...... 100 F.5 SGS 1-36 mass properties, from [19]...... 100 F.6 SGS 1-36 three-view drawing , from [33]...... 101 F.7 SGS 1-36 side drawing , from [6]...... 102 F.8 Schweizer SGS 1-36 Wing...... 103 F.9 Schweizer SGS 1-36 HT...... 103 F.10 Schweizer SGS 1-36 VT...... 103
H.1 Representation of kite velocities, with azimuth angle φazi, polar angle θele and course angle χ shown; from [29]...... 107 H.2 Representation of aerodynamic forces and velocities, with the azimuth an- gle φazi and polar angle θele shown; from [29]...... 108 H.3 Representation of aerodynamic forces, gravity force and velocities, with the azimuth angle φazi and polar angle θele shown; from [29]...... 108
I.1 Kite lift and kite weight during the simulation case 1...... 111 I.2 Kite lift and kite weight during the simulation case 2...... 111 xvi List of Figures List of Tables
5.1 Initial guesses, lower bounds and upper bounds of the optimization vari- ables in the optimization problem...... 27 5.2 Basin Hopping algorithm parameters used in the optimization...... 27 5.3 The masses and positions of the AP2 RW kite particles in PS model... 27 5.4 Comparison of the mass properties and the cg location of the AP2 RW kite and its PS model...... 28 5.5 XFoil Direct Analysis parameters for the AP2 RW kite...... 31 5.6 Wing & Plane Design tool advanced settings for wing, HT and VT of the AP2 RW kite...... 32 5.7 Control Surface Sign Convention...... 34 5.8 Tether spring stiffness in tension and compression...... 39 5.9 Radau5DAE solver parameters used in the simulation...... 48 5.10 Initial and simulation parameters...... 48
6.1 SGS 1-36 Sailplane fuselage drag calculation parameters...... 51
H.1 Averaged values of the system parameters during the approximate hori- zontal flight state, along with theoretical traction force and power..... 109 H.2 Tether force and traction power comparison of the simulation cases 3 and 4 with the theoretical analysis...... 110
xvii xviii List of Tables Nomenclature
Latin Symbols
2 Aside Fuselage Side Projected Area [m ] 2 Atop Fuselage Top Projected Area [m ] AR Aspect Ratio [-] b Span [m] BL Buttock Line [m] c¯ Mean Aerodynamic Chord [m] c Chord Length [m] ccs Chord Length of the Control Surface [m]
CD,t,eq Tether Equivalent Drag Coefficient [-]
CDt Tether Drag Coefficient [-]
CD Drag Coefficient(3D) [-]
Cd Drag Coefficient(2D) [-] ceff Effective Chord Length [m]
Cf,t Turbulent Flat-plate Skin Friction Drag Coefficient [-]
CL Lift Coefficient(3D) [-]
Cl Lift Coefficient(2D) [-]
CM Quarter Chord Moment Coefficient(3D) [-]
Cm Quarter Chord Moment Coefficient(2D) [-] croot Root Chord [m] df Fuselage Maximum Diameter [m] dt Tether Diameter [m]
xix xx Nomenclature
FB Body-Fixed Reference Frame [-]
FW Wind Reference Frame [-] FF Fuselage Form Factor [-] FS Fuselage Station [m] iHT Horizontal Tail Incidence Angle [rad] or [deg] iw Wing Incidence Angle [rad] or [deg] 2 Ixx Moment of Inertia about xB axis [kgm ] 2 Ixy Product of Inertia with respect to xB and yB axes [kgm ] 2 Ixz Product of Inertia with respect to xB and zB axes [kgm ] 2 Iyy Moment of Inertia about yB axis [kgm ] 2 Iyz Product of Inertia with respect to yB and zB axes [kgm ] 2 Izz Moment of Inertia about zB axis [kgm ] −1 kd Damping Coefficient [Nsm ] −1 ks Spring Constant [Nm ] lf Fuselage Length [m] ls0 Initial Segment Length [m] ls Length of Tether Spring-Damper Segment [m] lt Tether Length [m] M Aerodynamic Moment [Nm] th mi Mass of i Particle [kg] mkite Kite Mass [kg] mtp Mass of a Tether Particle [kg] N Number of Particles of the PS Model [-] n Number of Tether Spring-Damper Elements [-] S Reference Area [m2] 2 Swet,f Fuselage Wetted Area [m ] t Time [s] −1 va Apparent Wind Velocity [ms ] −1 vt,ro Tether Reel-Out Speed [ms ] −1 vw,exp Wind Speed at Altitude According to Power Law [ms ] −1 vw,log Wind Speed at Altitude According to Log Law [ms ] −1 vw,ref Reference Wind Speed [ms ] WL Water Line [m] th xi, yi, zi x, y, z Coordinates of the i Particle in FB [m] xB, yB, zB x, y, z Axes of Body-Fixed Reference Frame FB [-] xcg, ycg, zcg x, y, z Coordinates of the cg Position in FB [m] xW, yW, zW x, y, z Axes of Wind Reference Frame [-] Y¯ Spanwise Distance of the Mean Aerodynamic Chord [m] Nomenclature xxi
z Altitude [m] z0 Surface Roughness Length [m] zref Reference Altitude [m] p, q, r Roll, Pitch, Yaw Body Rates [deg s−1]
Vector Symbols
D Drag [N]
Df Fuselage Drag [N] exB , eyB , eyB Unit Vectors of Body-Fixed Reference Frame [-] Fd Tether Aerodynamic Drag Force [N]
Fg Gravitational Force [N]
Fsd Spring Damper Force [N] g Gravitational Acceleration [ms−2] L Lift [N] p Position [m] s Segment Vector [m] −1 sv Segment Relative Velocity [ms ] v Velocity [ms−1] −1 va Apparent Wind Velocity [ms ] −1 va,xz Apparent Wind Velocity Projected on XZ Plane [ms ] −1 va,xy Apparent Wind Velocity Projected on XY Plane [ms ] −1 vs Segment Velocity [ms ] Y State Vector [-] Y˙ State Derivative Vector [-]
Greek Symbols
α Angle of Attack [rad] or [deg]
αcs Additional Angle of Attack due to Control Surface Deflection [rad] or [deg]
αeff Effective Angle of Attack [rad] or [deg] β Sideslip Angle [rad] or [deg]
δcs Control Surface Deflection Angle [rad] or [deg] Γ Dihedral Angle [rad] or [deg]
κd Finite AR Correction Factor for Post-Stall Drag Coefficient [-]
κl Finite AR Correction Factor for Post-Stall Lift Coefficient [-] λ Taper Ratio [-] xxii Nomenclature
ΛLE Leading Edge Sweep Angle [rad] or [deg] ν Kinematic Viscosity [m2s−1] Ψ, Θ, Φ Tait-Bryan Roll, Pitch, Yaw Angles [deg] ρ Air Density [kg m−3] −3 ρt Tether Mass Density [kg m ] ξ Airfoil Camber [% of the chord] ζ Power Harvesting Factor [-]
