Active aerodynamic load control on wind turbines: Aeroservoelastic modeling and wind tunnel

Active aerodynamic load

control on wind turbines

Aeroservoelastic modeling and wind tunnel

experiments

Active aerodynamic load

control on wind turbines

Aeroservoelastic modeling and wind tunnel

experiments

PROEFSCHRIFT

ter verkrijging van de graad van doctor

aan de Technische Universiteit Delft,

op gezag van de Rector Magnificus Prof. Ir. K.C.A.M. Luyben,

voorzitter van het College voor Promoties,

in het openbaar te verdedigen op

Maandag 17 Oktober 2011 om 15:00 uur

door

Athanasios Konstantinou BARLAS

Dipl. Eng. in de werktuigbouwkunde techniek, University of

Thessaly, Griekenland

Samenstelling promotiecommissie:

Rector Magnificus voorzitter

Prof. dr. ir. G.A.M. van Kuik Technische Universiteit Delft, promotor

Prof. dr. G.J.W. van Bussel Technische Universiteit Delft, promotor

Prof. dr. ir. M.H.G. Verhaegen Technische Universiteit Delft

Prof. dr. Z. Gurdal Technische Universiteit Delft

dr. S.G. Voutsinas National Technical University of Athens

dr. D.E. Berg Sandia National Labs U.S.A.

F. Rasmussen RISØ DTU

Acknowledgements

The road has been rather long, somewhat winding, but definitely full of knowledge and experience.

Over these years in Delft it has been my good fortune to encounter some (but probably not too many) people who have given me more of their time, professional and personal help, and above all companionship and true friendship.

I would first of all like to thank my promotor Gijs van Kuik. Although he did not give me the traditional scientific support and supervision that a graduate stu-dent would expect from his professor, he allowed and encouraged me to grow into an independent researcher. This has possibly been the most valuable experience during the course of this PhD.

There are some fellow researchers whom I would specifically like to thank for the fruitful cooperation and support they have given me over the years: Jan-Willem van Wingerden and Teun Hulskamp. I definitely learned so much from them.

A big thanks also goes to the many students I supervised throughout these years, for giving me the joy of seeing them develop into mature engineers and for all the things they taught me.

A list that, alas, has far too many names on it to mention separately is that of all the co-workers that I have worked, talked, lunched with (and argued with) over the years. My gratitude goes out to all these former colleagues at the Wind Energy group, and especially Carlos Ferreira, for helping me so much and debating so often with me.

Separately from the above list, of course, I should mention my dear friends Ameya Sathe, Busra Akay, Jaume Betran, Theodor Chiciudean and Fanzhong Meng. I was very lucky to meet people with who I could connect so deeply.

Moving towards more personal acknowledgements, I would like give many thanks towards all my family and friends in Greece for their help, friendship and

I am, of course, particularly grateful to my girlfriend Irene Tselou for giving me so much encouragement and support during the last part of my PhD.

Roskilde, Thanasis Barlas

Contents

Acknowledgements

i

1 Introduction

1

1.1 Fatigue load reduction for large wind turbines . . . 1

1.2 Unsteady loads and controls for power regulation and load reduction 2 1.3 Active control vs. passive control . . . 5

1.4 Definition of the smart rotor concept . . . 6

1.5 Review of smart rotor research in aerospace . . . 6

1.5.1 Fixed wing applications . . . 6

1.5.2 Helicopter applications . . . 7

1.6 Review of smart rotor research in wind energy . . . 12

1.6.1 Early investigations . . . 12

1.6.2 Feasibility studies . . . 13

1.6.3 Control surfaces aerodynamics/aeroelastics investigations -modeling and experiments . . . 15

1.6.4 Wind turbine active load control simulations . . . 19

1.6.5 Aeroelastic stability of smart rotors . . . 28

1.6.6 Power regulation and sensitivity of load reduction to per-formance with smart rotor control . . . 28

1.7 Continuation of research - Research objectives . . . 30

1.8 Thesis Context and Overview . . . 31

2 Analysis of unsteady loads and fatigue

33 2.1 Introduction . . . 33

2.2 Definition of the reference 5 MW wind turbine model and simulations 34 2.3 Statistics and spectra of important load fluctuations . . . 36

2.3.2 Gust cases . . . 41

2.4 Control surface requirements . . . 42

2.5 Unsteady aerodynamic content . . . 46

2.6 Contribution of load frequencies to fatigue . . . 50

2.7 Effect of upscaling . . . 56

2.8 Implications for smart rotor control . . . 56

3 Analysis of smart rotor concepts for wind turbines

59 3.1 Aerodynamic control surfaces . . . 59

3.1.1 Flaps . . . 61

3.1.2 Microtabs . . . 62

3.1.3 Camber control (Morphing) . . . 64

3.1.4 Active twist . . . 65

3.1.5 Boundary layer control . . . 66

3.1.6 Concept comparison . . . 68

3.2 Actuators - Smart materials . . . 69

3.2.1 Conventional actuators . . . 70

3.2.2 Smart material actuators . . . 70

3.2.3 Application of smart material actuation . . . 73

3.3 Sensors . . . 74

3.3.1 Strain sensors . . . 76

3.3.2 Accelerometers . . . 77

3.3.3 Inflow measurements . . . 77

3.4 Controllers . . . 78

3.5 General design issues . . . 80

3.6 Concept comparison and preliminary choices . . . 80

4 2D aerodynamics modeling

83 4.1 2D unsteady aerodynamics modeling for airfoils with trailing edge flaps . . . 83

4.1.1 Unsteady aerodynamics theory for thin airfoils in the fre-quency and time domain . . . 83

4.1.2 Unsteady aerodynamic theory and modeling for thin airfoils with trailing edge flaps in the time domain . . . 95

4.2 2D unsteady aerodynamics model verification . . . 97

4.3 Results for the 2D case of an airfoil with a feedback controlled flap 98 4.4 Discussion . . . 102

CONTENTS v

5 Non-rotating experiments

111

5.1 Description of experimental setup and test cases . . . 111

5.1.1 Wind tunnel . . . 111

5.1.2 Blade design . . . 112

5.1.3 Actuators and sensors . . . 113

5.1.4 Real-time environment . . . 114

5.2 Results . . . 115

5.2.1 Feed-forward control cases . . . 115

5.2.2 Feedback control cases . . . 120

6 Full wind turbine aeroservoelastic modeling

127 6.1 Necessary modeling features for smart wind turbine aeroelastic tools 127 6.2 Description of numerical tool - DU SWAMP . . . 130

6.2.1 wind . . . 131

6.2.2 aerodynamics . . . 132

6.2.3 structural dynamics . . . 137

6.2.4 controllers . . . 145

6.3 Model verification . . . 147

7 Scaled rotor experiments

153 7.1 Experimental setup . . . 153

7.1.1 Wind tunnel . . . 153

7.1.2 Blade design . . . 154

7.1.3 Actuators and sensors . . . 157

7.1.4 Monitoring and controls environment . . . 158

7.1.5 Operational conditions . . . 158 7.2 Controls design . . . 159 7.2.1 System identification . . . 159 7.2.2 Feed-forward control . . . 162 7.2.3 Feedback control . . . 163 7.3 Results . . . 164

