Improve Energy Production by Using High Efficiency Smart Wind Turbine Blade

IMPROVE ENERGY PRODUCTION BY USING HIGH

EFFICIENCY SMART WIND TURBINE BLADE

JIALE LI

Submitted in partial fulfillment of the requirements

For the degree of Doctor of Philosophy

Dissertation Advisor:

Professor Xiong (Bill) Yu

Department of Civil Engineering

CASE WESTERN RESERVE UNIVERSITY

May, 2018

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUTAE STUDIES

We hereby approve the thesis/dissertation of

Jiale Li

Candidate for the degree of Doctor of Philosophy*

Committee Chair

Dr. Xiong Yu

Committee Member

Dr. Xiangwu Zeng

Committee Member

Dr. Yue Li

Committee Member

Dr. David Matthiesen

Committee Member

Dr. Mingguo Hong

Date of Defense

3/22/2018

*We also certify that written approval has been obtained for any proprietary material contained therein.

TABLE OF CONTENTS

TABLE OF CONTENTS ...... I

LIST OF TABLES ...... VIII

LIST OF FIGURES ...... X

ACKNOWLEDGEMENTS ...... XVI

ABSTRACT ...... XVIII

CHAPTER ONE. INTRODUCTION AND LITERATURE REVIEW ...... 1

1.1 Background ...... 1

1.2 Distributed Wind ...... 4

1.3 Research Motivation ...... 5

1.3.1 Innovative Wind Turbine ...... 6

1.3.2 Bio-inspired Blade ...... 8

1.4 Outline of Research ...... 12

CHAPTER TWO. WIND ENERGY RESOURCES IN CLEVELAND AREA ...... 15

2.1 Introduction ...... 15

2.2 Methodology and Technical Background ...... 15

2.2.1 Theoretical Wind Power Output ...... 15

2.2.2 Weibull Distribution ...... 16

2.2.3 Wind Power Density ...... 18

2.2.4 Turbulence Intensity ...... 18

2.2.5 Capacity Factor ...... 19

2.2.6 The Power Law Model...... 20

2.3 Description of the Measurement Sites and Equipment ...... 21

2.3.1 Site Description ...... 21

2.3.2 Northern Power® 100kW Wind Turbine ...... 23

2.3.3 Vestas® V-27 and Nordex® N-54 Wind Turbines ...... 25

2.3.4 ZephIR® LiDAR ...... 27

2.4 Measured Wind Data ...... 29

2.4.1 Monitored Wind Data ...... 29

2.4.2 Spatial-temporal Analysis ...... 37

2.5 Data Characteristic ...... 40

2.5.1 Turbulence Intensity ...... 40

2.5.2 Wind Power Density ...... 42

2.6 Validation of Monitored Data ...... 44

2.7 Conclusion ...... 51

CHAPTER THREE. ANALYSES OF THE EXTENSIBLE BLADE IN

IMPROVING WIND ENERGY PRODUCTION ...... 53 III

3.1 Overview ...... 53

3.2 Introduction ...... 54

3.3 Extensible Blade Concept ...... 55

3.3.1 Extensible Blade ...... 55

3.3.2 Specifications of Fixed Length Wind Turbine Blade ...... 57

3.3.3 Aerodynamic Model based on Blade Element Momentum (BEM) Theory 61

3.3.4 Validation of BEM Model in Power Output Prediction ...... 69

3.4 Structural Analysis ...... 71

3.4.1 Determine the Working Range of Wind Speed ...... 71

3.4.2 Modal Analysis ...... 74

3.4.3 Rainflow Cycle Counting Method ...... 78

3.4.4 Fatigue Damage Analysis ...... 82

3.5 Test Results ...... 85

3.6 Conclusions ...... 94

CHAPTER FOUR. EXPERIMENTAL STUDY ON THE PERFORMANCE OF

SMALL HORIZONTAL AXIS WIND TURBINE WITH BIO-INSPIRED BLADE ...... 97

4.1 Overview ...... 97

4.2 Introduction ...... 98

4.3 Setup of Wind Tunnel Tests...... 100

4.3.1 Wind Tunnel Components ...... 100

4.3.2 Wind Tunnel Control System ...... 102

4.4 Geometrical Design of Blade Models ...... 104

4.4.1 Model Description ...... 104

4.4.2 Blade Design ...... 113

4.5 Experimental Setups ...... 114 V

4.6 Experimental Results...... 117

4.6.1 Power Coefficient Curve ...... 117

4.6.2 Power Output ...... 124

4.7 Numerical Simulation and Discussions...... 134

4.7.1 Introduction ...... 134

4.7.2 Validation ...... 134

4.7.3 Simulation Model of S1223 Airfoil ...... 140

4.7.4 Modal Analysis ...... 146

4.8 Conclusions ...... 151

CHAPTER FIVE. SUMMARIES AND FUTURE WORK ...... 154

5.1 Summaries on This Research ...... 154

5.2 Future Work ...... 156

REFERENCES ...... 159

VII

LIST OF TABLES

Table 2-1 Wind shear coefficient α from ASCE 7-05 [81] ...... 21

Table 2-2 Northern Power 100 wind turbine parameters ...... 24

Table 2-3 Vestas V-27 and Nordex N-54 wind turbine parameters [87] ...... 27

Table 2-4 Locations and mean wind speed characteristics at 50m of the test sites ...... 30

Table 2-5 Seasonal theoretical capacity factors of the two wind turbine prototypes ...... 45

Table 2-6 V-27 wind turbine seasonal operation days and capacity factors ...... 46

Table 2-7 N-54 wind turbine seasonal operation days and capacity factors ...... 47

Table 3-1 Parameters for each section along the blade based on NREL prototype wind turbine [100]...... 60

Table 3-2 Blade extends type in the research...... 74

Table 3-3 Modal frequency...... 76

Table 3-4 Fatigue damage index ...... 84 VIII

Table 3-5 Comparison of total annual energy production by the original blade versus extensible blades at different sites...... 93

Table 4-1 Characteristics of the test models ...... 112

Table 4-2 Maximum power coefficient of the blade models ...... 120

Table 4-3 Total energy output ...... 133

Table 4-4 Modal Frequency ...... 147

LIST OF FIGURES

Figure 1-1 Global wind power capacity 2005-2016 ...... 2

Figure 1-2 Distributed renewable energy in modern electric grid ...... 5

Figure 1-3 Innovative wind turbines ...... 8

Figure 1-4 Biological functional surfaces [48] ...... 9

Figure 1-5 Leading-edge tubercles on humpback whale flippers ...... 12

Figure 2-1 Test sites wind map at 100 meters [85] (A: CWRU campus; B: in an industrial area; C: water intake crib 4 miles from the coast of Lake Erie) ...... 23

Figure 2-2 100kW wind turbine on CWRU campus ...... 24

Figure 2-3 Vestas® V-27 and Nordex® N-54 wind turbines ...... 26

Figure 2-4 Power curves of Vestas® V-27 and Nordex® N-54 wind turbines ...... 27

Figure 2-5 ZephIR® wind LiDAR ...... 29

Figure 2-6 Seasonal Weibull distribution at 50m (a) Site A; (b) Site B; (c) Site C ...... 35 X

Figure 2-7 Seasonal windrose compass distribution at 50m (a) Site A; (b) Site B; (c) Site

C ...... 36

Figure 2-8 Yearly wind speed distribution at the three sites (a) Site A; (b) Site B; (c) Site C

...... 40

Figure 2-9 Turbulence intensity of the three sites at 50m ...... 41

Figure 2-10 Wind power density of the three monitoring sites at 50m ...... 44

Figure 2-11 Capacity factors of V-27 wind turbine calculated with LiDAR-measured wind speed data versus monitored electricity output (with or without excluding the maintenance days) ...... 50

Figure 2-12 Capacity factors of N-54 wind turbine calculated with LiDAR-measured wind speed data versus monitored electricity output (with or without excluding the maintenance days) ...... 50

Figure 3-1 Schematic of the extensible blade concept (with an illustration of extension at the tip and middle of the blade)...... 57

Figure 3-2 DU-00-W-401 airfoil lift and drag coefficients [100]...... 59

Figure 3-3 Schematic of the blade with an example airfoil blade element (r is the distance from blade’s root to airfoil blade element, R is the blade radius, and the chord length is that of the straight line joining the leading and trailing edges of an airfoil)...... 59

Figure 3-4 Actuator disk model of a wind turbine in a stream tube [2] ...... 62

Figure 3-5 Blade element velocity components...... 63

Figure 3-6 Flowchart for calculating blade production power using BEM theory...... 69

Figure 3-7 Comparison of the blade element momentum (BEM) model’s predicted power output and the monitoring power output of the 100 kW wind turbine ...... 71

Figure 3-8 (a) Determination of the wind speed range for (a) blade extension at tip and (b) blade extension in the middle...... 74

Figure 3-9 The shapes of first four modes for (a) original length blade; (b) blade-extension of 20% at the tip; (c) blade extension of 20% in the middle...... 78

Figure 3-10 (a) Stress and strain cycles; (b) Load time history ...... 80

XII

Figure 3-11 Rain-flow cycle count for (a) Original length blade; (b) Middle extend 20% blade; (c)Tip extend 20% blade ...... 82

Figure 3-12 Comparison of the power curves for the original length blade, tip-extended blade and middle-extended blade...... 86

Figure 3-13 Weibull distribution of wind speed data at 10 min intervals at (a) Site A; (b)

Site B; (c) Site C...... 88

Figure 3-14 Statistical distribution of 10 min of energy output for the original blade and extensible blades at (a) Site A; (b) Site B; (c) Site C...... 92

Figure 4-1 Lab scale wind tunnel schematic...... 101

Figure 4-2 Block diagram of wind turbine system (adjust from [127]) ...... 103

Figure 4-3 (a) Ultimaker® 2+ 3D printer; (b) 3D printing material ...... 106

Figure 4-4 S1223 foil section ...... 107

Figure 4-5 Wind turbine Models ...... 107

Figure 4-6 Rotor blade models ...... 111 XIII

Figure 4-7 The blade twisting profiles about the quarter chord ...... 114

Figure 4-8 Wind speed variation ...... 117

Figure 4-9 Power coefficient for the rotor models ...... 123

Figure 4-10 Power coefficient increment...... 124

Figure 4-11 Power coefficient in time domain ...... 129

Figure 4-12 Mechanical power output in time domain ...... 133

Figure 4-13 Computational grid...... 136

Figure 4-14 Velocity field at different attack angle ...... 137

Figure 4-15 The comparison of simulation lift coefficient and experimental lift coefficient

...... 137

Figure 4-16 Airfoil pressure coefficient at different attack angle (a) α=0; (b) α=6; (c) α=12

...... 140

Figure 4-17 Computational domain and meshed grid...... 141

XIV

Figure 4-18 Pressure distribution and the pressure gradient lines near the blade surface of the models ...... 146

Figure 4-19 The first four modal shapes of the blades ...... 151

ACKNOWLEDGEMENTS

I would like to take this opportunity to express my deepest gratitude to my advisor

Professor Xiong Yu, for his valuable guidance and support during my graduate study. He gave me the privilege to work on several projects and the opportunity to attend academic conferences, which made me a quick starter in my academic life. He is a great mentor who

I gained experience that will benefit my entire life.