Abbreviations ac Aerodynamic Center cg Center of Gravity 2D Two Dimensional 3D Three Dimensional AoA Angle of Attack AWE Airborne Wind Energy CFD Computational Fluid Dynamics CWKP Crosswind Kite Power DOF Degree(s) of Freedom EOM Equation(s) of Motion FoE Figure-of-Eight HAWP High Altitude Wind Power HT Horizontal Tail KiteSim Kite Power System Simulator L/D Lift-to-Drag Ratio LE Leading Edge LEI Leading Edge Inflatable LLT Lifting Line Theory PS Particle System RW Rigid Wing TU Delft Delft University of Technology UAV Unmanned Aerial Vehicle VLM Vortex Lattice Method VT Vertical Tail Chapter 1
Introduction
Our world is going through a change, the share of the renewable energy technologies in the energy generation is increasing. Airborne Wind Energy (AWE) is a recent and state- of-the-art concept in the field of the renewable energy technologies. The humankind had invented the kites in ancient times and flown them since then for various reasons [40]. However, the concept of using AWE for the electricity generation needed to wait until the 20th century, where the general investigations were made in the 1930s and the more detailed attempts to produce electricity by the airborne windmills were taken place in the 1960s [20]. High Altitude Wind Power (HAWP) and Crosswind Kite Power (CWKP) constitute two main concepts of the AWE systems. The content of this thesis concerns the latter. In 1980, Miles L. Loyd published his famous paper ”Crosswind Kite Power” [18], which is considered as the scientific foundation of the CWKP systems. However, the research and the development of the CWKP systems stayed inactive until the late 1990s. The improvements in tether and wing materials along with automatic control and navigation systems led to the development of the first CWKP concepts in 2000s, where different solutions gave birth to the different concepts in the CWKP systems [7]. Delft University of Technology (TU Delft) has started the research on the AWE systems with the visionary work of Professor Wubbo Ockels. Ockels’ solution was the Laddermill concept which was developed in the late 1990s. The concept includes a series of multiple kites which are linearly connected by a cable that drives a ground based generator [22]. Since then, TU Delft KitePower Research Group has been continuing to the research and development of the CWKP systems and currently has a system demonstrator where a tethered Leading Edge Inflatable (LEI) kite is used in a pumping mode to generate electricity using a ground generator [30]. Dynamic simulations are essential during the development process of the complicated systems such as a CWKP system. Realistic simulations help to test the system without requiring to manufacture different kite configurations and conducting the flight tests. Modelling of the system components and the environment is necessary for dynamic sim- ulations. The modelling of kite, tether, generator and atmosphere is needed in order to
1 2 Introduction
simulate a CWKP system. The TU Delft KitePower Research Group currently has a simulation framework which is able to simulate the current pumping CWKP system demonstrator which has a LEI kite [10]. The LEI kites are represented either as a point mass model or as a particle system model in the framework. The current simulation is lacking a dynamic model to study Rigid Wing (RW) kites. Therefore, the purpose of this thesis is to develop a particle system model for a tethered RW kite for enhancing the current framework. By achieving this, the KitePower Research Group will have the ability to simulate RW kites which will be helpful for the development of the future CWKP systems. This thesis report is structured as followed. The Crosswind Kite Power Fundamentals chapter gives a background information about the CWKP technology aiming to give the reader an understanding on the physics behind the technology, the classification of the CWKP systems and the working principles of the pumping cycle operation. The Litera- ture Review chapter presents a literature review on the modelling and simulation, cross- wind kite modelling, KiteSim framework, particle system modelling, component buildup method and aerodynamic models for dynamic simulation. The Research Objective chap- ter gives the objective and the goals of the thesis. In the Methodology of the Simulation Modelling chapter, the methodology of the numeric simulation model and the simulation are explained. The Model Validation chapter presents the validation study of the par- ticle system model for rigid wing kite. The Simulation Results and Discussions chapter presents and discusses the dynamic simulation results of the several test cases conducted with the developed particle system model. Finally, in the Conclusions and Recommenda- tions chapter, the conclusion of the thesis is presented along with the recommendations for future developments. Chapter 2
Crosswind Kite Power Fundamentals
Anyone, who flew a kite before knows the feeling that the kite is pulling stronger while flying in a crosswind direction rather than while staying static in the air. That is the very basis of the Crosswind Kite Power, which is the term that Miles Loyd created when describing and studying the idea of power generation with tethered wings in a crosswind flight [18].
2.1 Physics of Crosswind Kite Power Systems
Using the ambient wind flow, a kite can be flown in a crosswind direction with an apparent wind velocity va, which is way higher than the ambient wind velocity vw. Hence, in crosswind flight, a kite experiences higher apparent wind velocity when compared to the static flight. A crosswind flying kite extracts a significant amount of power from the wind and this can be used for power generation [7], using either directly the high apparent wind velocity or its result in high aerodynamic forces and therefore high tether tension.