7.3.1 SISO feedback control . . . 165

7.3.2 MIMO feedback control . . . 165

7.3.3 MIMO feedback and feedforward control . . . 166

7.3.4 Summary of results and controller comparison . . . 166

8 Results from aeroservoelastic simulations of the scaled rotor case

171 8.1 Experimental rotor modeling in DU SWAMP . . . 171

8.2 Aeroelastic analysis . . . 173

8.2.1 Baseline comparison . . . 173

8.2.2 Load reduction comparison . . . 174

9.1 Simple SISO control concepts . . . 182

9.1.1 Model linearization and controller design . . . 183

9.1.2 Test cases . . . 185

9.1.3 Analysis of results . . . 185

9.2 Advanced SISO control concepts . . . 188

9.2.1 Model linearization and controller design . . . 189

9.2.2 Test cases . . . 189

9.2.3 Analysis of results . . . 190

9.3 Model Predictive MIMO control concepts . . . 190

9.3.1 Model linearization and controller design . . . 192

9.3.2 Test cases . . . 203 9.3.3 Analysis of results . . . 203

10 Conclusions

209 10.1 Discussion . . . 209 10.2 Future Work . . . 211

Bibliography

213

Summary

229

Samenvatting

231

Curriculum Vitae

233

Colophon

235

Chapter

1

Introduction

1.1

Fatigue load reduction for large wind turbines

The wind energy market has achieved a significant and rapid growth in the past decade. With the target of reducing the cost of energy and the increased trend towards offshore wind farms, existing rotor sizes have increased greatly. Rotors of more than 120m diameter are already in prototype stage. ’Outsmarting’ the

limits of scaling laws1_{, requires new technologies and production methods. With}

the intention to lower the cost per kWh, new trends and technological improve-ments have been a primary target of research and development. Reducing the cost of wind turbine blades has an effect on the cost of energy, but only a small percentage of the total. However, if an innovative blade design can result in de-crease in loading, the general cost will dede-crease, as rotor loads affect the loading of other components, as the drive train and the tower [47].

For large wind turbines, one important design driver is fatigue, amongst others. Fatigue is caused by various environmental and operational changes (analysed in the next section), and can become more important as the rotor size increases. Mit-igating the amplitude of the fatigue loads could, therefore, lead to a longer service life of blades, but also possibly to lighter blades, contributing to the reduction of the cost of energy.

1_{Power increases with the square of the rotor diameter and the mass of a solid structure with}

the power of three. Historically, the mass of blades has increased with the rotor radius to the power of 2.3 to 2.65

reduction

Defining unsteady environment and ways of influencing it

The loads acting on a wind turbine during operation can be divided into aerody-namic and gravity loads (external), and structural loads (internal). These loads are related by the aeroelastic coupling. The aerodynamic forces on the rotor are affected by the relative velocities on the blade sections. These velocities show fluc-tuating values during wind turbine operation. Most of these fluctuations are of a periodic nature (appearing in multiples of the rotor frequency) but also stochastic components are important. The rotational sampling of the incoming turbulent field is also indicated by 1P (once per revolution) and higher harmonic frequencies superimposed on the turbulence spectrum in the frame of reference of a rotating blade section. In general the following effects contribute to the total fluctuations comprising an asymmetrical inflow field:

• Horizontal or vertical wind shear • Tower shadow

• Turbulence (and rotational sampling of eddies) • Yaw and tilt misalignment

Furthermore, gravity forces on the rotor blades cause a periodic excitation of the rotor blade structural dynamics at the rotational frequency of the rotor. These can interact with structural modes of other components, e.g. tower and drive train.

To reduce fatigue loads during the operation of a wind turbine, control systems should be able to influence the structural loads [61]. In order to alleviate the de-scribed loads the control system of a wind turbine should be able to either reduce the fluctuations of the aerodynamic loads (indirectly influencing the structural loads) or add damping to the structural modes (directly influencing the struc-tural loads) [100]. Many approaches for load reduction control, using the existing full-span blade pitch system, have been proposed and will be summarized below. Evolution of wind turbine control systems for power regulation and load reduction Upscaling of the wind turbine rotors during the years has not led to significant changes in the blade structure. On the contrary, the blade loads control systems have evolved greatly [173]. Until the nineties, the wind turbines making use of the ‘Danish Concept’ combined constant rotor speed with stall of the flow around the rotor blades: increasing wind speeds automatically induce increasing drag forces that limit the absorbed power. All other control options were considered too complex. The simplicity of this concept has certainly contributed to the

1.2

UNSTEADY LOADS AND CONTROLS FOR POWER REGULATION AND LOAD REDUCTION 3

success of the ‘Danish concept’, but evolution toward large rotor sizes appeared to be uneconomical. Nowadays, all large wind turbines run at variable rotational speed, combined with the adjustment of the collective pitch angle of the blades to optimize energy yield and to control the loads. This was a big step forward: the control of the blade pitch angle has not only led to power regulation, but also to a significantly lighter blade construction due to the lower load spectrum and a lighter gear box due to reduced torque peaks. It is believed that further upscaling of wind turbine rotors will require more advanced load control systems for load reduction [94, 30].

Advanced pitch control

The next step in blade load control was Individual Pitch Control: pitch angle ad-justment per blade instead of collective. In theory, this can further alleviate the rotor loads, specially due to periodic effects (wind shear, tower shadow, upflow and shaft tilt). Not only may the blades benefit from this reduction, but also the drive train and nacelle structure. Focusing only on periodic loads, control strategies from helicopter research have been investigated. Cyclic Pitch Control (1P cyclic change in pitch, [100, 191]) and Higher Harmonic Control (pitch actions with multiples of rotor frequency, nP [44, 65]), have shown some potential of load reduction. Although the wind field effects cause a systematic azimuth-dependent variation in the aerodynamic conditions at a point on the blade, in practice, it is very difficult to achieve any real gains by superimposing a cyclic variation of the pitch angle per blade, because of the dominance of stochastic variations due to turbulence and variation of wind shear and upflow according to environmental conditions [3]. More advanced approaches of using the blade pitch mechanism for load reduction control purposes have been proposed, based on real feedback con-trol loops. Power regulation is always achieved through the collective pitch angle. Bossanyi [3, 4, 5] has proposed the use of additional load sensors on the blades (strain gauges, accelerometers) to superimpose an additional (individual) pitch demand to the collective pitch. Van Engelen and Van der Hooft [171] suggest a parametrization of feedback loops for Individual Pitch Control around 1,2 and 3P frequencies for load reduction, making use of the multi-rotational (or Coleman) transformation, while the same method is investigated for stability analysis by Bir [49]. In a different approach, Larsen et al [9], demonstrated significant fatigue

load reductions 2 _{by using Individual Pitch Control, based on local blade flow}

measurements (angle of attack and relative velocity). The fatigue load reductions are in the order of 9-31% for various wind turbine components. Individual pitch control results are compared with the Cyclic Pitch Control concept and appear more promising. All the above approaches show fatigue load reductions of 10-20%, although large and fast pitch changes are required, which will lead to excessive

2_{Fatigue load is a representation of the loading cycles experienced during power production}

over the full operational wind speed range, with the numbers of cycles weighted in accordance with the proportion of time spent generating at each wind speed.

reaction principle of quantities on the rotor. A recent, different approach is in-vestigated, based on the concept of feed forward control of the incoming wind field. Van der Hooft and Van Engelen [164] suggest the estimation of incoming wind speed based on energy balance and Hand et al [69] propose the use of a LIDAR (Light Detection and Ranging) system to directly measure the upwind incoming flow field and react with the pitch system. The individual pitch control approach has been further explored incorporating advanced control techniques for

further load reduction. In [86], a multivariable H2 3 individual pitch controller

with feed-forward wind disturbance rejection technique is utilized and it is shown

that better load reduction capabilities can be achieved. An H_∞ 4 _{multivariable}

controller is used in [101], where active tower damping is included. Load reduction comparable to simple PI (Proportional Integral) controller with a first order low pass filter was achieved and robustness to uncertainty in aerodynamic coefficients has been shown. Similar work with multivariable controls including tower mode active damping is published in [20] where reduction in fatigue loads is presented compared to a normal PI controller. Hand and Balas [102] have also shown the ef-ficiency of a disturbance accommodating controller, which incorporates properties of coherent turbulence inflow structures, in achieving load reduction compared to a normal PI controller.