I will never have been able to finish my dissertation without the guidance of my committee members. I would like to thank Professor Xiangwu Zeng, Professor Yue Li, and

Professor Mingguo Hong for their excellent teaching and valuable advises. Special thanks go to my committee member Professor David Matthiesen who shared his projects, valuable data, and participating my final defense committee.

I would also like to take the opportunity to thank all the faculty members for their help and support. Special thanks go to Professor Mario Garcia-Sanz, the director of the

Control and Energy Systems Center in CWRU and Dr. Fa Wang who provided great help on the wind tunnel test. In addition, I would like to thank our department secretary, Nancy

Longo, who is always willing to provide help. I really appreciate the assistances from our

XVI

department technician, Jim Berilla, who is a great engineer. I also want to thank the great wind engineer, John Yinglin for providing valuable data that used in my research.

I appreciate all my fellow civil engineering graduate students for their helps and accompanies throughout my study. They make the department of civil engineering a warm family.

I would like to thank the National Science Foundation, Ohio Department of

Transportation, Thinkbox of CWRU through its Student Project Fund, graduate school of

CWRU through the Travel Award, and the department of civil engineering of CWRU through the Teaching Fellowship for providing financial support. Their supports are highly acknowledged.

Finally, but most importantly, I would like to thank my wife, Dr. Xuefei Wang.

Without her endless love, supports and encouragements, I could not have finished this dissertation. My appreciation is also given to my parents and other family members, who give me so much love and supports.

XVII

Improve Energy Production by Using High Efficiency Smart Wind Turbine Blade

ABSTRACT

JIALE LI

Wind energy is considered as one of the most promising green energy sources for its renewable, sustainable, and worldwide availability. Traditional wind farms usually contain hundreds of wind turbines at locations with high quality of wind speed. However, there are more and more distributed wind turbines installed nowadays. Distributed wind turbines are installed at or near the point of end-use for the purposes of meeting on-site energy demand and are sometimes installed at locations with unfavorable wind quality.

Increasing the power efficiency to take advantage of both low and high wind speed is of great importance for the wind energy industry.

Optimal designs of the wind turbine blade have been the subject of extensive research, and significant progress has been accomplished in the past years. This study reviewed previous research to lay down a knowledge base for investigating innovative wind turbine blades. Two innovative wind turbine blades, extensible blade and bio-inspired XVIII

blade are introduced in this study. The extensible ‘smart’ blade will be extended at low wind speed to harvest more wind energy; on the other hand, it will be retracted to its original shape when the wind speed is above the rated wind speed to protect the blade from damages by high wind loads. An established aerodynamic model is implemented in this paper to evaluate and compare the power output of extensible blades versus a baseline conventional blade.

The bio-inspired blade is inspired from the leading-edge tubercles on the humpback whale flippers can improve the hydrodynamic performance of humpback whale. This research investigates the potential of bio-inspired blade technology to improve the performance in increasing wind energy output for the small horizontal axis wind turbine.

The high lift low Reynolds number airfoil S1223 was chosen in this research, and the wind tunnel test was conducted in the Control & Energy Systems Center at Case Western

Reserve University. The result shows that the blade with a shorter wavelength and larger wavelength of tubercles has better performance in increasing the maximum power coefficient. Additionally, the tubercles can delay the stall significantly comparing to the reference blade.

Overall, this research provides insights into the wind resource of different terrain XIX

types in Cleveland area and introduces details about two innovative wind turbine blades, which could increase the energy production of the wind turbines.

CHAPTER ONE. INTRODUCTION AND LITERATURE REVIEW

1.1 Background

Wind turbines have been used by human beings for more than 3000 years [1]. Its roles have evolved from performing mechanical work such as pumping, grinding and cutting to renewable energy production [2]. Modern wind turbines are typically horizontal axis turbine with two or three blades, which are results of optimal design from both efficiency and cost considerations. The willingness to reduce greenhouse gas (GHG) emissions and the increasing demand for energy keeps pushing the development of renewable energy.

Wind energy is considered as one of the most promising green energy sources for its renewable, sustainable, and worldwide availability [3, 4]. The installed capacity of the wind turbine has increased significantly after the year of 2000 for its beneficial regarding the security of energy supply and reducing greenhouse gas emissions [5]. Energy productions from fossil fuels lead to massive amounts of carbon dioxide emissions that contribute to global warming [6]. The development of renewable energies is a strategy to

mitigate the global climate change. Wind energy has been one of the world's fastest- growing renewable energy sources and is seeing surges in the United States and other nations [5, 7, 8]. Wind energy is renewable, sustainable, and locally available; furthermore, it is believed to feature a high return on investment compared to other renewable energy technologies. The average growth rate of installed wind power was approximately 28% during the last decade [9], and it is projected that the worldwide installation of wind energy will increase at an even faster pace in the next few decades [3, 10].

Figure 1-1 Global wind power capacity 2005-2016

In European countries such as Spain and Denmark, wind energy has already contributed 22.8% and 37.6% of the annual electricity demand throughout the countries.

In the United States, the total wind power capacity is about 76 GW by the end of year 2016, and it accounts 5% of U.S. electricity generation [11, 12]. It is expected to supply 10% of the electricity by 2020, 20% by 2030, and 35% by 2050 according to the U.S. Department of Energy (DOE) [13]. On the other hand, the newly installed wind turbines have been transformed from the traditional onshore wind farm to the offshore wind farm [14], the freshwater wind farm, and the distributed wind energy system. Coastal wind sources in the

United States are abundant and is a potential energy source [15]. The first commercial offshore wind farm in the United States, the Block Island Wind Farm, began to generate electricity at the end of 2015 [16]. In the United States, there are several large cities near the Great Lakes, and some researches have been seeking to develop wind energy in the

Great Lakes [17-19]. In Lake Erie specifically, an offshore wind demonstration project carried out by the Lake Erie Energy Development Corporation (LEEDCo), is planned to be built in 2018. Six of 3.45MW turbines are planned to be built off the Cleveland shoreline with a total output of 20.7MW [20]. In addition, more and more wind turbines named

‘distributed energy resources’ are installed to take advantages of their close proximities to

the existing electrical grid or manufacturing infrastructures [10].

1.2 Distributed Wind

Increasing both the wind power output and efficiency have been consistent goals for the wind energy industry. The proper siting of wind turbines is important to achieve such a goal. It is recommended that wind turbines be constructed at sites with high-quality wind resources. According to TC88-MC 2005 [21], wind resources are classified into four levels depending upon the characteristics of the average wind speed. Many investigations have been conducted into optimizing wind turbine locations for a particular wind farm [22].

These include constructing the aerodynamic model to account for the variations of wind flow over hills, ridges, valleys, offshore, and other types of complex topography. However, sites with high-quality wind resources are typically located in remote areas far away from cities [7]. More and more wind turbines named ‘distributed energy resources’ are installed at locations with lower-class wind resources [23-26] to take advantages of their close proximity to the existing electrical grid or manufacturing infrastructures as shown in Figure

1-2. It helps to reduce the development and transportation cost, which offsets to a certain extent the disadvantages of the low-quality wind resource [27]. In contrast to wind farms, which typically contain hundreds of wind turbines, distributed generators are mostly small- 4

scale power generators located close to the service loads. There are a significant number of wind turbines built as distributed energy resources. According to the report of the U.S.

DOE, the cumulative installed capacity of distributed wind turbines reached 992 MW spanning all 50 states by the end of year 2016 [28], representing nearly 75,000 units across

36 states, Puerto Rico, and the U.S. Virgin Islands. Effective utilizing wind resources at sites with low and medium wind speed helps to make wind energy production to be more geographically dispersed; this also helps to reduce the inherent variabilities of wind energy production [29, 30].

Figure 1-2 Distributed renewable energy in modern electric grid

1.3 Research Motivation 5

1.3.1 Innovative Wind Turbine

Improving the energy production at sites with low-class wind bears an important practice value. One potential method is to increase the wind turbine hub height [22, 31-34], which utilizes the benefit that the near-ground wind speed increases with elevation. There are, however, significant cost factors associated with manufacturing, logistic transportation, and the construction of components for the higher supporting tower. An alternative method is to develop innovative wind turbine blades technologies that achieve both improvements in production and resiliency. Significant progress has been made in this aspect. A new design of a dual-rotor wind turbine (DRWT), which includes rotors in both upwind and downwind directions, has been studied by No et al. [35]; the authors used the blade element momentum theory to calculate the aerodynamic forces and the torques generated from each of the rotor blades. This dual-rotor wind turbine is considered to have better performance in extracting energy than a conventional single-rotor wind turbine. The dual-rotor design, which includes rotors on both upwind and downwind directions, has also been studied by Hu at Iowa State University [36]. In Huang and Wu’s study [37], a balloon- type airfoil whose shape changes with the pressure distribution has been introduced. The blade is full of air and is able to change its shape according to the pressure distribution.