2.1.1 Power formula
Using idealized assumptions, Loyd [18] estimates the maximum power that a crosswind kite can extract from the wind as:
2 2 3 CL Pmax = ρ vw SCL (2.1) 27 CD
It is seen that the maximum power Pmax is proportional to the ambient air density ρ and ambient wind speed vw which are atmospheric parameters. Also, it is dependant on kite parameters which are kite area S, as well as kite lift and drag coefficients CL, CD. The 3 2 ratio CL/CD is an important kite parameter that effects the performance of a CWKP system considerably.
3 4 Crosswind Kite Power Fundamentals
Equation 2.1 neglects the kite weight and the tether effects. Moreover, it assumes a steady-state flight for the kite at zero elevation angle and at a position which is directly crosswind that is at zero azimuth angle.
2.1.2 Power harvesting factor
The ratio of the useful power P that is extracted from the wind by a crosswind kite of a surface S, to the wind power Pwind which is flowing through the cross sectional area of the same size with S is named as power harvesting factor ζ. It is important to note that the power harvesting factor is not an energy conversion efficiency; rather it is an performance parameter.
P P ζ = = 1 3 (2.2) Pwind 2 ρ vw S
Using Equation 2.1, the maximum power harvesting factor is obtained as in Equation 2.3. Diehl [7] states that the power harvesting factors of modern CWKP systems are around 5.5 and the existing highest experimentally obtained value is 8. Moreover, the rigid wings have higher power harvesting factor compared to flexible wings as a result of their higher lift-to-drag ratios, as discussed in Section 2.2.
2 4 CL ζmax = CL (2.3) 27 CD
2.2 Classification of Crosswind Kite Power Systems
The classification of the CWKP systems is not solely dependant on one concept. There is a large variety of interesting concepts in the area of CWKP systems. The most common classifications of the CWKP systems will be given here as a background information, the more detailed ones can be obtained from [5] and[7].
2.2.1 Power extraction mode classification
In CWKP systems, the power of the kite can be extracted using two different power extraction modes that Loyd describes as drag mode and lift mode [18]. The drag mode makes use of the high apparent wind velocity to drive an on-board wind turbine and the lift mode makes use of the tension in the tether either to drive a ground generator or to propel a vehicle on the ground. With this division of the lift mode, Diehl states that the CWKP systems can be divided in three groups regarding their means of power production [7]; on-board power generation systems, ground based power generation systems and vehicle propulsion systems. On-board power generation systems carry on-board turbines to generate electricity di- rectly from the high apparent wind velocity on the kite. Payne and McCutchen, described the idea of electricity generation with the on-board generators on a tethered sailplane in 2.2 Classification of Crosswind Kite Power Systems 5
their patent ”Self-Erecting Windmill” [23] in 1976. Right now, 40 years later, Makani Power [39] is working on this concept. Ground based power generation systems have a ground station where a drum and an electric generator is present. The tether tension that is generated by the kite is used to unroll the tether from the drum. Hence, the drum rotates and drives the generator. This concept is wide spread in the field of CWKP, for example, Ampyx Power, TwingTec and EnerKite are using this concept in their solutions [5]. Vehicle propulsive systems are not electricity generating systems. They make use of the tether tension directly to propel a vehicle on the ground. SkySails [11] is a company that is using this concept.
(a) TU Delft LEI kite de- (b) TU Delft Kiteplane (c) Makani Power design sign design from [36]
Figure 2.1: Examples of flexible, hybrid and rigid wing kite designs
2.2.2 Flexible wing vs. rigid wing classification
Diehl [7] gives another classification in the CWKP systems which is between the flexible wings like LEI kites or ram-air kites and the rigid wing like sailplanes or aircrafts. Flexible wings are lighter kites due to their structures; they maintain their shapes using the airflow around them. They do not fly at very high speeds; thus, they are easily controllable by pilots. All these aspects also result in the fact that the flexible kites are safer for the operations near humans. Rigid wings, contrary to flexible wings, have heavier structures, which is due to the fact that they maintain their shapes without the need of the airflow around them. The rigid structure allows the use of more efficient profiles in the design, resulting in higher lift-to- drag ratios; thus, can reach higher speeds compared to the flexible wings. These aspects make the rigid wings more useful for power production; however, also more dangerous in case of a crash. There also exist some kites that have hybrid designs, making use of both flexible and rigid elements in their structures in order to have a design which can have the benefits of the flexible and rigid wing designs. 6 Crosswind Kite Power Fundamentals
2.2.3 Other classifications
There exist more classifications that can be made for CWKP systems. Diehl describes the use of multiple wing system in [7]. Although this concept is not built yet, the concepts can be classified considering the number of kites of the system. Cherubini classifies the ground based CWKP systems with respect to their means of steering and control [5]; such as systems which are using on-board actuators or airborne control pods or multiple tether lines as seen in Figure 2.2.