Individual pitch control can provide further load reduction and is still under re-search. Nevertheless, some issues that can limit the load reduction using pitch control can been identified. Large multi-MW blades can limit the speed of the pitch actuator to less than that needed for load reduction control. On the other hand, the excessive use will lead to wear of the pitch bearings and actuator. The results from the previous research efforts show that the demanded pitch angles and rates are relatively high, especially when trying to reduce fluctuations caused by turbulence. In [45] the dynamics and stability of a hydraulic pitch actuator are simulated, and it can be seen that the behavior of the actuator can limit the fast reaction time needed for load control. More distributed control is required in order to achieve considerable load reduction of the fluctuations in the asymmetric inflow field of large rotors. This can be achieved by using control devices along the blade span. Distributed active control based on real-time measured quantit-ies (being loads, accelerations or inflow states) can deal with fast changes in local aerodynamic loads. This is the target of smart rotor control, that is analyzed through this research.

3_H

2 is a modern control theory method for linear-quadratic optimization problems, where

external disturbances are assumed to be Gaussian white noise.

4_H

∞is a modern control theory method for linear-quadratic optimization problems, when

1.3 ACTIVE CONTROL VS. PASSIVE CONTROL 5

1.3

Active control vs. passive control

Although this thesis focuses on active aerodynamic control, some remarks have to be made for passive aerodynamic control concepts. Such concepts are often referred to as ’aeroelastic tailoring’. This is the case when rotors have specified, tailored aerodynamic and dynamic properties so that the response of the blades to excitations can be such that the excitations are alleviated. Passive systems have been and are investigated in many research programs and can be divided in two categories:

1. Coupling between two or more degrees of freedom

• Tension-torsion coupling: for increasing rotor speed, the centrifugal force twists (a part of) the blade by material (fiber) coupling, or by a discrete torsion mechanism [146].

• Bend-twist coupling: the (downwind) deflected blade is forced by ma-terial fiber geometry to twist towards feather [72, 12].

• Sweep-twist coupling: the blades have sweep in the rotor plane, giving a center of mass and aerodynamic center aft of the blade root center. Lift on the ”sweeped” part causes the blade to twist towards feather [110].

2. Adding flexibility to one or more degrees of freedom • Flexible blade root (flexbeam, teeter) [174, 175] • Flexible part of the blade (passive tip) [174, 175]

One example of such a passive control approach was the FLEXHAT program during the 90’s in the Netherlands [174, 175]. By using a passively activated tip and a flexible blade root the wind turbine configuration used in FLEXHAT led to fatigue load reductions of 30% in the blades and 100% peak shaving of the torque loads measured in field tests. The load reductions were convincing but no commercial follow-up was realized. The complicated mechanisms used on the blades and the not directly up-scalable concept were the main reasons.

In passive control techniques the actuator parameters are set at the design stage, so all passive controls are inherently open-loop. So, passive control systems may only be effective over a limited range of operating conditions and there may even be conditions for which a passive control system degrades system performance. Likewise, since most engineering flow phenomena contain complex unsteady mo-tions, the ability of a passive device to control these unsteady motions is inherently limited [6]. Also, although passive control systems are usually chosen mainly for their simplicity, for wind turbine applications, such systems are not necessarily simple, and possibly not reliable and easily maintainable.

On the other hand, in (closed-loop) active control, one utilizes measurements of the current state along with a model of the system to devise a new control that

target state. In consequence, active control techniques offer significantly more flexibility, especially when dealing with changes in a flow state, although they can be inherently complex [6].

Thus, this type of control requires that both actuators and sensors be designed and utilized in an effective way, so it can be realized with different kinds of sensors, actuators, controllers and aerodynamic devices. So for a fatigue load reduction control objective, one should identify certain ways to apply active control tech-niques efficiently, especially when variations in excitations (mainly the incoming wind field) are not easily predictable. Because of advances in material and control technology such kind of systems are now widely available and can possibly com-pete with passive systems in terms of simplicity and reliability. The investigation of such control concepts is the target of various research programs for the applic-ation of active load control for wind turbines. When referring to active control devices with built-in intelligence the term ”smart” or ”adaptive” is commonly used.

1.4

Definition of the smart rotor concept

By definition, a smart structure involves distributed actuators and sensors and one or more microprocessors that analyze the responses from the sensors and use integ-rated control theory to command the actuators to apply localized strains/displacements to alter system response [66]. Therefore, by smart rotor control, the active aerody-namic load control by using distributed devices with built-in intelligence is meant. More detailed and fast aerodynamic control can an mitigate the loading due to stochastic components of the wind, but it can also contribute to the challenges as-sociated with unsteady phenomena. Small, low inertia aerodynamic surfaces can both result in fast control reaction time and distributed control over the asym-metric incoming wind field. The advances in materials and control technology have contributed to the development of such systems. Similar concepts have been explored in aerospace research with applications to fixed and rotary wings.

1.5

Review of smart rotor research in aerospace

1.5.1

Fixed wing applications

The problem of actively controlling aeroelastic responses has been a major concern through the history of aerospace. Controlling structural responses using aerody-namic means can have various beneficial results like flutter suppression, fatigue load alleviation, gust alleviation, noise reduction or increased ride quality. Thus, long-term research programs have been investigating various applications in air-craft wings. The historical perspective of the subject has been well documented by Mukhopadhyay [154]. Programs like the Active Flexible Wing (AFW) [138],

1.5 REVIEW OF SMART ROTOR RESEARCH IN AEROSPACE 7

the Benchmark Active Control Technology (BACT) [36], the Smart Wing [143] and the Active Aeroelastic Wing program [38] have demonstrated the ability of active control systems to deal with aeroelastic instabilities, reduce loads or im-prove performance. In these campaigns, important issues have been analyzed, like unsteady aerodynamics, control surface design, actuator dynamics, controller design for load reduction and flutter suppression, and aeroservoelastic modeling, simulation and wind tunnel testing of actively controlled wings.

Moreover, various research activities in aerospace have been oriented towards the use of adaptive materials and integrated systems for aeroelastic control. The use of smart material actuators has been considered as an effective solution for control surface actuation. Also, the concept of morphing airfoils as aerodynamic control surfaces has been explored (an idea not so new, since it was used by the Wright brothers in their first successful flight in 1903). An interesting overview of early aerospace research programs about controlled aeroelastic response using such concepts is documented in [129]. Research on the topic is ongoing with nu-merous recent publications on controlling aeroelastic response with trailing edge flap devices for gust alleviation or flutter suppression [132, 140, 141].

1.5.2

Helicopter applications

Although investigations of controlling aeroelastic responses in typical airfoil sec-tions or wings offers the basis for every attempt of aeroelastic control, the concept of applying this idea to wind turbine rotor blades can be approached more real-istically considering similar applications in rotorcraft research. The concept of active control on rotor blades, especially by using smart structures (actuators, sensors, controllers) has been thoroughly studied in the field of helicopter tech-nology. The interest for smart rotor control in helicopter rises mainly because of the importance of vibration and noise reduction at the rotor. In this literature field a lot of topics have been studied, including control surface concepts, smart materials, smart actuators, design options, control strategies, modeling and ex-perimental testing.