The authors used the numerical simulation to simulate an NACA0012 airfoil blade and came out with the result that this innovative blade can achieve better aerodynamic performance than the conventional blade. Bhuyan and Biswas [38] described an

unsymmetrical cambered airfoil blade for a vertical axis wind turbine (VAWT) which achieves improved performance in self-starting and a high power coefficient. Bottasso [39] investigated a novel passive control concept to mitigate loads and suppress vibrations of wind turbines via a flap or a pitching blade tip that moves passively in response to blade vibrations.

Innovative wind turbine blades have drawn attentions of researchers with the purpose of improving the production of the small wind turbine. Small wind turbines are working under low Reynolds number flown, which ranges from 104 to 106 flow according to

Lissaman [40]. Based on this method, a significant amount of low Reynolds number airfoil was designed for small horizontal axis wind turbines in order to achieve better startup and low wind speed performances. Wang [41] proposed a novel Darrieus vertical axis wind turbine, whose blades deform automatically into the desired geometry and improve aerodynamic performance. Some innovative wind turbine design are shown in Figure 1-3.

Figure 1-3 Innovative wind turbines

These previous efforts primarily look at dynamically changing the cross-sectional shape of the airfoil in response to wind directions. Meanwhile, the diameter of the rotor is another major factor determining the maximum energy output. Longer blades feature larger sweep areas, and hence capture more kinetic energy. This leads to a lower cost per kilowatt- hour of energy produced, which has been validated by numerous studies [42-46].

According to Jureczko et al. [47], the manufacturing cost of a wind turbine blade is about

15%–20% of the total wind turbine production cost. Improving the total power output of a wind turbine via optimizing the wind turbine blades presents an important opportunity to increase the turbine’s cost efficiency.

1.3.2 Bio-inspired Blade

“Bio-inspired” means that learning from nature can give us important inspiration to 8

develop new methods and approaches to build the superior structures and construct the artificial advanced materials. The surfaces of different creatures have evolved into unique hierarchical microstructures with superior function to match their survival environment as shown in Figure 1-4 [48]. In addition to the above-mentioned innovations, wind turbine blades are also inspired by aquatic lives in bionics.

Figure 1-4 Biological functional surfaces [48]

Researchers have looked into innovative wind turbine blades technologies inspired by aquatic life to improve the production and resiliency of wind turbine. Bionics means imitating nature to solve engineering problems. It has been known for a long time that bluff bodies have the effects of altering the regular vortex shedding and drag characteristics [49,

50]. The tubercles on the leading edge of humpback whale flipper, characterized by different sinusoidal amplitudes and wavelengths (Figure 1-5), are unique morphological structures of this animal, which could increase the hydrodynamic performance of the whale.

The flipper has a wing-like, high aspect ratio planform [51]. In addition, the tubercles on the leading edge of the flipper act as a mechanism to delay the stall [52], allowing the flipper to maintain a high lift coefficient at high attack angle, giving the whale a good maneuverability [53]. Fish and Battle [51] for the first time analyzed the hydrodynamic performance of the humpback whale flipper from a 9.02 m humpback whale that beached on Island Beach State Park, New Jersey. From the real whale flipper, the authors found out that the planform and symmetrical cross-sectional designs of the flipper indicate hydrodynamic adaptations for generating lift and minimizing drag [51]. Watts and Fish [54]

made a numerical simulation of adding whale flipper tubercles on the airfoil NACA 643-

021. They compared the lift and drag forces of the wing with leading-edge tubercles versus

the same wing without tubercles at a 10-degree angle of attack. They found a 4.8% increase in lift, a 10.9% reduction in induced drag, and a 17.6% increase in the lift to drag ratio. In the study of Miklosovic et al. [55], they conducted the wind tunnel test on a scale model of whale flipper with and without tubercles using the airfoil NACA 0020. The result shown the model with tubercles delayed the stall angle by 40% while increasing lift force and decreasing drag force. In the study carried out by Bai et al. [56], the performance of the vertical axis wind turbine with tubercle blade was even lower than a straight blade. The airfoil shape used in their research was the NACA 0015 airfoil. In the numerical simulation down by Zhang and Wu [57], the mechanical torque of the blade with tubercles increased significantly at high wind speed and the airfoil was chosen as NACA 0021. Huang et al.

[58] used the leading edge protuberances blades on a small scale horizontal axis wind turbine (HAWT). They used an SD8000 airfoil, and the experiment was carried out for smooth leading edge and comparative models with leading-edge tubercles. It was found that the usage of leading-edge protuberances might be significantly helpful if the small- scale HAWT system was designed to operate at low wind speeds, whereas it was not ideal to be applied to the one specially designed for high wind speed.

Figure 1-5 Leading-edge tubercles on humpback whale flippers

1.4 Outline of Research

The dissertation analyzes wind data collected in 10-minute time intervals from three locations in Cleveland area to assess wind characteristics and presents two innovative wind turbine blades, which can increase the energy harvesting efficiency. Wind speed monitoring data at three different sites are then used to predict the annual wind energy output with different blades. The work is organized as follows:

 Chapter One introduces the background of this study, and reviews the recent

development of innovative wind turbine blades. The importance to develop

innovative wind turbine blades with higher efficiency has been emphasized.

 Chapter Two analyzes wind data collected from three locations in Cleveland area.

Specifically, the statistical analyses of wind data include the Weibull shape and

scale factors, turbulence intensity, and wind power density. In addition, the three

utility scale wind turbines, which used in this study, are introduced. The power

output of the prototype wind turbines Vestas® V-27 and Nordex® N-54 between

2013 and 2015 are used to validate with those predicted based on wind speed model

derived from LiDAR measurement.  Chapter Three describes the feasibility analyses of an innovative, extensible

‘smart’ blade technology aims to significantly improve the wind turbine energy

production. Monitored wind data of chapter two are used to predict the annual wind

energy output of original blade and extensible blades. The effects of extension on

the dynamic characteristics of blade are also analyzed.

 Chapter Four investigates the bio-inspired blade technology in increasing wind

energy output. The high lift low Reynolds number airfoil S1223 was chosen in this

research, and the wind tunnel test was conducted in the Control & Energy Systems

Center at Case Western Reserve University. Seven types of wind turbine blade, 13

including traditional blade and bio-inspired blade with different wavelength and

amplitudes of tubercles, are studied in this research.

 Chapter Five summarizes the work and major contributions. It also proposed the

challenges and outlook the future research interests, which deserve further

investigations.

CHAPTER TWO. WIND ENERGY RESOURCES IN CLEVELAND AREA

2.1 Introduction

This chapter analyzes wind data collected in 10-minute time intervals from three locations in Cleveland area to assess wind characteristics. Specifically, the statistical analyses of wind data include the Weibull shape and scale factors, turbulence intensity, and wind power density. This part analyzes the wind data collected over 5 years in 10-minute time intervals. In addition, the wind speed data at one of the locations is used to evaluate the capacity factors of prototype wind turbines that are subsequently installed in 2012. The power output by the two prototype turbines between 2013 and 2015 is used to compare with those predicted based on wind speed model derived from LiDAR measurement. The results show that the estimated capacity factor of wind turbines from LiDAR data is satisfactory, when maintenance days are excluded. The data presented in this part of the paper will be used in the following study.

2.2 Methodology and Technical Background

2.2.1 Theoretical Wind Power Output

The wind power output of a turbine is described by the well-known equation [59]:

1 P C Av3 (2.1) 2 p

3 where P is the power output; ρ is the density of air (typically 1.225 kg/m ); Cp is the power coefficient; A is the rotor swept area, and v is the inflow wind speed. The equation shows clearly that for the same type of wind turbine at a certain air density condition, the power output is significantly affected by the inflow wind speed.

2.2.2 Weibull Distribution

The wind speed data observed from various studies with different observation methods usually have widely different parameters and cannot be used to compare the wind power potential directly. Therefore, a general wind distribution is needed to represent wind energy characteristics. The two most common distributions are the Weibull and the Rayleigh functions, where the Rayleigh distribution is a subset of the Weibull distribution [60]. In recent years, the Weibull distribution has been commonly used, accepted, and recommended in literature to express the wind speed frequency distribution [61] as it gives a better fit for measured monthly probability density distributions than other statistical functions [62, 63]. The density function of the Weibull distribution in its simplest form is

[64, 65]:

kv fe(v) ( )k1 (v/c)k (2.2) cc

where f(v) is the probability density function, also refer to as pdf; v is wind speed (m/s); c is the scale factor (m/s) and k is the shape factor. A higher value of c indicates a higher wind speed is higher and the value of k indicates the wind stability. There are several methods used to calculate the parameters of the Weibull wind speed distribution for wind energy analysis such as mean wind speed and standard deviation (MWS), the method of moments (MM), the power density method (PDM), the graphical method (GM), and the maximum likelihood method (ML) [60, 66, 67]. In this research, the Weibull distribution is fitted to time-dependent wind speed data using the maximum likelihood method MLM because of the limitations imposed by our 10-minute time intervals [60]. Then the shape factor and scale factor could be calculated as follows [68, 69]:

NN1 vk ln(v ) ln(v ) ii11i i i k N (2.3) vk N i1 i

1/k 1 N k cv  i (2.4) N i1

where vi is the average wind speed in time step i and N is total the number of nonzero wind

speed data points. Eq. (2.3) should be solved by iterations with an initial guess k.

2.2.3 Wind Power Density

The power of wind could be estimated according to the well-known equation [2]:

1 P() v  Av3 (2.5) 2

in which P(v) is the wind power (W). To indicate the wind resources at a certain site, the wind power density (WPD), which describes how much kinetic energy from the wind per area can be transformed into energy production [70], is the most commonly used parameter.

The WPD equation is shown as follows [71]:

 Pk(v) 1 3 p(v ) f (v)dv  c3  ( ) (2.6) 0 Ak2

in which p(v) is the wind power density (WPD) in W/m2, Γ is the gamma function which has a standard form of [62, 72]:

 (x ) eux u 1 du (2.7) 0

2.2.4 Turbulence Intensity

The causes of wind speed turbulence are obstacles on the soil surface, roughness length of the terrain, and local dynamic variations of the air pressure and temperature [73].