Figure 2.2: Different ground based CWKP steering and control concepts, from [5]
2.3 Working Principles of Crosswind Kite Power Systems
As mentioned in the classification of CWKP systems section, there are on-board power generation, ground based power generation and vehicle propulsion systems. The TU Delft KitePower research group is currently using a ground based power generation system with pumping cycle operation. Therefore, in this section, pumping cycle power generation is explained. The operation of the pumping cycle systems do share some common principles with the other CWKP systems. However, there exists differences, for which Diehl gives the necessary explanation in [7]. The main components in a CWKP system using pumping cycle operation are ground station, tether and kite. The pumping cycle operation can separated into two phases; namely traction (reel-out) and retraction (reel-in) phases as shown in Figure 2.3. The traction phase is where the kite is flying a crosswind maneuver such as a figure-of- eight (FoE) pattern. Due to the crosswind flight, the high aerodynamic load generated by the kite increases the tether tension and the tether reels-out from the drum in the ground station and rotates it. Consequently, the drum drives the generator. Once the maximum design tether length is reached the traction phase ends and kite goes through a fast transition phase in order to start the retraction phase. In the retraction phase, the kite is de-powered where its aerodynamic load is minimized and the tether tension is decreased. In this stage, the de-powered kite is reeled-in by an electric motor. De-powering can be done in different ways such as by decreasing the angle-of-attack, or by stopping the crosswind maneuver, or else by flying in the sides of the wind window. Once the retraction phase is over, the kite starts again with the traction phase. 2.3 Working Principles of Crosswind Kite Power Systems 7
Wind
Reel-out (traction) phase: Reel-in (retraction) phase: energy generation energy consumption
Figure 2.3: Traction and retraction phases of the pumping cycle operation, from [38]
In the pumping cycle operation, the system generates power by driving the generator during the traction phase and consumes power for reeling-in the kite in the retraction phase. Due to the de-powering of the kite, the required power for the retraction is significantly lower than the power generated in the traction phase. Therefore, a net power production occurs over the pumping cycle operation.
Figure 2.4: 3D representation of the figure-of-eight flight pattern, from [25] 8 Crosswind Kite Power Fundamentals Chapter 3
Literature Review
3.1 Modelling and Simulation
Modelling and simulation is a fundamental discipline for the development of complex engineering systems. A simulation framework can be helpful for the design, optimization and operation of the aerospace systems. As a matter of fact, the simulations are based on the mathematical models. Therefore, model building is necessary for every simulation. The modelling process is stated as an iterative process in [34]; a mathematical model is established on the laws of physics and further it is refined by the experimental results until a satisfying model is developed which is good enough for the scopes but with the accepted limitations. The simulations generate simulated data for a real system using the mathematical models. On the other hand, experiments make use of the physical models and collect observed data for the real system. The observed data coming from the experiments is also used in the iterative development process of the mathematical models for the verification and validation of the models. Yun and Li describe simulations as a method to obtain information through the use of assumed system models which presents techniques for analysis, decision-making, design- ing and training [42]. Moreover, they state that the simulations can decrease the flight periods, cost and risk. Stevens et al. express that the high cost of building and flight testing of a real aircraft leads to the development of the modelling and simulation disci- pline [34]. They indicate that the use of mathematical models together with computer simulations provides a relatively cheaper way for the performance evaluation and design improvisation, also creates the means to investigate the accidents and develop training simulators. At the moment, the modelling and simulation technology is an important discipline which is widely used in the aerospace industry, for the development of unmanned aerial vehicles (UAV), general aviation aircrafts, passenger aircrafts, military aircrafts, rotorcrafts, wind turbines, etc.
9 10 Literature Review
Comparatively, CWKP systems are complex systems using aerial and ground components. Therefore, the modelling and simulation technology is helpful for the development of the CWKP systems in following aspects:
Testing new design configurations • Checking flight characteristics • Conducting performance analyses • Implementing automatic control systems • Training kite pilots •
3.2 Crosswind Kite Modelling
As the simulation technology plays an important role for the development of CWKP systems, modelling of the kite is required in order to have a CWKP system simulation. Many kite modelling approaches can be found in the literature such as point mass, particle system, rigid body, multi plate, multi body and finite element models. A comparison of these modelling approaches is shown in Figure 3.1 considering the complexity and the computation time. The literature review on crosswind kite modelling is done for point mass, particle system, rigid body and multi plate approaches.
Figure 3.1: Complexity vs computation time for different kite modelling approaches, taken from [37]
Williams et al. investigate point mass model, rigid body model and flexible multi plate model for surf kites in [41]. They state that the point mass model is the simplest model representing the kite. However, the lack of rotational inertia in the point mass model makes it not useful for high level control algorithms. On the other hand, it is useful enough 3.2 Crosswind Kite Modelling 11
for basic flight trajectory and performance analysis. The rigid body model represent the kite with six degrees of freedom (DOF), which is required for the fully investigation of the flight dynamics of the kite. Hence, rigid body models give the position, velocity, attitude and rotational rate outputs for the kite by making use of the forces and moments acting on the body. The multi plate flexible model is built up as a series of hinged plates representing the kite. Williams et al. state that this model is useful for the representation of the deformation of the kite which is a common case for flexible kites [41]. Ruppert gives a detailed list of the CWKP system models available in the literature [26], along with an overview for the different modelling approaches as can be seen in Figure 3.2.
Figure 3.2: Description of point mass, rigid body, particle system (lumped mass) and multi plate models, taken from [26]
The rigid body modelling is a common modelling approach for the kites, especially for RW kites. There is an enormous amount of literature and research on the rigid body models for aircraft flight dynamics [8], [34]. A RW kite is basically an aircraft, more specifically an unmanned glider with the only difference that it is connected to a ground station by a tether. Hence, the previous literature and research on aircraft rigid body models are helpful for the CWKP modelling, with the only difference that when getting the equations of motion (EOM) of a rigid body kite model, the contribution of the tether forces and moments needs to be included. Figure 3.3 shows a model of the kite and a tether of a CWKP system [12], where a rigid body model is used for the kite and a particle system model for the tether, the forces f and moments m acting on the rigid body with mass M and inertia tensor I are also shown. Terink et al. use a rigid body modelling approach in the simulation of a tethered inflatable kiteplane [36], where the kiteplane is connected to a single tether with two bridle lines. 12 Literature Review
Figure 3.3: Representation of the rigid body model of the kite and the particle system model of the tether, from [12]
In order to include the kinematic constraints coming from the bridle, the Lagrange’s equations are used to obtain the EOM of the rigid body model. Winch launching of gliders and CWKP operations show some similarities. The glider is connected to a winch via a tow cable which is reeled during take-off and climb. This is similar where a CWKP kite is connected to a ground station via a tether which is reeled in and out during a pumping cycle. Thus the research on the glider launch modelling can be helpful for the crosswind kite modelling. Santel investigates glider winch launch using a numeric simulation, where the glider is modelled as a rigid body and the EOM of the rigid body model is obtained using Newton’s second law [27], [28].