However, some differences exist between helicopter and wind turbine applica-tions. Firstly, some operating parameters are different: Helicopter blades ex-perience higher tip speeds, load frequencies, centrifugal and aerodynamic forces (compared to their size). Moreover, the high amounts of scheduled maintenance required for helicopters are a given fact, whereas wind turbine blades need to have low maintenance requirements; considering the limited maintainability of offshore wind turbines, the use of devices that raise additional requirements is not easily justified. On the other hand, wind turbine blades are of much larger size, very much cost-driven and reliable, but not so much limited by weight (compared to helicopter blades used for flight). This leads to some restrictions but also advant-ages, concerning active control applications. Also, the aerodynamics of a wind turbine are in many ways parallel to the ones found in helicopter rotors. Similar

blade airloads and performance, as well as predicting the dynamic stresses and aeroelastic response of the blades [78]. On the other hand, wind turbines are subjected to some other complicated effects like wind shear, turbulence, tower shadow and wakes of other turbines. Airloads acting on helicopter blades (mostly at forward flight) are highly periodic due to the common variations in both the local angle of attack and the relative velocities seen by the blade sections during one revolution.

Major research programs have been running over the years evaluating previous research studies, aerodynamic control device concepts, actuators selection, smart materials and feasibility for rotor control. Review articles like the ones of Straub [87] and Chopra [66], analyze available concepts. More specifically, for control con-cepts, pitch control, twist control, camber control and moveable control surfaces (trailing edge flaps or servo tabs actuated by smart materials) are proposed. Also, smart materials for actuation purposes are reviewed (piezoelectric, electrostrict-ive, magnetostrictelectrostrict-ive, shape memory alloys (SMA) and electrorheological fluids) and actuator configurations are analyzed. Smart materials are favorable for actu-ation purposes due to several reasons: compact size, large actuactu-ation displacements with low energy requirements and fast frequency broadband response. A lot of experience in smart control for helicopter applications has been gained through the past 20 years resulting in various successful applications. Some representat-ive examples are well summarized in [66], where research achievements from the long term smart rotor program at the University of Maryland are mainly presen-ted. Investigations focused on closed-loop wind tunnel testing of Mach-scaled or Froude-scaled models, incorporating smart material actuated control devices. Discrete and embedded piezoceramic actuators as well as SMAs have been util-ized. The concepts of trailing edge flaps, active tips and active full-blade twist have been explored. The potential for load reduction has been demonstrated. In further research investigations, Roget and Chopra [26] performed closed loop control wind tunnel tests on a four bladed Mach scaled rotor with individually controlled trailing edge flaps. The actuation was based on piezoelectric bender actuators. System identification was used for controller design. Simultaneous reduction of 1 and 4/rev components of fixed-frame loads is demonstrated (43% reduction). A comprehensive summary of smart-material solutions for aeroelastic and vibration control has also been prepared by Giurgiutiu [153]. Active blade twist and active flaps concepts are reviewed, together with variety of smart ma-terial actuation concepts.

The subject of aeroservoelasticity has gained significant interest in rotorcraft re-search during the last decade. Especially the concept of actively controlled flaps has been greatly explored. The general perspective of the topic has been summar-ized very well in [119, 120, 48]. More recent work focuses on optimization of active flap control for vibration reduction and performance enhancement ([118, 121]).

1.5 REVIEW OF SMART ROTOR RESEARCH IN AEROSPACE 9

concept achieved control authority results

active twist with embedded AFC (numerical model) 1-2 deg tip twist 10% reduction in torsional loads active twist with embedded AFC (1/6 Mach scaled)) 0.5 - 0.75 deg tip twist

reduced torsional strain and vertical hub force

active twist with embedded PZT wafers (1/8 Froude scaled) 0.35-1.1 deg tip twist

-active twist with embedded PZT (Froude scaled) ±0.4 deg tip twist over 10% rotor thrust authority torque plate PZT actuated pitch (Froude scaled) up to 10 deg

40 and 8% reduction in flight control and aircraft gross weight

active blade tip with induced-strain rotary actuator (1/8 Froude

scaled) 2-2.5 deg tip pitch

-active blade tip with piezo-induced bending-torsion coupled

composite bean 1.7-2.9 deg aerodynamic thrust authority up to 30%

active twist with bending-torsion PZT wafers (1/8 Froude

scaled) 0.3-0.5 deg tip twist

-servo-flap with piezo bimorph actuators (Froude scaled) up to ±8 deg -servo-flap with piezo bimorph actuators (Mach scaled) ±5.7 deg to ±10 deg -servo-flap with mechanically amplified piezo-stack (X-frame) (1/6

Mach scaled) up to 10 deg

-servo-flap with mechanically amplified piezo-stack (L-L) (1/6

Mach scaled) 8 to 19 deg p-p

-evelon with piezo bimorph actuators (Mach scaled) ±5 deg to ±10 deg reduction in vibratory loads in forward flight trailing edge flaps with neurocontroller (numerical model) ±5 deg elimination of periodic blade disturbances trailing edge flaps with piezo actuator with neurocontroller (hover

stand Mach scaled) ±5 deg

suppression of 1/rev and induction of 2/rev loads

trailing edge flaps with multilayered piezobimorphs with

neurocontroller (Froude scaled) ±5 deg 80% reduction in vibratory loads trailing edge flaps with electromagnetic actuator (full scale at

whirl stand) ±5 deg to ±8 deg

-trailing edge flaps with piezo stack (X-frame) (full scale at whirl

stand) 2-4 deg.

-trim-tab with bi-directional SMA torsion tubes actuator ±7.5 deg -trailing edge flaps with C-block actuators (blade section in wind

tunnel) 15-25 deg peak-to-peak

-trailing edge flaps with piezo-stacks and L-L (blade section in

open jet wind tunnel) ±10 deg

-trailing edge tab with SMA wires (blade section in open-jet wind

tunnel) ±20 deg

-trailing edge flaps on a swashplateless rotor (numerical model)

±4.7 deg (added to ±7.1 deg for

primary control) up to 90% reduction in 4/rev hub loads trailing edge flaps with piezo-stacks amplification (real scale test

flight) ±10 deg 50% to 90% reduction in vibratory loads

Table 1.1: Developed concepts and achieved results in smart structures concepts for helicopters (compiled from [66, 26, 119, 120, 48, 118, 121, 153]).

Daimler Chrysler Research Labs and DLR [37, 23, 24, 27]). After long term ex-perience in higher harmonic and individual blade control techniques, the active flap concept for vibration reduction is pursued. A full scale rotor is developed based on a BK117/EC145. Actively controlled piezoelectric actuated trailing edge flaps are used on each blade. The system is tested during flight, in open-loop and closed-loop configuration and shows excellent performance in reduction of vibrat-ory loads (50-90% reduction).

Although the field of smart rotor research for helicopters is vast, and it is not the purpose of this thesis to fully cover it, based on the most important research and development effort, different concepts can be compared. Based on literature (especially references [66, 26, 119, 120, 48, 118, 121, 153] provide a large amount of information) a table of most important achievements in this field has been com-piled (table 1.5.2). The different implemented aerodynamic devices and actuators can be seen, together with details on the implementation (model, wind tunnel testing, scaling) and capabilities (control authority, loads reduction). In fig. 1.1, a schematic of the most important concepts is presented. The general layout of various actuation options for each case is also illustrated. By analyzing the various achievements, some conclusions can be drawn. Firstly, it is clear that maximum control authority can be achieved by using trailing edge flaps in combination with mechanically amplified smart material actuation. This has also been proven in real scale applications. Active twist concepts with embedded smart material fea-tures has proven interesting, but limited control authority can be provided. Also, the blade structure is changed considerably, affecting weight and stiffness proper-ties. Also limited variability in the control authority is possible (and no variable spanwise control). On the other hand, discrete hinged devices, although offering great performance in loads reduction as appearing from the mentioned investiga-tions, can require a complicated internal structure with pitch links, rods, etc. All these results are of great interest for wind turbine smart rotor applications, and should be taken into account as lessons learned from the helicopter research.