There are many statistical descriptors of turbulence, which may be useful, and the turbulence intensity is one of those parameters. Turbulence intensity (TI) is a measure of the overall level of wind turbulence [2]. It is a relative indicator of turbulence and not an absolute value [74]. Its formal equation is shown below [75]:

 TI  u (2.8) u

in which TI is the turbulence intensity and σu is the standard deviation of the wind speed variations about the mean wind speed ͞u, usually defined over 10 min.

2.2.5 Capacity Factor

The capacity factor is defined as the ratio of the average power output to the rated output power of the generator and is given as follows [2, 76]:

kk Paveexp[ (v c / c) ]  exp[  (v r / c) ] k C ff 100% kk  exp[  (v / c) ] (2.9) Pr(v r / c) (v c / c)

in which Pave is the average power output of the wind turbine (kW), Pr is the rated electrical

power of the wind turbine (kW), vc is the cut-in wind speed, vf (m/s) is the cut-off wind

speed, and vr (m/s) is the rated wind speed respectively.

2.2.6 The Power Law Model

The power law model is also referred as Hellman exponential law model, which is firstly proposed by Hellman in 1914 [77]. It correlates wind speeds at two different heights in the power law format as [78]:

 h2 vv21  (2.10) h1

where v1 and v2 are the wind speed at different heights h1 and h2, respectively. The exponent

α is the Hellmann exponent, which also refers to as the wind shear coefficient (WSC).

According to Eq. (2.10), the WSC is derived as:

ln(v / v )   21 (2.11) ln(h21 / h )

Eq. (2.11) is an engineering empirical formula used to express the turbulence near the ground [79] with no explicit physical basis [80]. It is commonly used for the estimation of 20

wind loads for structural designs as documented in American Society of Civil Engineers

(ASCE 7-05), including the estimation of the wind loads on the wind turbine tower and blades. Table 2-1 shows the WSC for different landscapes based on ASCE 7-05. Among these, urban and suburban areas include wooded areas with numerous closely spaced. Open terrain includes flat open countries, grasslands, and all water surfaces obstructions having heights generally less than 9.1m. Flat, unobstructed areas and water surfaces outside hurricane-prone regions include smooth mud flats, salt flats, and unbroken ice [81].

Besides the terrain type, it has also been found that the WSC at the same location varies with the height, hour of the day, time of the year, temperature, etc. [82-84].

Table 2-1 Wind shear coefficient α from ASCE 7-05 [81]

Landscape type WSC α

Urban and suburban areas 1/4

Open terrain 1/6.5

Flat, unobstructed areas and water surfaces 1/9

2.3 Description of the Measurement Sites and Equipment

2.3.1 Site Description

The state of Ohio has great wind energy potential. According to the American Wind

Energy Association (AWEA), the installed wind energy capacity in Ohio was 443MW at the end of 2015, and even larger potential wind energy resources have not been developed yet. Both onshore and offshore locations are chosen in this research in order to compare the offshore wind energy potential with onshore energy potential in the same area. Figure

2-1 shows the annual average wind speed at 100 meters in northern Ohio near Lake Erie.

The figure shows that the wind speed becomes higher from inland to offshore and the highest wind speed is on the lake far away from the coast. The dataset used in this research consists of three locations shown in the following figure: Data from site A comes from the anemometer installed in a 37m tall utility-scale wind turbine located on the campus of

CWRU. The surrounding topography could be described as city area with tall buildings.

Data from site B comes from the LiDAR, which is in an industrial area 4 miles away from the Lake Erie coastal line. This study is based on a dataset collected by laser-based ZephIR® wind LiDAR. The LiDAR system is positioned at an open terrain location with no obstacles and the monitoring height was set at five different heights; include 70m, above ground level. The topography of the site nearby could be described as a flat area with buildings under 10m height. Data from site C comes from the LiDAR located on a water intake crib

4 miles away from the coast of Lake Erie, and the monitoring height was set at three different heights 30m, 40m, and 50m above water level.

Figure 2-1 Test sites wind map at 100 meters [85] (A: CWRU campus; B: in an

industrial area; C: water intake crib 4 miles from the coast of Lake Erie)

2.3.2 Northern Power® 100kW Wind Turbine

The prototype wind turbine used in this research is a 100kW utility-scale wind

turbine (Northern Power® 100) located on the campus of CWRU in Figure 2-2. The

key parameters of the turbine are shown in Table 2-2. The wind turbine was installed

in November 2010 with financial support from the Ohio Third Frontier Program. The 23

primary role of the turbine is to serve as a research testbed for electrical and mechanical research [86].

Figure 2-2 100kW wind turbine on CWRU campus

Table 2-2 Northern Power 100 wind turbine parameters

Configuration Description

Model Northern Power® 100

Design Class IEC IIA

Design Life 20 years

Hub Heights 37m

Power Regulation Variable speed, stall control

Rotor Diameter 21m

Rated Wind Speed 14.5m/s

Rated Electrical Power 100kw, 3 phase, 480 VAC, 60/50 Hz

Cut-In wind speed 3.5m/s

Cut-out wind speed 25m/s

The Northern Power installed the monitored system in the wind turbine to collect its operation data continuously, which include data on the wind speed, direction, output power, etc. in 10-minute interval.

2.3.3 Vestas® V-27 and Nordex® N-54 Wind Turbines

Two commercial wind turbines Vestas® V-27 and Nordex® N-54 were built on the studied sites in 2012 and have generated electricity to the grid ever since. Three years

(2013-2015) of wind power capacities from the monitored wind production data is used as a comparison basis to with the theoretical wind power capacity. Both wind turbines are three bladed upwind horizontal wind turbines. The V-27 wind turbine has 30m hub height with 3.5m/s cut-in wind speed and 26m/s cutout wind speed. The N-54 wind turbine has much higher hub height, which is 70m above the ground level. The cut-in wind speed and the cutout wind speed of the N-54 wind turbine are 3m/s and 25m/s respectively. Both of

the wind turbines’ rated wind speed is 14m/s. The wind turbines’ power curves are shown in Figure 2-4.

Figure 2-3 Vestas® V-27 and Nordex® N-54 wind turbines

Figure 2-4 Power curves of Vestas® V-27 and Nordex® N-54 wind turbines

Table 2-3 Vestas V-27 and Nordex N-54 wind turbine parameters [87]

Performance Description Model Vestas® V-27 Nordex® N-54 Manufacturer Denmark Germany Rated Power 225kW 1MW Number of blades 3 3 Rotor diameter 27 m 54 m Cut-in wind speed 3.5 m/s 3 m/s Cut-out wind speed 26 m/s 25 m/s Rated wind speed 14 m/s 14 m/s Hub height 30 m 70 m

2.3.4 ZephIR® LiDAR

Light Detection and Ranging (LiDAR) is a remote wind sensing technology capable of measuring three-dimensional relative wind velocity at multiple fixed distances from the optical transceiver [88]. The basic LiDAR principle relies on measuring the Doppler shift of radiation scattered by natural aerosols carried by the wind [89] such as water droplets, pollen, or dust. These aerosols will reflect the laser light and this light is measured in the

LiDAR’s photodetector and is known as backscatter. The analog signal is sampled at 100

MHz and converted at 200kHz to power spectra using a 256-point time domain FFT (Fast

Fourier Transform) [90]. The accuracy of the LiDAR system has been tested against cup anemometer with wind speed correlations of 0.99 or better [91, 92]. This study is mainly based on a data set collected by laser-based ZephIR® wind LiDAR. The LiDAR consists of a continuous-wave laser with a cone angle of 30.6° to the zenith. The measurement of the radial velocities is by 1.55μm laser, which is in the infrared range and will not harm human eyesight. The ZephIR® LiDAR can be rapidly deployed, configured, and redeployed with both onboard data storage system and wireless data transmissions systems.

The ZephIR® LiDAR obtains approximately 150 line-of-sight measurements in 20 milliseconds and 3 seconds of data to derive horizontal and vertical wind speed components and wind direction. In any given azimuth direction, LiDAR measures the line of sight or

radial velocity, VLOS, which contains resolved components of the horizontal wind speed (u) and vertical wind speed (w) according to the following equation [71]:

VLOS abs ucos(   d )sin(  )  w cos(  ) (2.12)

in which θ is the instantaneous azimuth angle and θd is the wind direction. By assuming

the flow is uniform throughout the entire measuring volume, u, w, and θd can be obtained

using a non-linear least squares method. In this research, the wind speeds data at 30m and

70m were analyzed corresponding to the hub heights of the prototype wind turbine.

Figure 2-5 ZephIR® wind LiDAR

2.4 Measured Wind Data

2.4.1 Monitored Wind Data

In order to compare the wind energy potential of the three sites, the wind data need to be adjusted to the same height in the first step. According to the Power law equation described in the last section, the wind speed of the three sites were all adjusted to 50m

height for comparison. Site A is on the campus of Case Western Reserve University surrounded with several buildings up to five floors. Site B is in an industrial area surrounded with low-rise workshop buildings and trees. Site C is a water intake crib on

Lake Erie about four miles away from the shoreline. The Power Law wind shear coefficient was chosen as 0.4 for site A, 0.25 for site B, and 0.1 for site C according to Table 2-1 based on the terrain types. The seasonal wind speed characteristics derived with theoretical methods using Eqs. (2.2-2.7) for the three locations at 50m are shown in Table 2-4.