3.3 TU Delft Dynamic Kite Power System Simulator
Fechner et al. describe a dynamic model of the pumping kite power system that is used in the Kite Power System Simulator (KiteSim) framework in [10], where the models of the atmosphere, tether, winch and kites are explained in detail. In this thesis, the KiteSim models for atmosphere, tether and winch are used directly; only slight modifications are done when required. In this section, these models and the implementation of the KiteSim is briefly explained; for further details reader should refer to [10].
3.3.1 Atmospheric model
KiteSim includes three atmospheric models for the calculation of the wind speed at the altitude; the power law profile, the logarithmic law profile and the fitted profile which is a linear combination of the power and logarithmic law profiles [10]. 3.3 TU Delft Dynamic Kite Power System Simulator 13
The air density at the altitude is calculated assuming the standard average sea-level air density as 1.225 kg/m3 which decreases exponentially with the altitude.
3.3.2 Tether model
The tether is modeled as a particle system with a fixed number of point masses1 connected with spring-damper elements2. The point masses representing the tether are equally placed on the tether creating n number of segments. Figure 3.4 shows the particle system representation of the tether with 6 segments, where PKCU represents last tether particle. The reel-out and reel-in of the tether is also simulated by varying the tether length as a function of tether reel-out speed. The forces acting on a tether particle are consisting of the weight of that particle, the spring-damper forces exerted on that particle by adjacent segments and the half of the aerodynamic drag of the adjacent segments.
Figure 3.4: Particle system representation of the tether (points P0 to PKCU) and the four point kite (points A to D), Pc does not represent a particle, taken from [10]
3.3.3 Kite models
KiteSim framework has two different kite models representing the LEI kites that are used in the kite power system demonstrator of the research team: a point mass model and a particle system model which is a four-point model3. The point mass model is showed in Figure 3.5a. This is the simplest modeling approach where the kite is represented as one point mass placed at the end of the tether. All of the external forces acting on the kite are directly applied at the point mass; specifically these forces are lift, drag, side and gravitational forces. This model does not include the
1Point masses of a particle system model are also denoted as particles in this thesis 2Spring-damper elements are also denoted as segments in this thesis 3The particle system model that is used in [10] is denoted as four-point model in this thesis. 14 Literature Review
(a) Point mass model (b) Four-point model
Figure 3.5: Representation of the point mass and four-point model, from [10] rotational inertia of the kite, it only includes the three translational DOF. Thus, the ac- curacy of the model is not sufficient to use in the development of control algorithms. This model is used for the flight path simulation and optimization, and for the initialization of the four-point model. The four-point model is the simplest particle system model that can include the rotational inertia of the kite in the simulation. It represents the kite with four discrete point masses (A, B, C and D) connected to each other by spring-damper elements as shown in Figure 3.5b. Total kite mass is distributed to the four particles using empirical equations. The kite geometry is parametrized using the kite width, kite height and bridle height of the LEI kite. Moreover, the positioning of the particles is done by empirical equations. This model attaches three aerodynamic surfaces to the top and side particles (B, C and D in Fig. 3.5b) for the aerodynamic model. The accuracy of the four-point model enables the development and optimization of control algorithms.
3.3.4 Winch model
The winch is modelled as assembly of an asynchronous generator, a gearbox and a drum. During the operation the tether can be reeled-in, reeled-out or braked. The winch is modelled by including the inertial dynamics of the winch and the torque-speed profile of the generator.
3.3.5 Implementation
The differential equations of the particle system consisting of the tether and kite particles together are formulated as an implicit problem. Furthermore, the problem is extended to include the winch states and residuals. The obtained implicit problem is solved by the Radau5DAE solver of Assimulo simulation package. 3.4 Particle System Modelling 15
KiteSim presents a soft real-time capable model, written in a general purpose program- ming language with possibility to adapt for different CWKP systems.
3.4 Particle System Modelling
The particle system modelling approach models the bodies as a finite number of point masses connected by spring-damper elements. Particles are defined as objects that have a mass, position and velocity. They react to the applied forces, but do not have spatial extent. A particle is represented by a point in space and it has the translational degrees of freedom in x, y and z directions. As being a point, it cannot have an orientation; thus, it does not have rotational inertia. However, multiple particles can be interconnected to form lines, areas and volumes leading to obtain rotational inertia as a system of particles. Figure 3.6 shows a particle, a line formed out of two particles and two triangles formed out of five particles. Here, the line has rotational inertia about y and z axes; it still does not have rotational inertia about the x axis. Likewise, the triangles have the rotational inertia about all axes.
Figure 3.6: Representation of a particle and the systems of particles formed by the particles along with local axis frames, from [37]
A particle system consists of the particles and the entities that apply forces to these particles. Knaap explains that a body can be subjected to the body and surface forces and these forces will result in internal forces [37]. Thus, in particle system modelling, the body is modeled by a number of particles connected by forces that represent the internal forces in the body. Hence, in a particle system where the particles are interconnected by spring-damper elements, these forces will be the spring-damper forces. The calculation of the spring-damper forces is done via Hooke’s Law and the damping force equation, using the positions and relative velocities of the connected particles. The motion of a particle system can be described using the Newton’s second law:
F(x, v, t) = Ax¨ (3.1) where A is the matrix of the particle masses, F is the vector function of the net force acting on the particles which is dependent on the states of the system (position vector x 16 Literature Review
and velocity vector v of the particles) and time. Introducing v = x˙ , the motion of the particle system can be written as a system of first-order differential equations as given below, from [37].