By further studying the various attempts for smart rotor control in helicopters some conclusions can be drawn also from the design point of view. Because of the strong periodic nature of airloads in helicopter blades, some investigations have focused on applying high frequency aerodynamic control to reduce these fluctu-ations, instead of real feedback control based on measured quantities (e.g. Higher Harmonic Pitch control). The use of aerodynamic control surfaces (trailing edge flaps, tabs, moving tips) on the blades gives the advantage of faster control with smaller deflections (due to the large moment arm near the blade tip) for reduction of blade root moments, without using full blade pitching that is inefficient due to the use of the swashplate and the larger inertia. Also, because of the small size and thickness of the helicopter blades, the potential of smart actuating devices was identified early. Smart materials can provide high energy density with small size and low power consumption. Because of the large centrifugal forces and generally

1.5 REVIEW OF SMART ROTOR RESEARCH IN AEROSPACE 11

{

-PZT stack -electromagnetic mechanical amplification PZT bender(s) SMA wires θtip PZT torsion tube PZT sheets / AFC

Active tip Active twist Active flap

Figure 1.1: Schematics of smart structures concepts for helicopters (PZT: Lead Zirconate Titanate, AFC: Active Fiber Composite, SMA: Shape Memory Alloy).

large loads on the sections (compared to their size) the maximum aerodynamic effect of the control surface is a big issue. The maximum achieved deflections of the control surfaces by using smart actuators is the most important parameter for such applications. Various amplification mechanism have been generally used in order to achieve bigger displacements. Most smart materials exhibit low strains and moderate forces for large scale applications. In order to make them applicable as discrete actuator devices, mechanical amplifiers, where force and strain capab-ilities of the material are interchanged, are used to increase the strain. Several configurations have been proposed. Usually, this kind of mechanical amplification system uses parts as rods, arms, frames etc to deliver amplified displacement or power to specific control devices from the actuators. Always a trade-off between force and displacement is taking place. Furthermore, a significant attempt was made to use embedded actuation on the blades which results in shape morph-ing (camber control) or twistmorph-ing (active twist). Unique methods utilizmorph-ing active fiber composites showed shape control capability, although generally the use of small deflection surfaces is preferred due to simplicity, reduced weight and power consumption. From the control objective point of view, smart rotor application approaches in helicopters have managed to develop efficient systems, which, with

bration and noise reduction.

1.6

Review of smart rotor research in wind energy

Although some preliminary investigations for active control using devices on the blades had been made during the 90’s, research regarding smart rotor control for wind turbines is a relatively new, innovative and ongoing part of research at various wind energy research institutes. Interest in the subject has increased during the past years, in connection with general research in evaluation of ad-vanced controls for load reduction on modern large wind turbines. Various re-search investigations of applying smart rotor control concepts on wind turbines are reviewed, focusing on active control solutions. Concepts and methods which, through simulations and experimental approaches demonstrate the potential for load reduction, are presented.

1.6.1

Early investigations

Preliminary investigations of aerodynamic control devices on wind turbine blades are performed by the National Renewable Energy Laboratory (NREL) during the 90’s in the USA. These aileron-type of devices are analyzed to be used for power regulation purposes and aerodynamic braking. Series of wind tunnel experiments are performed, simulations quantifying the devices performance, and also field tests, examining different configurations. In [11] five trailing edge devices are investigated to determine their potential for use as wind turbine aerodynamic brakes. These devices are compared mainly according to the achieved lift to drag ratio reduction and drag increase. The spoiler-flap concept is considered the best choice. In [134] extensive 2D wind tunnel tests of these devices are conduc-ted, analyzing various aerodynamic parameters for a range of angles of attack, control configurations and sizes. The control devices are evaluated also numeric-ally during rotating operation using a BEM (Blade Element Momentum theory) code. The overall performance of a wind turbine with such aerodynamic control devices is predicted, running simulations for various configurations (parametric fixed device configurations). In [135], field experiments using a 20kW horizontal-axis wind turbine that incorporates variable-span, trailing edge aerodynamic con-trol devices are presented. The target of these rotational, atmospheric tests is the quantification of the influence of span-wise 3D effects, by comparing aerodynamic parameters with the 2D experimental data from the previous wind tunnel research work. Although, only fixed configurations of three trailing edge devices are used during the tests, so no active control concepts are tested, this research work is of great importance, since it comprises a realistic investigation of the aerodynamic parameters associated with control devices used in variable span-wise length and the 3D flow effects associated with their performance. Specifically, a ‘softening’

1.6 REVIEW OF SMART ROTOR RESEARCH IN WIND ENERGY 13

of the stall behavior (∆Cl around stall) is observed, compared with the infinite

span (2D) results. Also, it is stated that the use of 2D experiments and data simulations underpredicted the effective reduction in lift for short span devices near the tip. The reason is suggested to be connected with the effect of strong vorticity being shed due to the device uploading, which reduces the lift in the inboard section, thus enhancing the performance of the device in terms of power regulation. Such effects are considered of great importance when designing vari-able span-length aerodynamic control devices for rotational applications, based on 2D measurements and modeling.

1.6.2

Feasibility studies

The concept of active control of wind turbine aeroelastic responses using local aerodynamic devices on the blades, although receiving great interest, has not been fully treated as a whole, studying the feasibility of implementation in modern sys-tems and analyzing all design parameters. Some recent works try summarizing available knowledge and future steps [75], but focus mostly on evaluation of certain actuation mechanisms. In Delft University of Technology this preliminary evalu-ation and knowledge base has been compiled during past years. Firstly, research work, concerning feasibility studies on smart rotor control for wind turbine applic-ations, has been conducted by Marrant, van Holten and van Kuik in the project ‘Smart Dynamic Rotor Control for Large Offshore Wind Turbines’. The results of this study are summarized in [61]. This research deals with the inventory of rotor design options and possible load reductions. The fluctuating loads on a wind tur-bine are described and possibilities of influencing fatigue loads or structural loads are discussed. Active rotor control concepts, which include pitch control concepts (collective, cyclic, higher harmonic), individual blade control (part-span pitch, aileron control, active twist) and active damping of blade and tower vibrations, are presented. Also, semi active and passive control options are discussed (passive tips, self-twisting blades, compliant blades). Present techniques are summarized with regard to sensors, actuators, aerodynamic devices and control strategies, and their application on large offshore pitch-regulated variable-speed wind turbines. Regarding sensors, strain gauges, accelerometers and force sensors are analyzed. Piezoelectric force sensors at the blade root are considered a feasible solution for the measurement of aerodynamic loads. Optical fibers are considered expensive and not well established for measuring strains on blades. Passive accelerometers are considered a good solution due to their bandwidth and low frequency limit. Regarding control strategies, four control strategies that had been developed to actively suppress vibrations in rotorcraft are analyzed: a feed forward adaptive control algorithm, a Fourier synthesis algorithm, a real-time adaptive neural net-work controller and an iterative learning controller. Considerations regarding the connection between controller design and wind turbine design are also pointed out. Regarding actuators, the categories analyzed are: conventional (pneumatic,

orheological, shape memory alloys (SMA), electrostrictive, piezoelectric, magneto-strictive). It is concluded that the smart materials with the best prospective for actuation in wind turbine blades are piezoelectric and SMAs, which can be used for discrete or distributed (embedded) actuation if necessary, in combination with an amplifier. Furthermore, aerodynamic rotor control concepts are summarized (full-span and part-span pitch control, blade twist control, microtabs, camber control, aileron control - flaps). Aerodynamic control with trailing edge flaps or microtabs was considered the most feasible concept due to high frequency capab-ilities and good structural and safety features.