Table 2-4 Locations and mean wind speed characteristics at 50m of the test sites

Mean Weibull S.D. SEM Weibull wind Scale WPD Time (m/s CV (%) Shape Wind Class speed factor c (W/m2) ) factor k （m/s） (m/s) (m/s) Site A Spring 4.37 2.26 51.72 0.015 1.72 4.78 107.49 2-Marginal Summer 3.54 1.63 46.05 0.010 2.01 3.90 47.87 1-Poor Autumn 4.58 2.22 48.47 0.014 1.90 5.04 110.31 2-Marginal 41°30'08.6" N 81°36'19.9" W Winter 5.02 2.18 43.43 0.014 2.25 5.57 126.52 2-Marginal

Site B Spring 5.44 2.74 50.37 0.024 2.11 6.16 180.47 2-Marginal Summer 4.41 1.91 43.31 0.017 2.46 4.98 83.98 1-Poor Autumn 5.15 2.44 47.38 0.022 2.25 5.82 144.15 2-Marginal 41°36'07.8"N 81°29'48.7"W Winter 6.16 2.86 46.43 0.027 2.30 6.97 242.44 3-Fair Site C Spring 8.52 3.73 43.78 0.021 2.44 9.62 608.99 6-Outstanding Summer 6.48 2.90 44.75 0.020 2.40 7.32 272.21 3-Fair Autumn 7.90 3.36 42.53 0.024 2.49 8.90 476.81 4-Good 41°32'53.7" N 81°44'58.7" W Winter 9.95 3.80 38.19 0.025 2.77 11.18 889.51 7-Superb

The table shows that the highest mean wind speed occurs in the winter and the lowest mean wind speed occurs in the summer for all three sites. The mean wind speed of

site A, B, and C in the winter are 5.02 m/s 6.16 m/s, and 9.95 m/s respectively. It is clear that the mean wind speed increases when the surrounding terrain becomes flat as the mean wind speed of site C is the highest and mean wind speed of site A is the lowest for all seasons. The variations in the range of wind speeds is least at site A and largest at site C, due to meteorological and oceanographic factors. The diurnal effects can also be observed from the Figure 2-8. The wind speed at the offshore site C is typically higher from 12:00 to 22:00 during winter and spring, which is possibly due to sea breeze circulations. Strong solar heating causes the overlying air to rise on the coastland during the day and thus the air over the adjacent water surface flows inland to replace the rising air. On the contrary, a reverse but weaker circulation develops overnight. At site A, the lowest Weibull scale factor of 3.90 m/s occurs in the summer and the highest value of 5.57 m/s occurs in the winter.

The same trend can be found at site B and site C. The Weibull shape factor k varies from

1.72 in the spring to 2.25 in the summer at site A, varies from 2.11 in the spring to 2.46 in the summer at site B, and varies from 2.40 in the spring to 2.77 in the winter at site C. The relatively high value of k indicates that the variation of mean wind speed about the annual mean is small in the summer for site A and B, and winter for site C.

Wind power density, measured in watts per square meter, indicates how much

energy is available at the site for conversion by a wind turbine. The seasonal wind power density (WPD) is also calculated in Table 2-4. The wind class is classified based on the

WPD according to National Renewable Energy Laboratory standard [93]. It demonstrated that the wind class is poor in the summer for site A and B but fair for site C. The wind class is marginal and fair for site A and B in the winter but is superb for site C in the winter.

The wind rose and the frequency histogram of seasonal wind speeds was plotted in

Figure 2-7 and 2-8 for all three sites. According to the International Standard IEC 61400-

12 and other international recommendations, the two-parameter Weibull probability density function is the most appropriate distribution function of wind speed data [94]. The

Weibull distribution and the frequency distributions of the wind speed matched very well on the monitoring site in this research as shown in Figure 2-7; therefore, the Weibull distribution factors could successfully be applied for the wind potential assessment and used to calculate the annual production of the wind turbine prototype in this paper. The wind rose diagrams, which were constructed using the measurements of wind speeds and corresponding wind directions, provide useful information on the prevailing wind direction and availability of directional wind speed in different wind speed intervals [74]. The seasonal wind direction frequencies for the three sites at 50m have been plotted in Figure

2-8 based on the mean wind speed data of up to five years (site A: Jan. 2014-Dec. 2014; site B: May 2011 to Apr. 2012; site C: Jan. 2006-Dec. 2010) in 10-minute time intervals of wind speed. An obvious seasonal trend with higher wind speed in the winter and lower wind speed in the summer were observed at all three locations. The effect of topography is obvious at all three sites as the dominant wind directions in different seasons are all different. The wind direction at the site A is dispersive as the dominant wind directions come from both NNW and S in the spring and the summer, but come from S and SSW during autumn and winter. For site B, the dominant wind directions in the spring, summer, and autumn are from the S. At site B, the percentage of wind speed from S is 17% in the spring, 14% in the summer, and 13% in the autumn. The dominant wind directions in the winter, however, come from NE and ENE with a probability of 14% and 15%, respectively.

The wind directions for offshore site C tends to be more stable compared with the other two sites, the dominant wind directions come from both NE and SSE in the spring but come from SW to SSE in the other seasons.

(a)

(b)

(c)

Figure 2-6 Seasonal Weibull distribution at 50m (a) Site A; (b) Site B; (c) Site C

(a)

(b)

(c)

Figure 2-7 Seasonal windrose compass distribution at 50m (a) Site A; (b) Site B; (c) Site C

2.4.2 Spatial-temporal Analysis

In order to analyze the annual and diurnal cycles of the wind speed at three different terrain conditions, the contour plot is used to depict the wind distribution in this research.

Within each month of the year, the wind data is averaged at the same time point with a 10 minutes time interval diurnally. Algorithms are programmed in the commercial software

MATLAB (MATLAB R2017a by MathWorks®) to process the data. One year of wind data for each site has been used to create the contour plots, and for site C specifically the wind data is from 2007. The results are shown in Figure 2-9.

For the 3D plot on the left of Figure 2-9, the X-axis is hour of the day, the Y-axis is the month, and Z-axis is the wind speed value. The annual average wind speed of each site is also plotted as a plane in the 3D plot for comparisons. The 2D contour plot can be obtained by projecting the 3D plot on the XY-plane as demonstrated in Figure 2-9. The 2D plot helps to compare the diurnal and annual cycles of wind speed simultaneously. The annual average of the wind speeds at the three sites are 4.38m/s, 5.29m/s, and 7.37m/s, respectively. The 3D plots indicate that there are significant differences between the annual average wind speed and the time dependent wind speed at every time point during the year. In addition, the wind speed shows a noticeable trend of seasonal variation. For all three sites, the wind speed is higher in 37

winter and lower in summer, and there is an obvious low wind speed belt from May to

September. It can also be observed that the variation range of the wind speed is higher onshore compared to offshore. The variation range of the wind speed at site A is the smallest and site C is the largest, which is because of the meteorological and oceanographic factors. The diurnal effects can also be observed from the figures. The wind speed at the offshore site C is higher from 12:00 to 22:00 in winter and spring, which is mainly because of sea breeze circulations.

Strong solar heating causes the overlying air to rise on the coastland during the day, and hence the air over the adjacent water surface flows inland to replace the rising air. On the contrary, a reverse but weaker circulation develops overnight [95]. At the onshore site A, however, the diurnal cycle of the wind speed is mainly affected by the heat exchange between the near ground surface and higher air layers. The wind speed at site A is higher from 16:00 to 22:00 all the year round. Because of the high heat capacity of water, the sea breeze effects are slow to change with time compared to the diurnal effects caused by heat exchange between air boundary layer over land [96]. The nearshore site B does not show a clear diurnal trend, which is because it is in the transition zone of the atmospheric boundary layer.

(a)

(b)

(c)

Figure 2-8 Yearly wind speed distribution at the three sites (a) Site A; (b) Site B; (c) Site C

2.5 Data Characteristic

2.5.1 Turbulence Intensity

The monthly turbulence intensity (TI) at three sites at 50m height can be calculated using Eq. (2.8), and the results are shown in Figure 2-10. In general, wind speed at site A has a higher value of TI than site B and site C for most of the month in spring, summer, and winter. However, the TI value of site A is the smallest of the three sites in October and

November. The highest TI value occurs at site A in April, which reaches 55.89%, and the lowest TI value occurs at site B in January, which reaches 29.27%. Site C has the lowest

TI value for six months out of a year, and the annual average TI value of site C is the lowest 40

of the three sites. The annual average turbulence intensities of the sites A, B, and C, are

46.49%, 43.81%, and 41.91% respectively. It can be concluded that all the three data recorded regions are characterized by a high level of wind turbulence. Although the offshore site C has the lowest TI among the three sites, it does not show a significant decrease compared to onshore sites. This is because, in this research, the test site is about

4 miles away from the shoreline, and it is still affected by the internal boundary layer formed as wind blows from land to the lake.

Figure 2-9 Turbulence intensity of the three sites at 50m 41

2.5.2 Wind Power Density

The annual variations of wind power density (WPD) at 50m height evaluated using

Eq. (2.6-2.7) for the three sites are shown in Figure 2-11, and the power densities show large variations from site to site. In general, Figure 2-11 shows that the mean wind power density follows the same trend as the seasonal mean wind speed of all sites, which is lower during summer and higher during winter. Site C has the highest value of WPD and site A has the lowest. The maximum wind power density occurs during afternoons in the month of February at site C and the value is 1075 W/ m2. The minimum value occurs at site A in

July morning and the value is 15 W/m2. The figure indicates that site C has the greatest wind power density; site A has the lowest wind power density, and site B is in between which annual average values of 93 W/m2, 168 W/m2, and 470 W/m2 respectively.

(a)

(b)

(c)

Figure 2-10 Wind power density of the three monitoring sites at 50m

2.6 Validation of Monitored Data

The former discussed two commercial wind turbines Vestas® V-27 and Nordex®

N-54 are selected to calculate the theoretical energy output in this sector. The theoretical seasonal capacity factors are calculated according to Eq. (2.9) and the results are shown in

Table 2-5. The capacity factors are computed using the monitoring one-year wind speed

data at both 30m and 70m heights and power curves of the two prototype wind turbines.