d x d x v = = (3.2) dt ˙x dt v A−1F(x, v, t)
This system of differential equations can be numerically integrated in time to solve and simulate the motion, using implicit and/or explicit methods for the numerical integration. The dynamic behavior of a rigid body can be accurately represented by its particle system model [37]. However, there are some concerns that are needed to be taken into account when representing a rigid element in a particle system. The first concern is the rigid body rotation and translation. Rigid bodies can be exposed to forces which are applied anywhere on them. When a force lies on a line which crosses the center of gravity of the rigid body, this force will cause the translation of the rigid body as seen on the left in the Figure 3.7. When a pair of parallel forces of the same magnitude but in opposite direction is present, these forces will cause the rotation of the rigid body as seen on the right in the Figure 3.7. Therefore, the representation of the loads applied to a rigid body can easily be determined by finding the net force through the center of gravity and the net moment which can be represented by two opposing forces as stated.
Figure 3.7: Representation of the external forces that cause the translation (left) and the rotation (right) on a rigid body, from [37]
The second concern is the representation of the center of gravity, the mass moments of inertia and the products of inertia of the rigid element. The particle system model should have equivalent mass properties along the local axis frame as the real body in order to represent the physics of the real body.
3.5 Component Buildup Method
In the component buildup method, the aircraft is divided into the discrete components representing the basic elements of the aircraft such as wings, horizontal tails, vertical tails, fuselage, propellers. The component forces and moments are locally calculated at 3.5 Component Buildup Method 17
each component and further the total forces and moments acting on the aircraft center of gravity (cg) are obtained using the contribution of each component. This method is widely used in aircraft modeling and simulation, with some examples seen in [31], [35], [16]. Selig models the full-envelope aerodynamics of the small UAVs for real-time simulations using a component buildup approach in [31], where the aircraft is split into wing, horizon- tal tail, vertical tail, fuselage and propeller components. For each component, a different model is developed to calculate the component forces and moments. The component forces and moments are calculated using the lookup tables that cover the full flight en- velope of the aircraft. The components are superposed and the component forces and moments are summed to obtain the total forces and moments acting on the aircraft cg; while doing this the interaction effects between the components are included. The local apparent flow at each component is determined using aircraft speed and rotation, wind speed, turbulence and the interference effects. For the wing and tail components, the aerodynamic coefficients and the induced angle of attack of the components are found using the lookup tables as a function of control surface deflection and angle of attack. Terink also uses a component buildup method in the simulation of tethered kiteplanes [35], the aerodynamic forces and moments of the components are locally calculated at the right wing, left wing, horizontal tail and vertical tail. Then, the contribution of these aerodynamic surfaces is used to calculate the total aerodynamic forces and moments at the kiteplane cg. The discretization can be seen in Figures 3.8 and 3.9.
Figure 3.8: The discretization of the kiteplane aerodynamic surfaces, from [35]
Figure 3.9: Illustration of lift and drag forces on the discrete elements of the kiteplane, from [36]
Khan and Nahon use the component buildup method for the dynamic modeling of an UAV with high maneuvering capabilities [16]. They state that the component buildup 18 Literature Review
method ensures that real aircraft phenomena such as induced roll or adverse yaw are directly included into the model, by calculating the aerodynamic forces and moments at the separate components instead of getting the total forces and moments of the aircraft as a whole. The aerodynamic forces are assumed to be acting at the aerodynamic center (ac) of the components; hence, the ac is chosen as the reference point for the calculations of the forces.
3.6 Aerodynamic Models for Dynamic Simulation
An aerodynamic model is required in order to calculate the aerodynamic loads acting on the particle system model representing the RW kite. The selection of the appropriate aerodynamic model can be critical for the dynamic simulation; as the complexity of the aerodynamic model increases, the required computing time also increases. The aerodynamic model that is used in KiteSim [10] consists of lift and drag coefficient curves that are functions of angle of attack as shown in Figure 3.10. These curves are obtained by using a model for unstalled and stalled airfoils and further modifications are made to better fit the curves for the LEI kites. In the simulation, the lift and drag coefficients are obtained as a function of angle of attack, which is calculated using the apparent wind velocity, steering settings and depowering settings. Further the aerody- namic lift and drag forces acting on the model are calculated, the aerodynamic moment is not taken into account. The control of the current LEI kite is being done with the depowering and steering lines. The former is changing the angle of attack of the whole LEI kite resulting in the pitch control, the latter is creating an angle of attack differential between the left and right sides of the kite resulting in steering control. Hence, the control of the LEI kite is modeled with the angle of attack changes at the aerodynamic surfaces.