In more recent overview and feasibility studies, more focus is put on analysis of the different options for aerodynamic control devices. In [157, 158], various devices and their performance, as known from literature, is evaluated. Flaps, microtabs, microflaps, vortex generators, suction/blowing, plasma actuators, synthetic jets, morphing are covered. Many of these concepts are evaluated as promising for load control purposes, and more focus is put on microtabs.

Table 1.2: Concept matrix of available options for distributed smart rotor control. In another feasibility inventory presented in [82] for the UPWIND work pack-age ‘Smart Rotor Blades and Rotor Control’, the state of the art in smart rotor knowledge is summarized and analysis of different concepts is performed (see fig. 1.2). The most important inventory analysis results are included in section 3.

1.6 REVIEW OF SMART ROTOR RESEARCH IN WIND ENERGY 15

1.6.3

Control surfaces aerodynamics/aeroelastics investigations -

model-ing and experiments

In aerospace research, investigation of the performance of aerodynamic control devices has always played a vital role in active control concepts. A lot of know-ledge has been gained in this field regarding aerodynamic modeling and experi-mental evaluation of different options. For wind turbine blades, certain require-ments exist for similar use of such concepts. In order to investigate the possibility of controlling fluctuating loads on wind turbine blades, research programs have focused on analyzing the aerodynamic efficiency of certain devices/surfaces, for possible use for wind turbine blades load alleviation. Simulations and wind tunnel tests at the 2D or non-rotating blade level quantify parameters which are import-ant for the intended control purposes.

Flaps

Trailing edge aerodynamic devices, like flaps or ailerons, have been considered as a concept of high potential. Trailing edge flap devices for wind turbine blades have been thoroughly investigated by Risø (The Danish National Laboratory for Sustainable Energy, now Risø DTU). Especially, attention has been drawn on the concept of variable geometry trailing edge (fig. 1.2), since the option of smoothly deforming the aft part of an airfoil using smart materials is possible with modern technology advances, and the potential of using such an approach is of great in-terest. In [115] a CFD (Computational Fluid Dynamics) study is carried out to determine the effect of the size and shape of the variable trailing edge geometry on the aerodynamic characteristics of a wind turbine airfoil. Three different shapes of trailing edge geometry are analyzed: rigid, soft curved and strongly curved. From the static simulations it is concluded that soft curved flaps with flap chord

to section chord ratios ranging from cf/c = 0.05− 0.10 would be optimal because

of the great influence in lift with insignificant drag penalty. From the dynamic measurements, it is concluded that the amplitude of the lift generated on an os-cillating airfoil could be reduced significantly by the counteracting movement of

the flap for a wide range of reduced frequencies5 _{(k=0.09 - 0.36).}

In [131] a 2D aeroelastic model is developed, based on a panel code and a spring-damper system for an airfoil with deformable trailing edge. For control, a simple PD (Proportional Derivative) control algorithm is used, with a target control strategy to minimize the tip deflection variation of the blade. The res-ults show the potential of such a control: The standard deviation of the airfoil displacements has been reduced to 25% of the value corresponding to no control, during 2 sec. simulations. All the other simulations (100s turbulence, gust) show

5_{The reduced frequency k is defined as k = ωc/2V , where ω is the angular frequency of the}

Figure 1.2: Airfoil trailing edge camberline with deformable trailing edge geo-metry [30]

considerable attenuation of the oscillation amplitudes.

In [98, 99], a potential flow analytical method for the unsteady 2D force distribu-tion on a variable geometry thin airfoil undergoing arbitrary modistribu-tion is described. In addition to already developed potential flow analytical expressions for unsteady aerodynamics of thin airfoils, usually described as thin plates with the addition of flat control surfaces (see Theodorsen [150] and Leishman [52]), this method adds the option for a smooth deflection of the airfoil shape by superposition of chordwise deflection mode shapes.

This analytical model was used by Buhl et al [93, 94], coupled with a linear spring/damper model for the elastic deformation of the airfoil. An optimal con-trol strategy is used to minimize the fluctuations on the airfoil normal force. The analysis showed that when the airfoil experienced a wind step from 10 to 12 m/s the standard deviation of the normal force could be reduced by up to 85% when the flap is controlled by the input of the airfoil flapwise position and velocity, while reductions of up to 95% could be obtained when the flap is controlled by the input of the angle of attack. When the airfoil experienced a turbulent wind field, the standard deviation of the normal force could be reduced by 81% for control based on measured angle of attack. The maximum reduction using a combination of flapwise position and velocity is 75%. Calculations showed that the effect of a time lag in the actuators and sensors significantly reduces the efficiency of the control algorithm (fig. 1.3). Likewise, the effect of a low maximum actuation

1.6 REVIEW OF SMART ROTOR RESEARCH IN WIND ENERGY 17

velocity reduces the efficiency of the control algorithm.

The mentioned analytical model described in [98, 99] was expanded by Andersen to account for dynamic stall conditions, as described in [57, 28, 29]. Necessary data for the flap unsteady lift contribution during stall are taken from wind tunnel measurements, and validated with them. In [90], the same model is used to inves-tigate the potential of using pressure sensors for flap control. Simulations show that simple control algorithms, based on one pressure difference measurement over the pressure and suction sides of an airfoil, result in a good load reduction potential, with up to 74 % reduction in RMS values of the flapwise loads. Behrens [111] has also simulated unsteady motion of trailing edge flaps using an Immersed Boundary method. Unsteady calculations are carried out at different angles of attack and flap angles. It is shown that this method delivers good results when comparing the pressure distributions with those calculated on full body-fitted meshes.

Figure 1.3: Turbulent wind input response. The reduction of the standard devi-ation of the normal force as a function of the time lag. Ay,By: Control based on airfoil flap-wise position and velocity, Aα: Control based on airfoil angle of attack [94]

The investigation of variable trailing edge geometry was further extended at Risø by building a prototype and performing wind tunnel tests [79]. A profile section of 2m was fitted with 36 piezoelectric bender actuators at 10% of chord length. The thin curved actuators were directly fitted to the trailing edge of the

sponses of the lift due to the flap deflection measured in the experiments, were also modeled with an indicial function formulation. Iin order to reduce the lift force fluctuations generated by pitching the profile, the flaps were actuated in a prescribed way: reduction of up to 82% in lift force was measured in a prescribed pitch and flap motion (with an appropriate phase delay between pitching and flapping).