The range of the capacity factor for V-27 wind turbine varies from 5.37% in the summer to

17.89% in the winter. The overall trend of the capacity factor is lower during summer and higher during winter. The similar trend is observed from the capacity factor of N-54 wind turbine at the height of 70m. The capacity factors of the N-54 wind turbine are higher than

V-27 as expected as the wind speed at 70m is higher than that at 30m as shown in section

4.2. The annual theoretical average capacity factors of the two wind turbines are 12.36% and 16.68% respectively.

Table 2-5 Seasonal theoretical capacity factors of the two wind turbine prototypes

V-27 Capacity Factor N-54 Capacity Factor Season (%) (%) Winter 17.89 24.45 Spring 13.17 18.45 Summer 5.37 8.67 Autumn 13.00 15.16 Average 12.36 16.68

Table 2-6 and 2-7 show the actually seasonal capacity factors of the two wind turbines from 2013 to 2015 recorded by the meters installed in the two wind turbines. The tables also show the operation days of the two wind turbines, which excludes the maintenance and service days. The data for summer and autumn of 2013 was missing because some 45

technical issues happened in the meter’s data acquisition system. The issues were solved in the October of 2013 and the meter functions properly afterward. For the V-27 wind turbine, the operational days varies from 60 days to 82 days for each season. As expected, the capacity factor of the V-27 wind turbine is higher during winter and lower during summer because of the wind pattern observed in section 4.2. The capacity factors of the V-

27 wind turbine vary from 3.23 in the summer of 2015 to 16.46% in the winter of 2013.

The difference between theoretical capacity factors and the real capacity factors are also shown in the Tables. The negative value means the real capacity factors are lower than the theoretical capacity factors. Therefore, the real capacity factors are overall smaller than the theoretical capacity factors.

For the N-54 wind turbine, the actual capacity factors vary from 3.50% in the summer of 2013 to 16.87% in the winter of 2014. The difference between a theoretical capacity factor and monitoring capacity factor reaches 9.93% in the spring of 2015. Table 2-6 and

2-7 illustrate that the maintenance and service days take a significant amount of time for both wind turbines. Part of the reason might because these two wind turbines are all used wind turbines, which served 10 to 20 years.

Table 2-6 V-27 wind turbine seasonal operation days and capacity factors 46

Days Compare with Capacity Year Season of theoretical Capacity Factor Factor (%) operation (%) Winter 62 10.89 -7.00 Spring 68 9.31 -3.86 2013 Summer N.A. N.A. N.A. Autumn N.A. N.A. N.A. Winter 82 16.46 -1.43 Spring 74 10.37 -2.80 2014 Summer 61 3.83 -1.54 Autumn 60 9.19 -3.81 Winter 74 11.68 -6.21 Spring 65 7.68 -5.49 2015 Summer 59 3.23 -2.14 Autumn 71 10.89 -2.11

Table 2-7 N-54 wind turbine seasonal operation days and capacity factors

Compare with Days of Capacity Year Season theoretical Capacity operation Factor (%) Factor (%) Winter 71 15.41 -9.04 Spring 72 12.69 -5.76 2013 Summer 50 3.50 -5.17 Autumn 73 11.01 -4.15 Winter 65 14.59 -9.86 Spring 67 11.25 -7.20 2014 Summer 69 6.39 -2.28 Autumn 52 13.37 -1.79 Winter 73 16.87 -7.58 Spring 54 8.51 -9.93 2015 Summer 66 5.01 -3.66 Autumn 48 6.97 -8.19

In Table 2-6 and 2-7, the actual capacity factors of the two wind turbines show a significant difference between the theoretical capacity factors. However, the normal operation days also vary significantly from season to season. This part of the research aims to evaluate if the difference between theoretical energy output and actual energy output was because of the maintenance and service days. The theoretical capacity factors, recorded data capacity factors, and capacity factors only consider operation days (except maintenance and service days) are compared in this section and the results are shown in

Figure 2-12 and 2-13. The negative value means the real energy output was higher than the theoretical energy output and the positive value means lower. Table 2-6 and 2-7 illustrate that both wind turbines have higher energy output than theoretical in the autumn of 2015.

This might be because the wind speed in the autumn of 2015 is higher than the wind speed in the autumn of the monitored year.

From Figure 2-12, it can be seen from the plot that the theoretical capacity factor line is higher than the capacity factor derived from recorded data. Further observation showed that the capacity factor only considers operation days is overall closer to the theoretical capacity factor except in the winter of 2014. The difference average between the theoretical capacity factors and actual data capacity factors is 3.58%; however, the difference between

the theoretical capacity factors and capacity factors only considering operation days is only

0.57%.

The N-54 wind turbine follows the same trend as shown in Figure 2-13. It depicts that the theoretical capacity factors are much higher than the recorded data capacity factors but very similar to the capacity factors only consider operation days. The results indicate that average difference between the theoretical capacity factors and recorded data capacity factors is 6.22%. However, the difference between the theoretical capacity factors and capacity factors only considering operation days is only 0.31%.

From the discussion, it can be concluded that the accuracy of predicting the long-term wind energy output using LiDAR data is acceptable if the wind turbine maintenance and service days are taken into consideration. This demonstrates the reliability of LiDAR for accurate wind energy assessment.

Figure 2-11 Capacity factors of V-27 wind turbine calculated with LiDAR-measured wind speed data versus monitored electricity output (with or without excluding the maintenance days)

Figure 2-12 Capacity factors of N-54 wind turbine calculated with LiDAR- measured wind speed data versus monitored electricity output (with or without excluding the maintenance days)

2.7 Conclusion

This chapter describes a comprehensive analysis of wind energy resources at three locations at onshore, nearshore, and offshore locations near Lake Erie in the Cleveland area.

On-site monitored wind data recorded by the ZephIR® LiDAR system and anemometers at

10-minute time intervals are used for this investigation. The wind speed probability distributions, power density distributions, turbulence intensities, and power capacity factors are presented in detail. Furthermore, one year of wind data is presented to demonstrate the spatio-temporal variations, which shows diurnal and seasonal differences.

Two prototype wind turbines are used to calculate the potential energy output at one of the three sites. The wind energy production predicted from field measured wind speed data by

LiDAR is used to compare with the theoretical wind energy output with the actual wind energy output of two utility-scale wind turbines over three-year period. The seasonal wind capacity factors calculated directly using the LiDAR monitored wind speed data and manufactur power curves showed good agreements with the actual energy production data when the wind turbine maintenance and service days are excluded from consideration.

Overall, the results from this study indicate that LiDAR provides a reliable measurement of wind speed under field conditions.

The spatio-temporal analysis of the wind data shows that there is higher wind energy potential in the winter than in the summer for all three locations. There is also an obvious low wind speed belt from May to September. The results indicate that at 50m above ground level, the wind energy potential at the offshore location is 3-5 times higher than that nearby onshore area, with an annual average wind speed is 8.2m/s and an annual mean power density equal to 562 W/m2. It can also be observed that the variation in wind speeds was higher for onshore locations compared to offshore. The variation range of wind speeds at site A is the smallest and at site C is the largest, which is because of the meteorological and oceanographic factors. The high wind speed period at the offshore site is longer than that at the onshore site, which is mainly because of the high heat capacity of water. The offshore sea breeze effects are slow to change with time compared to the inland diurnal effects caused by heat exchange. The turbulence intensities of the three sites are all at a high level, but the offshore site C is slightly lower than the onshore sites A and B. This is because the test site is not far enough from the coast and is still affected by the internal boundary layer formed as wind blows from the land to the lake.

CHAPTER THREE. ANALYSES OF THE EXTENSIBLE BLADE IN

IMPROVING WIND ENERGY PRODUCTION

3.1 Overview

This chapter describes the feasibility analysis of an innovative, extensible blade technology. The blade aims to significantly improve the energy production of a wind turbine, particularly at locations with unfavorable wind conditions. The innovative ‘smart’ blade will be extended at low wind speed to harvest more wind energy; on the other hand, it will be retracted to its original shape when the wind speed is above the rated wind speed to protect the blade from damages by high wind loads. An established aerodynamic model is implemented in this paper to evaluate and compare the power output of extensible blades versus a baseline conventional blade. The model was first validated with a monitored power production curve based on the wind energy production of a conventional turbine blade, which is subsequently used to estimate the power production curve of extended blades. The load-on-blade structures are incorporated as the mechanical criteria to design the extension strategies. Wind speed monitoring data at three different sites, which described in chapter two are used to predict the annual wind energy output with different blades. The effects of extension on the dynamic characteristics of blade are analyzed. The results show that the

extensive blade significantly increases the annual wind energy production (up to 20% to

30%) with different blade extension strategies. It, therefore, has the potential to significantly increase wind energy production for utility-scale wind turbines located at sites with low-class wind resource as distributed wind turbines.

3.2 Introduction

Due to variations in the wind conditions across different sites, it is difficult for a fixed length blade to match the varying characteristics of installation sites. In fact, commercial wind turbine manufacturers supply wind turbines of similarly rated power outputs with different blade lengths for sites with different wind conditions (e.g. General

Electric). A new concept of variable length blade or telescope blade is proposed recently to increase the power output and the annual energy production of the wind turbine [97-99].

The smart blade has an extensible length that will adjust itself according to the incoming wind speed. The blade will extend at low wind speeds to harvest more wind energy, and it will retract to its original shape when the wind speed is above the ‘rated wind speed’ to ensure structural safety. Therefore, it will produce more energy while protecting the blade from possible damage under high wind speeds. Although this variable length blade has been proposed for several years, there is very limited information on the aerodynamic performance characteristics of this blade. In addition, the blade concepts in previous studies are only extended at the blade tip.

This research analyzes the concept of the smart blade with the extensible length adjusted according to the incoming wind speed. The blade will extend at low wind speeds to harvest more wind energy and it will retract to its original shape when the wind speed is above the ‘rated wind speed’ to ensure structural safety. The performance of this extensible blade was analyzed using blade element momentum (BEM) theory, which is an accepted method by the wind industry for wind turbine blade aerodynamic calculation and therefore provides practice feasible conclusions. The BEM model is firstly validated with the field- monitored energy production data of regular wind turbines. The performance of the extensible blade is then analyzed using field-monitored wind speed data at a few onshore and offshore sites around Lake Erie. The results show the promise of the extensible blade to improve energy production at sites with a low class of wind resources.

3.3 Extensible Blade Concept

3.3.1 Extensible Blade

coefficient is decided by the mechanical structure of the rotor, with the theoretical maximum given by the Benz limit. The rotor swept area is decided by the length of the blade. The blade length is typically controlled from safety consideration to prevent the structural failure at a critical high wind speed which is rarely exceeded in the turbine service life. In this sense, the fixed length blade is not optimized to work under low wind speed conditions. The low wind speed allows the blade length to be increased to improve the wind turbine production while posing no threat to its structure safety.