Figure 3.10: Lift and drag coefficients versus angle of attack curves for LEI kites in KiteSim, from [10]
The control of the RW kite is done with the control surfaces; conventionally with ailerons, elevator and rudder for rolling, pitching and yawing. Thus, the control surfaces should also be included in the aerodynamic model to represent the effect of the control surface de- flections on the aerodynamics of the RW kite. Considering these aspects, a more detailed 3.6 Aerodynamic Models for Dynamic Simulation 19
aerodynamic model should be used for the RW kite simulations. Terink uses a strip theory aerodynamic model for the determination of the aerodynamic loads on a kiteplane [35], the aerodynamic forces and moments are locally calculated at the right wing, left wing, horizontal tail and vertical tail. Further, the contribution of these aerodynamic surfaces is used to calculate the aerodynamic force and moment vector of the kiteplane. The flight of a RW kite shows strong nonlinearities due to the large flight envelope. These nonlinearities should be included to a dynamic simulation in order to have realistic results. Abdallah et al. investigate the nonlinearity of an aircraft simulation model in [2], where the model uses approximately 50 lookup tables for aerodynamic and propulsive data for a wide range of angle of attack, sideslip angle, throttle deflection and control surface deflections. Thus, the nonlinearities in the aerodynamic and propulsive coefficients are directly represented in the simulation by means of these lookup tables. Selig models the full-envelope aerodynamics of the small UAVs for real-time simulations using a component buildup approach in [31], where the methods used to model the wing, fuselage and tail surfaces of conventional aircraft are presented in detail. In the component buildup method used in the simulation, the component forces and moments are calculated using the lookup tables that covers the full flight envelope of the aircraft. Further, these components are superposed and the component forces and moments are summed to obtain the total forces and moments acting on the aircraft cg; while doing this any interaction effects between the components are included. Selig states that the component buildup method is a regular method that is used in a wide range of modeling and simulation platforms. In the method, the aircraft is split into wing, horizontal tail, vertical tail and propeller. For each component, a different model is developed to calculate the component forces and moments. The local apparent flow is determined using aircraft speed and rotation, wind speed, turbulence and the interference effects. Selig states that the use of traditional stability derivative methods is inconvenient for the nonlinear aerodynamic behavior. For the RW kite model that will be developed for KiteSim, lookup tables can be used to calculate of the aerodynamic forces and the moments. Lookup tables can be generated using many different sources such as flight test data, wind tunnel measurements, analytical calculations, Computational Fluid Dynamics (CFD) results, empirical methods [31]. Sequeira et al. investigate different aerodynamic models (stability derivatives, strip the- ory, vortex lattice method and panel method models) for dynamic simulation of aircraft in [32]. The strip theory, the vortex lattice method (VLM) and the panel method as aerodynamic models are compared as follows. In the strip theory, the geometry of the aircraft is divided into discrete segments, and the aerodynamic loads are calculated on each of these segments. The total aerodynamic loads are calculated by summing the loads on each segment. In [32], Sequeira et al. state that the strip theory can be employed as a rapid development and behavioral estimation tool for different aircraft geometries, and any aerodynamic surfaces can be specified by this method. The model in the article implements rectangular wings with symmetrical airfoils and does not include the skin friction drag, downwash and wake modeling. The model calculates the local angle of attack and the local sideslip angle at the aerodynamic center of the each segment. Then, by using a reduced thin-airfoil theory, it calculates the lift and drag coefficients of the 20 Literature Review
segments. Further, the lift and drag of each segments are computed. Finally, these forces are accumulated at the center of gravity of the aircraft to get the total aerodynamic load of the aircraft at the cg. In VLM [32], the wing bound vorticity is represented by a lattice of constant dipole panels. Using the vortex ring elements, the radiation condition is satisfied. Moreover, the strengths of the vortex rings are satisfying the ”no normal flow through the mean surface” condition. The vorticity is represented by use of vortex wake filaments in a wake sheet lattice. The strength of the wake sheet satisfies the zero spanwise vorticity Kutta condition at the trailing edge. This method calculates the loads using the vortex strengths and the free stream velocity. The method does not include the induced drag. Sequeira et al. mention also about several drawbacks of this method. The use of a simple quasi-steady flat sheet wake model is a source of error. Also, the usage of a low order ring vortex model causes slow convergence with an increased panel discretization. The lack of body thickness can introduce additional errors. The panel method that is used in [32] is the FastAero panel method which achieves a fast solution for the unsteady potential flow around the bodies with thickness. Sequeira et al. mention that with the traditional panel method wake models, the interaction of the wake with the downstream surfaces creates problems and further states that the usage of FastAero eliminates these problems. Gohl and Luchsinger use an in-house developed VLM aerodynamic model in [12]. They state that this method calculates the aerodynamic forces fast and therefore real time dynamic simulation is possible. The VLM is used to calculate the lift and the induced drag. The viscous drag is calculated by using XFoil. In a CWKP system, a kite will be exposed to a large range of angle of attack during the operation. Hence, the simulation must include a post-stall aerodynamic data for the kite. Obtaining the post-stall aerodynamic data is not possible using linear aerodynamic models. Empirical methods, CFD analyses and experiments are some ways to obtain the post-stall aerodynamic data. Bianchini et al. investigate different models for the extrapolation of the post-stall lift and drag coefficient in [4]. They compared Viterna-Corrigan, Montgomerie, Beans&Jakubowski, Kirke and AERODAS models with CFD simulations and experiments. One of the find- ings from the comparison is that the lift coefficient has a good agreement in the models for the deep-stall region where the flow tend to be independent from the profile shape and the airfoil tend to act like a thin flat plate. They conclude that Kirke, Montgomerie and AERODAS models provide a more accurate extrapolation of post-stall data than the others. Chapter 4
Research Objective
The KiteSim framework developed by the TU Delft KitePower research team has the capability to simulate the LEI kites using the point mass model and the four-point particle system model. As mentioned in Section 2.2, the CWKP systems can have different designs; flexible, rigid or hybrid wings. The LEI kites are categorized under the flexible wing designs. RW kite designs are investigated and used by several companies and academic groups in the world. The current lack of simulation capability to simulate a RW kite in the KiteSim framework gives birth to the research objective and the fundamental research question of this thesis which are therefore formed as:
Enhance the KiteSim framework to include the capability of the dynamic simulation of a rigid wing pumping crosswind kite power system by implementing a particle system model to represent the tethered rigid wing kite.
Can a rigid wing kite be accurately modelled using the particle system modelling approach?
In reaching this research objective, two main research questions and the following sub- questions are needed to be answered:
1. How can a particle system model for a tethered rigid wing kite be added to the current modelling and simulation framework?
(a) How can the current four-point model for the LEI kites be modified to imple- ment rigid wing kite dynamics? i. Is the four-point model a sufficient particle system model for representing a rigid wing kite or should the number of points representing the kite be increased? ii. How will the aerodynamic surfaces and the parts of the rigid wing kite be represented in the particle system model?