Microtabs

The use of microtabs as aerodynamic devices for load control on wind turbine blades has been proposed and extensively investigated by van Dam et al. [163, 162, 68, 159, 160]. The effect of varying tab location, height and width has been simulated by van Dam et al. using CFD. Results show an increase of up to 50% for the lift coefficient (Cl) in the linear range of the lift curve. The percentage is larger for low angles of attack and decreases at higher angles. Also, data showed that a 1% of chord tab placed at 5% of chord from the trailing edge provided the best compromise for lift, drag and volume constraints in the trailing edge. Regarding increase in drag, from the experimental work of van Dam et al. it can

be seen that an increase ∆Cd of up to 0.025 (250 drag counts) can be noticed at

the case of a deployed microtab (20% increase compared to the baseline airfoil

with Cd of 0.01). For a change in lift coefficient of ∆Cl= 0.2 the drag penalty is

20 drag counts in a representative case of a deployed microtab. For comparison:

for a change in lift coefficient of ∆Cl= 0.2, the drag penalty is 10 drag counts in

a representative case of a deployed trailing edge flap [115]. Although the increase in drag strongly depends on the angle of attack surface deployment and chosen airfoil, in the case of microtabs it seems to be slightly increased. Also, noise issues are believed to be connected with the deployment of microtabs. In the work of Oerlemans [137] it was shown that microtabs produce a high level of trailing edge noise but only an increase in broadband noise when in spanwise gapped configur-ations.

3D CFD simulations were also conducted [160] in order to investigate the effect of gaps between spanwise distributed tabs. The relationship between tab solidity ratio and change in lift was found to be highly linear, which is important for con-trol purposes. So microtabs show distinct relationships between tab-gap sizing and the resulting level of load control. Also, 2D experiments in the UC Davis Wind Tunnel were performed in order to measure the aerodynamic performance of fixed and actively controlled MEM tabs. The experiments were conducted

at Re = 1x106 _{for the two blade sections (fixed tabs and remotely controlled}

integrated tabs) for different locations and heights (for the fixed tab) and com-pared to CFD calculations. Results show good aerodynamic performance, with

an achieved ∆Clof up to 0.4. Furthermore, in [159] unsteady CFD simulations of

1.6 REVIEW OF SMART ROTOR RESEARCH IN WIND ENERGY 19

with the previous experimental ones. The studies show the unsteady aerody-namic behavior of the microtabs during deployment and compare it to the one of ‘microflaps’ (i.e. tiny trailing edge flaps). It was concluded that, in general, the global temporal response is independent of these aerodynamic devices. In [124], this research is extended to include validation studies with experimental results comparing shedding frequencies and Strouhal numbers for static, deployed microtabs and microflaps.

Boundary layer control devices

In addition to utilising aerodynamic control surfaces used, like flaps and microtabs for aeroelastic control, boundary layer flow control methods seem appealing. Act-ive flow control is a vast field of research on its own. An interesting overview can be found in [6]. Using such techniques for load reduction on wind turbine blades is an idea already proposed [82], but not thoroughly explored. In [192, 193], the use of Synthetic Jets (Glezer2002) for controlling blade flow and blade vibrations is investigated. Wind tunnel tests were performed. Global flow measurements were conducted, where the moments and forces on the blade were measured and also the flow field over the blade was quantified using Particle Image Velocimetry (PIV). Using synthetic jets, the flow over the blade was either fully or partially reattached, depending on the angle of attack and the Reynolds number. Fur-thermore, by either changing the momentum coefficient of the synthetic jets, the number of synthetic jets used, or by using different driving waveforms, propor-tional enhancement of the moments and forces, as well as the reduction of the blades vibrations were obtained. Moreover, feedback control wind tunnel tests are performed, activating the synthetic jets oscillation at the presence of stall. In this way, stall-induced vibrations are largely alleviated, reducing the overshoot and contributing to faster decay of oscillations. In general, the potential for load reduction was shown, although the investigations were limited to the concept of solely reducing dynamic stall vibrations.

Also, the use of plasma actuators has been explored for wind turbine load con-trol applications. In [136], surface-mountable, single dielectric barrier discharge (SDBD) plasma actuators on wind turbine airfoils are investigated computation-ally and experimentcomputation-ally. In one case a single SDBD plasma actuator is used close

to the trailing edge, and achieve a ∆Cl= 0.08 shift in the lift curve. It is stated

that this performance adds linearly with more actuators along the chord span. In another case, an airfoil is modified with flow separation ramps. It is shown that the actuator can recover the lost lift at lower angles of attack, providing a shift

in lift of ∆Cl= 0.4.

1.6.4

Wind turbine active load control simulations

Although quantification of the potential in load control can be seen in the pre-vious ‘local’ investigations (2D models and experiments), the necessity of more

operation is obvious. During the past years various research efforts have investig-ated the global aeroservoelastic problem when using local aerodynamic surfaces on the blades. There is a variety of concepts and methods on this investigation, which are analyzed in this section.

At the Delft University Wind Energy Research Institute (DUWIND) a prelimin-ary comparison of different concepts for smart rotor control of wind turbines was carried out by Marrant [60]. Four different smart rotor blade concepts are com-pared based on their potential to reduce fatigue loads for particular dimensions, and on their aerodynamic efficiency, bandwidth and complexity. The fatigue load case during normal power production is examined, comparing load calculations for the conventional blade and the ‘smart’ blade. A three-dimensional, one compon-ent turbulcompon-ent model and a wind shear model are used for the time-varying wind field input. The benchmark wind turbine used is the DOWEC (Dutch Offshore Wind Energy Converter) concept 6MW turbine. A time-marching BEM model is used with no structural dynamics for the blade, which is considered rigid and un-deformed. The maximum load alleviation capacity of the smart structures is used in the analysis, where it is assumed that the smart rotor blade knows exactly the wind field state at every time step. Moreover, as a first approximation, the smart blade is assumed to react instantaneously to the load change, i.e. no controller is used. The four smart blade concepts compared are: trailing edge flaps, microtabs,

camber control (which changes the Cl− α curve) and active twist (which changes

the angle of attack). The variations in blade root bending moment are calcu-lated for the baseline blade and the smart blade incorporating different spanwise lengths of smart devices. The smart devices reduce the loads by changing the

∆Cl or the angle of attack in response to full knowledge of the wind input. The

limited bandwidth of the devices is also taken into account by cutting off the maximum frequency of the Fast Fourier Transform (FFT) of the blade root flap bending moment. Rainflow counting and Miner’s law are used for determining the fatigue damage in order to compare the different concepts. The comparison value used is the ratio of the total fatigue damage of each smart concept over the conventional blade (overall relative damage ratio). The actuation of all concepts is based on piezoelectric actuators, except for the camber control concept which

is assumed to be actuated by an inflatable structure concept. The values of ∆Cl

or ∆α and maximum bandwidth of these actuation concepts are taken from lit-erature. Results are shown in fig. 1.4. From this figure it can be seen that active

trailing-edge flap control/active camber control (∆Cl = ±0.4) is about twice as

effective as microtab control (∆Cl= ±0.3). Only microtabs with a larger tab at

the lower surface (∆Cl = −0.55to0.3) can keep up with the active trailing-edge

flap/active camber control concept up to 15% smart structure length. For active trailing-edge flaps, active camber control and microtabs, smart structure lengths of 30% are most efficient for the reduction of fatigue loads. Active twist achieves reasonable performance, but only when using actuators over the full blade length.

1.6 REVIEW OF SMART ROTOR RESEARCH IN WIND ENERGY 21

Figure 1.4: Comparison of smart rotor blade concepts with infinite bandwidth [60].