The basic idea of the extensible blade is to increase the blade length at lower wind speed to produce more energy; the blade will turn to its original length when the inflow wind speed exceeds the rated wind speed. Therefore, an improved power output curve will be achieved for all the working conditions while mitigating safety risk. To study the technical feasibility without losing generality, two types of blade extension scenery are analyzed to assess the benefits of the extensible blade in wind energy production; i.e., (1) extension at the middle of the blade and (2) extension at the tip of the blade (Figure 3-1).

In this study, the extensible part of the blades is assumed to have the same foil size as the connection parts. The parameters of the prototype turbine blade are first determined.

Aerodynamics analyses are conducted on the extensible blade at different extension conditions. The performance of the extensible blade in energy produced is compared with regular blade using the wind speed data at three different sites in Lake Erie area, Cleveland,

OH, USA.

Figure 3-1 Schematic of the extensible blade concept (with an illustration of

extension at the tip and middle of the blade).

3.3.2 Specifications of Fixed Length Wind Turbine Blade

The baseline turbine model is built based on a 100 kW utility-scale wind turbine

(Northern Power® 100, Northern Power Systems, Barre, VT, USA), installed on the campus of Case Western Reserve University. The key parameters of the turbine are shown in Table 2-2. The manufacturer power curve is plotted in Figure 2-2. The turbine was installed in November 2010 with financial support from the Ohio Third Frontier Program.

The primary role of the turbine is to serve as a research test-bed for wind energy research

[86]. The data of the wind turbine can be accessed from the online monitor system provided

by the Northern Power, for instance, wind speed, wind direction, wind turbine rotational speed, energy output etc.

Wind energy is produced due to the lift force on the blade produced by the incoming airflow, which drives the rotor. The airfoil shape characteristics are the essential factors determining the lift force. The model used in this research is based upon the airfoil DU-00-

W-401 from the well-known NREL prototype wind turbine. Because the blade profile data is unavailable for the 100 kW prototype wind turbine, the model used in this research is based upon the airfoil DU-00-W-401 from the NREL prototype wind turbine [100] and scaled down to the length of a 100 kW turbine. The detailed profile data of the NREL prototype wind turbine is available for research purposes. The lift and the drag coefficient of DU-00-W-401 are plotted in Figure 3-2. As a simplification, the airfoil is assumed to have the same shape from root to tip of the blade, with a decreasing chord length. The chord length of a blade is defined as the width of the wind turbine blade at a given distance along the length of the blade (Figure 3-3). In this study, the rotor shape of the NREL wind turbine is scaled down to the 21 m diameter blade with the corresponding chord lengths scaled and shown in Table 3-1.

Figure 3-2 DU-00-W-401 airfoil lift and drag coefficients [100].

Figure 3-3 Schematic of the blade with an example airfoil blade element (r is the

distance from blade’s root to airfoil blade element, R is the blade radius, and the chord length is that of the straight line joining the leading and trailing edges of an

airfoil).

Table 3-1 Parameters for each section along the blade based on NREL prototype

wind turbine [100].

Radius (m) Twist (Deg) Chord (m) Airfoil Shape

0.48 13.31 0.59 Cylinder

0.93 13.31 0.64 Cylinder

1.39 13.31 0.69 Cylinder

1.96 13.31 0.76 DU-00-W-401 a

2.65 11.48 0.78 DU-00-W-350

3.33 10.16 0.74 DU-00-W-350

4.02 9.01 0.71 DU-97-W-300

4.70 7.79 0.67 DU-91-W2-250

5.39 6.54 0.63 DU-91-W2-250

6.07 5.36 0.58 DU-93-W-210

6.76 4.19 0.54 DU-93-W-210

7.44 3.13 0.50 NACA64618 b

8.13 2.32 0.46 NACA64618

8.81 1.52 0.42 NACA64618

9.38 0.86 0.38 NACA64618

9.84 0.37 0.35 NACA64618

10.29 0.11 0.24 NACA64618

10.52 0 0.15 NACA64618 a DU stands for Delft University; b NACA stands for National Advisory Committee for Aeronautics.

3.3.3 Aerodynamic Model based on Blade Element Momentum (BEM) Theory

The basic principle of a wind turbine is to extract kinetic energy from the wind and is illuminated by an actuator disk in the one-dimensional tube is shown in Figure 3-4. The effect of an actuator disk is to induce a step drop in pressure while allowing a continues speed at the rotor area [101, 102]. The actuator disc causes the drop in the velocity of the inflow uniform wind speed Utot at the rotor area to a lower speed of Utot(1-a). In which a is the axial induction factor. When the pressure changes across the annulus disk in the rotor area, a thrust force is created acted on the annular element.

Figure 3-4 Actuator disk model of a wind turbine in a stream tube [2]

The aerodynamic analyses are conducted on the original fixed blade as well as the extensible blades using the blade element momentum (BEM) theory. BEM theory is a classical analysis method of wind turbines [103], which has been widely accepted for blade performance analyses; an established model such as BEM is selected for these analyses so that the results provide a practical assessment of the new blade technology. BEM is composed of two different theories; i.e., blade element theory and momentum theory [104].

Blade element theory assumes that blades can be divided into small elements that act independently of the surrounding elements and operate aerodynamically as two- dimensional airfoils as shown in Figure 3-5, in which α is the attack angle. The characteristics of blade responses (drag and life on each element) are determined by the angle of attack of incoming wind, which is the angle between the center reference line of

the geometry and the relative incoming flow W (Figure 3-5). The momentum theory assumes that the loss of air pressure or the generation of turning momentum in the airfoil blade element is caused by the work done by the incoming airflow [105]. The BEM theory couples these two theories together and calculates the total lift and momentum via an iterative process [106]. The model is subsequently used to determine the power output at a given wind speed.

The BEM theory is implemented via customized code developed with MATHCAD

(Mathcad 15.0, Parametric Technology Corporation, Needham, MA, USA). Details of the implementation procedures for the BEM theory are provided in the Figure 3-6.

Figure 3-5 Blade element velocity components.

The actual wind flow acting on the turbine rotor is rather complex and can be simplified by the use of the blade element theory. The velocity components in the radial positions of the blade can be expressed regarding the wind speed, the axial induction factor

(a), tangential flow induction factors (a’) and the rotational speed of the rotor (Ω). The axial flow induction factor (a) and the tangential flow induction factor (a’) are critical parameters in the BEM theory. Figure 3-5 illustrates the conceptual model to calculate the lift and draft forces on each airfoil blade element. The airfoil is assumed to have a blade pitch angle of β, and the wind acts on the airfoil with an attack angle α. The pitch angle is the angle between the blade chord and normal direction of the rotor plane, which is an important parameter for maximizing blade lift and determines the load acting on the blade.

The component of wind velocity in the direction of the blade is ignored as it does not contribute to the torque on the blade rotation. Therefore, the inflow angle ϕ which is the intersection angle between the inflow wind velocity and the rotation plane of the blades, satisfying the following relationship:

   (3.1)

BEM theory does not include the effects of tip losses, hub losses, skewed values, dynamic stall, and tower shadow. The lift and drag forces generated by the airfoil along the blade and the momentum equations are used to produce the induction factors. The calculation step was then organized into a series of equations that can be solved iteratively, which is further elaborated in the following sections.

An important assumption of the blade element momentum (BEM) theory is that the lift and drag forces acting on a blade element are solely responsible for the momentum which caused by the air passing through the blade swept annulus[107]. Lift and draft forces are determined by the relative wind velocity act on the airfoil. The wind velocity perpendicular to the rotor plane is the inflow wind velocity Utot(t) reduced by the amount of a × Utot(t) due to axial interference (i.e., (1 − a) × Utot(t)). Assuming the rotor rotates with angular speed Ω, the blade element at a distance r from the rotor axis will be moving with a tangential speed Ωr [108]. When the wind passes through the rotor plane and interacts with the moving rotor, a tangential slipstream (or wake rotation) of wind velocity a’Ωr is introduced. The resultant inflow wind velocity about the rotor blade W is shown in

Figure 3-6 and can be calculated via the procedures are shown in the following:

2 2 2 W Utot (1  a )  [  r (1  a ')] (3.2)

And the inflow angle  could also express using the velocity:

Ua(1 )   arctan[tot ] (3.3) ra(1 ')

To calculate the relative incoming wind speed, W at each position r along the length of the blade and for each total wind speed Utot, the axial flow induction factor a and tangential flow induction factor a’ need to be calculated first. Typically, this is done via an iterative numerical procedure, with the basic steps as follows [2, 109, 110]:

a. Assume an initial choice of a and a’. (for example a = a’ = 0 as an initial guess).

Ua(1 ) Calculate the inflow angle via   arctan[tot ], where Ω is the rotor angular ra(1 ')

speed. b. Calculate    ; c. Read Cl and Cd from the lift and drag coefficient curves shown in Figure 3-2 with the

result of α from step b. Calculate the coefficient of sectional blade element force

normal to the rotor plane Cx and coefficient of sectional blade element force parallel

to the rotor plane Cy:

CCCx l cos  d  sin

CCCy l sin  d  cos d. Substitute 퐶푥 푎푛푑 퐶푦 into the following expressions to calculate new values for a and

a’

a rr2 ()CCxy 1a 4 sin22 4sin

a'  C  ry 1a ' 4 sin cos

Cr()  3 r 2r e. Evaluate convergence of the solution by comparing the calculated a and a’ from step

e with the assumed a and a’ from step a.

f. If the differences between values are smaller than designated threshold, the process

stops. Otherwise, update a and a’ values and continue the iteration between (b) and (e)

until the results converge. g. Take the result of a and a’ into Eq. (3.2) to calculate the relative wind speed W.

The procedure shown above applies to different types of turbine blades. It also needs to note that as the lift and drag coefficients vary with attack angle, variable pitch wind turbine modulates the wind attack angle by dynamically adjust the pitch angle of blades.