21 22 Research Objective
iii. How will the structure of the particle system model be created? iv. What is the effect of the tether connection point(s)? (b) Which aerodynamic model for the calculation of the aerodynamic forces and moments should be chosen? i. Can the strip theory, vortex lattice method and/or panel method be used? ii. What are the available algorithms for implementing these models? iii. How can the post-stall aerodynamics be included? iv. How can the aerodynamic model be implemented to the simulation? v. How can the control surfaces be modelled?
2. What can be a validation process for the implemented rigid wing particle system models?
(a) Can a previously designed rigid wing kite be simulated? (b) Are there any previous tethered rigid wing kite data for the validation? (c) Can flight test data of a similar aircraft be used to validate the rigid wing kite model?
The following sub-goals are needed to be successfully completed in order to reach the research objective stated earlier.
Getting familiar with the current modeling and simulation framework • Modeling the rigid wing kite as a particle system model • Using the appropriate aerodynamic model for the aerodynamic surfaces • Implementing the particle system model of the rigid wing kite to the framework • Validating the particle system model of the rigid wing kite • Evaluating the dynamic simulation of a rigid wing kite design • Chapter 5
Methodology of the Simulation Modelling
5.1 Rigid Wing Particle System Model Structure
The particle system (PS) modelling approach can represent the geometric shape and the mass property of any object if the structure of the PS model is formed accordingly. The PS model is created by the discretization of the particles and the formation of the spring- damper elements. The rigid wing PS model contains 6 discrete particles connected to each other by 13 spring-damper elements as shown in Figure 5.1. The chosen PS model structure presupposes the following limitations and assumptions on the RW kites that are to be simulated.
The RW kite has a conventional configuration; with a wing and a conventional tail • (horizontal tail, vertical tail)
The fuselage is considered as a very thin rod and therefore negligible. • The RW kite has a single line tether and the tether connection point is at the cg • location of the kite.
The plane of symmetry of the RW kite is xBzB plane; Ixy = Iyz = 0 • The wings of the RW have small dihedral angles, and the horizontal tail has zero • dihedral angle.
The control surfaces of the RW are ailerons, elevator and rudder which are extended • full span of the wings, HT and VT.
23 24 Methodology of the Simulation Modelling
m2
m0 m1 m5
m3 m4
Figure 5.1: Representation of the PS model on a conventional RW kite, with 6 discrete point masses and 13 spring-damper elements
5.1.1 Discretization of the particles
The number of the discrete point masses representing the RW kite in the PS model is chosen considering the tether connection point on the RW kite and the representation of the aerodynamic forces and moments acting on the RW kite. The tether force is an external force and in a PS model an external force can be applied only to the point masses of the model. Thus, the PS model must have a point mass at the location of the tether connection point of the RW kite. In CWKP systems, the tether connection point is generally chosen at the cg of the kite or very close to the cg. Hence, for the model, a point mass is placed at the cg location of the RW, shown as m0 particle in the Fig. 5.1. The aerodynamic forces (lift and drag forces) are also external forces and must be applied only to the point masses. The aerodynamic surfaces of the RW kite are broken down as the right wing, left wing, horizontal tail and vertical tail. Each aerodynamic surface has an associated point mass and the aerodynamic forces that are generated by that surface is to be applied to that associated point mass. The positions of these associated point masses are chosen as the mean aerodynamic center of the corresponding aerodynamic surfaces. m2, m3, m4 and m5 particles in Fig. 5.1 show the particles representing the right wing, left wing, horizontal tail and vertical tail respectively, and they are located at the mean aerodynamic centers of the aerodynamic surfaces. The aerodynamic moments are also external loading and needed to be applied only to the point masses. However, in particle dynamics, particles do not have rotational degrees of freedom; therefore, the aerodynamic moments cannot be directly applied to the particles. The aerodynamic moments are applied with an alternative method that is explained later in the report, in Subsection 5.6.3. In order to apply this method, m1 particle is placed at the mirror symmetry of the m4 particle with respect to the yBzB plane of the RW kite. Considering these aspects, PS model of the RW kite is consisting of 6 particles and the simulation requires the positions and the masses of these 6 particles. The calculation of the positions and the masses of the particles is done via an initialization script (RW init v2.py). The initialization script requires an input file containing the mass and geometric properties of the RW kite, an example of the input file can be seen in AppendixA which contains the data for AP2 RW kite of the Ampyx Power. Geometric 5.1 Rigid Wing Particle System Model Structure 25
input values that are span (b), taper ratio (λ), root chord (croot), dihedral angle (Γ), leading edge sweep angle (ΛLE), the positions of the root leading edge points of the aerodynamic surfaces and the location of the cg of the kite are used to determine the positions of the particles. Input mass properties are the kite mass (mkite), cg position, mass moments of inertia (Ixx,Iyy,Izz) and products of inertia (Ixy,Ixz,Iyz). They are used to determine the masses of the particles once the particle positions are calculated. The mean aerodynamic chord (¯c) and the spanwise location of the mean aerodynamic chord (Y¯ ) are calculated according to Equation 5.1a and 5.1b. They are geometrical parameters of the kite required for the calculation of the particle positions. The equation for the spanwise location of the mean aerodynamic chord is done with the assumption that the lift is proportional to chord [24]; moreover, for the vertical tail, the span of the vertical tail is considered twice of the vertical tail height; bVT = 2 hVT.
2 1 + λ + λ2 c¯ = croot (5.1a) 3 1 + λ b 1 + 2λ Y¯ = (5.1b) 6 1 + λ
The input positions of the root leading edge (LE) points and the cg of the RW kite are given in the aircraft station coordinate system. The origin of the Fuselage Station (FS) is at the nose and the positive direction is to the back. The origin of the Buttock Line (BL) is at the centerline and the positive direction is to the right. The origin of the Water Line (WL) is at the cg level and the positive direction is upwards. The particle positions are calculated in the body-fixed reference frame as given below.
x0 = 0, y0 = 0, z0 = 0 (5.2a)