The first investigations with aeroservoelastic simulations of full wind turbine models with control devices are reported by NREL. In [16] a PI closed-loop con-troller is used in the aeroelastic code FAST (with the AeroDyn module). The controller is designed based on system identification with the objective of con-trolling ailerons (on the outer 30% blade span) for power regulation. Look-up tables are used for the aerodynamics of the ailerons. The response of the system to specific wind input conditions (gust, smooth turbulence) with and without con-trol is investigated. The concon-trolled ailerons reduce the response time to a step-gust wind input and yielded reasonable performance for a range of wind speeds and input conditions. In [15] a different approach for the design of the controller was used. The FAST code is used, in conjunction with system identification tools, to generate a wind turbine dynamic model for use in active aileron control design. The load reduction in fluctuations (gust or smooth turbulence) for the aileron controlled cases is evident, but only quantified in time series plots of root flap bending moment in the references.

In later research work [148], the investigation of microtab aerodynamic devices for load control is carried out, in a full wind turbine model, using input,

effects of the microtab devices are incorporated only in the form of adjustment in static lift and drag based on the experimental and computational results of van Dam et al. [163, 162, 68]. The full aeroelastic model is linearized and expressed in the fixed reference frame, using a multi-blade coordinate transformation (see [117]). A control strategy based on a LQR (Linear Quadratic Regulator) state space controller with full state knowledge is developed, which includes individual blade pitch control and controls the turbine operation differently in distinct oper-ation regions. A step change in logarithmic wind shear exponent is simulated, and control response with a traditional PI controller for (collective) blade pitch, in-dividual pitch and microtab control are compared in terms of reduction of blade tip deflections. Also, peak and fatigue loads are calculated based on load case defined in the standards of IEC (International Electrotechnical Commission) [1]. The variations in tip deflection are quickly reduced with the microtabs. With small control actions, the microtabs showsignificant load reduction potential. Dif-ferent extreme loads are reduced up to 9% and fatigue loads up to 25% with the microtab control. It is seen that individual pitch control slowly adapts to the change in wind input conditions and reduces the tip deflections, but only using large and quite fast pitch actions. Microtab control adapts faster to the change and reduces tip deflections faster (fig. 1.5). Similar work with integration of mi-crotabs in aeroelastic modeling and control is also presented in [161], using FAST. In [30] the research work at Risø on the 3D modeling of a wind turbine rotor with actively controlled, deformable trailing edge geometry is presented. BEM is used together with the elastic modeling of a rotating blade, which includes the spanwise distributed control surfaces. The unsteady flap aerodynamics and cam-berline dynamics are the same as described in [93, 94, 98]. The blade is modeled as a cantilever beam using modal representation. The turbine in this case is using a 33 meter long blade. PID (Proportional Integral Derivative) controllers using input signals from local flapwise deflections or accelerations on the flapped sections are implemented. Effects of system time lag, flap power consumption and signal noise are included. Rainflow counting and W¨ohler curves are used to determine the equivalent loads, which are minimized by a simplex-type optimiz-ation scheme, finding the optimal control for the considered case. The numerical investigations show a huge load reduction potential very dependent on time delay. The computational tests showed fatigue load reduction potential of up to 64%. Equivalent flapwise root bending moments were reduced, although with reduced potential (40%) when signal noise, actuator time lag, flap mass and maximum power consumption were added. Moreover, optimal placement and dimensions of flaps are investigated (fig. 1.6). It can be seen that an 11 meter flap gives equi-valent load reduction of more than 60%. Also, by dividing the flapping sections into split flaps of different lengths, the effects of these split flaps are investigated. According to the authors, in this way, it is possible to more effectively damp out energy from more vibrational mode shapes.

1.6 REVIEW OF SMART ROTOR RESEARCH IN WIND ENERGY 23

Figure 1.5: Comparison of calculated aeroelastic response using (from top to bottom) collective pitch, individual pitch, and microtabs controls for a step change in wind shear at 14 m/s [148].

Figure 1.6: Calculated equivalent load for the flapwise root bending moment for undivided large spanwise flaps as function of the total spanwise length of the flap (lengths of split flaps are also shown on the graph) [30].

In a recent research work [31], spanwise distributed deformable trailing edge geometry (DTEG) actuators are integrated in a full aeroelastic model of the Up-wind/NREL 5MW reference wind turbine, using the code HAWC2. The un-steady aerodynamics of these sections are modeled based on the work presented in [57, 28]. The normal baseline torque and pitch controllers of the reference wind turbine are used as developed by NREL. A non-traditional control scheme, based on physical reasoning, is used to control the individual deflections of the DTEG below and above rated for load reduction. The advanced flap controllers use a combination of input signals: inflow measurements (angle of attack and result-ant local velocity) from Pitot tubes located at the leading edge of the flapped sections, blade root bending moments and blade pitch signal. The DTEG signal contributions from the inflow measurements and the blade pitch angles are based on theoretical models. The performance of the integrated DTEG controllers is shown, under various turbulent wind conditions (fig. 1.7) and wind step cases. A fatigue reduction of 33% in the tower root moment is obtained for 7 and 18m/s turbulent wind cases. Furthermore, a reduction of 16% in the tower ultimate root

1.6 REVIEW OF SMART ROTOR RESEARCH IN WIND ENERGY 25

moment over a 10 minute series is seen at 18m/s. The fatigue in the flap-wise blade root moment is also decreased 48% in the same case. Depending on mean wind speeds and choice of control parameters, it is seen that the average power can also be regulated. An increased mean power production of 1.5% is seen. An-dersen presents an extension of this research work, in [32], where HAWC2 is also used together with an optimization algorithm to determine the optimal strain sensor and flap locations in terms of fatigue load reduction performance (the time series of loads and control actions can be seen in fig. 1.8).

Figure 1.7: (from top down) free wind at hub height, electrical power, flapwise blade root moment, tower root moment in flow-wise direction, collective pitch speed, flap deflection angles for 11m/s turbulent wind input [31].

Lackner [104, 105, 106] also investigated the integration of trailing edge flaps on a full wind turbine model of the Upwind/NREL 5MW Reference Wind turbine,

using GH Bladed c . The research work addresses how trailing edge flaps perform

for fatigue load reductions, and how they perform relative to an individual pitch control (IPC) approach. The traditional feedback control approach implemented for load reduction, utilizes a multi-blade coordinate transformation (see [117]), so that variables in the rotating frame of reference can be mapped into a fixed frame of reference. Single input single output control techniques for linear time invariant systems are then employed to determine the appropriate response of

Figure 1.8: (Left) Fatigue load reduction as a function of sensor and flap location. (Right) Optimized high-frequency time constant (control parameter) as a function of sensor and flap location [32].

1.6 REVIEW OF SMART ROTOR RESEARCH IN WIND ENERGY 27

Figure 1.9: Power spectra density of flapwise root moment. Comparison between baseline controller (SC), individual pitch control (IPC(, and flap control (IFC) [105]. Top figure shows zoom-in area around 1P peak.

the trailing edge flaps based on the loads on the blades. No distributed control is investigated (i.e. one flap per blade). The use of trailing edge flaps and this control approach is shown to effectively reduce the fatigue loads on the blades, relative to a baseline controller. The load reduction potential is also compared to an alternative individual pitch control approach. It is seen that active flap control is comparable to IPC but can also contribute to high frequency load reduction (see fig. 1.9).

In [13], a more advanced aeroservoelastic modeling approach is used, utilizing vortex-theory based aerodynamic models. First, a 2D investigation on a section with trailing edge flap is carried out, using a panel code with viscus-inviscid interaction formulation. The structural responses with and without a simple PID flap controller to impulsive and sinusoidal excitations are shown. A 3D investigation, on the blade level, was also carried out, using a free-wake vortex particle model coupled with a FE-type beam model for the Upwind/NREL 5MW reference wind turbine rotor. An excitation caused by an exponential wind shear with exponent 0.2 is used. A maximum reduction of 30-35% is achieved in the