The relative wind velocity gives rise to aerodynamic lift and drag forces acting on each segment of the blade, which can be calculated as follows:

1 F()() r C C r W2 r (3.4) Ll2

1 F()() r C C r W2 r (3.5) Dd2 where C(r) is blade chord length; r stands for the distance from the hub of a section of the blade; Cl is the lift coefficient, Cd is the drag coefficient.

The differential torque act on a blade section is

1 dTrdFF sin  F cos    WCrC2 ( ) (3.6) T L D2 y

The shaft power is calculated via total torque and rotor angular speed

r P T   dT  (3.7) m 0

PCPw p m (3.8)

where Pw is wind turbine production power, Ω is rotor speed; Pm is shaft power; Cp is the power coefficient.

In summary, the driving force on a wind turbine is generated by lift force when the wind flows past the airfoils. The lift force increases with attack angle, which is also accompanied by increases in undesirable drag force. While the tangential component of lift force supports blade rotation, drag force opposes it at the same time. Therefore, a wind turbine will achieve the best performance when the ratio of lift force to drag force is maximum, or at its optimum attack angle. Airfoil cross sections are aligned in a way to operate at close to optimum attack angle. The torque is dependent on the blade section chord length (C), and the relative inflow wind velocity W, which varies along the blade length. They are also dependent on the air density. The power output can be calculated by multiplying the rotational speed and the torque acting on blades. The procedure is also illustrated in Figure 3-6.

Figure 3-6 Flowchart for calculating blade production power using BEM theory.

3.3.4 Validation of BEM Model in Power Output Prediction

For implementing the BEM model analyses, the 10.5 m prototype blade is divided into

30 sections each with a width of 0.35 m. The number of section and the section width are determined based on the results of a sensitivity study, which achieves computational efficiencies while ensures the accuracy.

The performance of the developed BEM model in power output prediction is firstly validated by utilizing the monitored power production data from the 100 kW small wind turbine described in Section 3. The data used in this study was collected at 10 min time interval between September 2014 and August 2015, which includes the air density, the rotational speed of the blade, the wind speed at hub height, and the power output. Measured wind speed, blade rotational speed, and air density are used to calculate the power output using the described BEM model. The monitored power outputs at different wind speeds are compared with those predicted by the BEM model in Figure 3-7. Also shown in this figure are the curve fitting of the measured or BEM model predicted power output. In general, the predicted power output performance matches well with the measured data. The monitored total power output and total energy output during the one-year period is 388.87

MW and 64.8 MWh respectively.On the other hand, the power output and energy production predicted by the BEM model is 397.52 MW and 66.25 MWh respectively. For wind speeds under 6 m/s, the curve fitted turbine power curve from the BEM model prediction is slightly beneath that from the monitoring power production data; the trend reverses for wind speeds larger than 12 m/s. One of the causes is the limited amount of data available at high wind speed range. Overall, the maximum error between the BEM model’s 70

predicted output and monitored data is within 2.2%. The comparison with the monitored wind turbine power output data validated that the BEM model is accurate in predicting the wind energy output. Subsequently, the validated BEM model was utilized to analyze the performance of the proposed extensible blade in the subsequent section.

Figure 3-7 Comparison of the blade element momentum (BEM) model’s predicted

power output and the monitoring power output of the 100 kW wind turbine

3.4 Structural Analysis

3.4.1 Determine the Working Range of Wind Speed

The extended blade is subjected to a higher wind load. Therefore, determining the range of working wind speed is firstly conducted to ensure the safety of the blade. Since

the focus of this study is to assess the feasibility of the extensible blade for improving energy production, simplified mechanical analyses are conducted rather than sophisticated evaluations. The maximum allowed working wind speed is determined based on the corresponding bending moment on the extended blade, whose value should not exceed the bending moment of the original length blade at the cut-out wind speed [111].

With these criteria, the ranges of working wind speeds for two types of extensible blades are analyzed; i.e., (1) different extent of extension at blade tip; and (2) the different extent of extension in the middle of the blade. By using the BEM model, maximal in-plane and out-of-plane bending moments in the original blade and extended blades at different wind speeds are shown in Figure 3-8, and the intersection points are limits that determine the range of operational wind speed for the extensible blade. Ranges of safe working wind speeds corresponding to the different extension of the blade are determined, which are summarized in Table 3-2. As a note, from an operation perspective, the scheme of extension is designed to be simple (i.e., extension at steps of 25%, 20%, 0%) so that blade extension is not too frequent to conserve the energy needed for blade actuation. More sophisticated extension schema can be designed based on further analyses of wind characteristics.

(a)

(b)

Figure 3-8 (a) Determination of the wind speed range for (a) blade extension at tip

and (b) blade extension in the middle.

Table 3-2 Blade extends type in the research.

Extension Method Wind Speed Range (m/s) Extension (%)

3–10 25

Extensible blade with tip extension 10–14 20

14+ 0

3–10 20

Extensible blade with middle extension 10–14 10

14+ 0

3.4.2 Modal Analysis

A modal analysis determines the vibration characteristics (natural frequencies and mode shapes) of a structure. The natural frequencies and mode shapes are important parameters affecting the response and design of a structure for dynamic loading conditions.

A good design for reducing vibration is to separate the natural frequencies of the structure from the harmonics of rotor speed [112]. The modal analysis of the extensible blade helps us understand how the natural frequencies change, thus avoid resonance when the large amplitudes of vibration could damage the wind turbine.

The FEM software COMSOL® (COMSOL 5.0, COMSOL, Inc., Burlington, MA,

USA) is used to calculate the un-damped modal characteristics of the turbine. The wind turbine blade is considered as a cantilever beam with blade root fixed. The program solves the following eigenvalue problem [113] utilizing the model’s stiffness and mass matrices.

[KM2 ]{ } {0} (3.9)

Eq. (3.9) is a typical real eigenvalue problem; therefore, ϕ has a non-zero solution if the value of its determinant coefficient is zero.

Typically, only the first few natural modes are of interest for structural engineering design as they typically contain most of the modal mass and have natural frequencies close to the excitation frequency of the wind. In this research, only the first four natural frequencies are considered; as the finite element model considered here is a simplification of the structure intended to capture global structural dynamic demands, the higher mode results will likely be less accurate. For comparison purposes, both the tip-extend and middle-extend strategy are extensions of 20% of its length.

Table 3-3 presents the results of modal analysis with the first four modes. Overall, increasing the length of the blades reduces its natural frequencies. From the results of the modal analysis, the dominant vibration mode for the horizontal across wind direction has a natural frequency of 1.356 Hz for original length blade, 0.8982 Hz for middle extend

20% blade and 1.104 Hz for tip extend 20%. The natural vibration mode shapes are shown in Figure 3-9.

Table 3-3 Modal frequency.

Model Original Length Middle Extend 20% Tip Extend 20%

Shape (Hz) (Hz) (Hz)

1 1.356 0.898 1.104

2 5.267 4.044 2.648

3 6.443 4.221 5.336

4 12.147 8.892 6.025

(a)

(b)

(c)

Figure 3-9 The shapes of first four modes for (a) original length blade; (b) blade-

extension of 20% at the tip; (c) blade extension of 20% in the middle

3.4.3 Rainflow Cycle Counting Method

Fatigue analysis in a variable amplitude load environment such as wind loading on a wind turbine requires extraction of stress (or load) cycles from non-harmonic response histories. This requires the use of a cycle counting method to quantify fatigue damage, which can be related to establishing S-N curves for fatigue design or evaluation. A cycle counting scheme must be used to reduce a complex irregular time loading history into a series of constant amplitude events. Cycle counting methods were initially developed for fatigue damage studies in aeronautical structures [114]. Various cycle counting methods are used including level crossing counting, peak counting, simple range counting [115] and rain-flow counting [114]. This research applies the rain-flow cycle counting method as defined in [116].

The rain-flow cycle counting method was initially proposed by M.Matsuiski and

T.Endo [114] to count the cycles or the half cycles of strain-time signals. The concept is 78

illustrated in Figure 3-10 where a load-time history is shown in Figure 3-10(b) with two small cycles bcb’ and efe’ and one big cycle adg according to rain-flow method. The corresponding cyclic stress-strain curve is shown in Figure 3-10(a). The rain-flow counting method rule can be qualitatively described as follows: a) The rain flow begins at the inner side of each peak or valley and it will stop if it meets

the opposing peak or valley larger than a departure one, e.g. a drop starts from a will

stop at d which consists a half cycle ab’d. b) The rain flow will also stop if it meets another rain flow drop from the top, e.g. a drop

starts from c will stop at b’ which consists a half cycle cb’. c) Following the upper two steps, recorded all full cycle numbers and amplitudes. And

leave the uncoupled cycles. This is called the first stage. d) Combination the uncoupled cycles from the first stage to couples and make the second

stage count of the combined couples. The total cycles are the summation of the first

and second stage.

(a)

(b)

Figure 3-10 (a) Stress and strain cycles; (b) Load time history

By applying the wind speed history data to the numerical simulation wind turbine blade model, the bending moment of the wind turbine blade in ten minutes’ interval is calculated. Fatigue damage is measured through cycle counts of the wind turbine blade root overturning. Thus, in this research, a MATLAB rain-flow program is used to produce cycle count numbers for different amplitude bins of the analytical base overturning moment history. The program is developed by [117] and can be reached at [118]. The cycle counting results of the analytical model for the three types of wind turbine blades is shown in Figure

3-11. It shows from the figure that there is a shift of the bending moment of the two extend blade types compare to the original length blade which means the extend wind turbine blade increases the blades’ bending moment significantly.

(a)

(b)

(c)

Figure 3-11 Rain-flow cycle count for (a) Original length blade; (b) Middle extend

20% blade; (c)Tip extend 20% blade

3.4.4 Fatigue Damage Analysis

For Typically, S–N (stress amplitude versus a number of cycles to failure) curves have been used to estimate the fatigue failure of a material. In a variable amplitude load environment, Miner’s rule [119] is often applied to quantify fatigue demands. It states that if there are k different stress levels and the average number of cycles to failure at the ith

stress, Si, is Ni, then the damage fraction, Cf, is:

k ni   C f (3.10) i1 Ni

Si: Stress range in bin i

ni: Number of stress cycles accumulated in the bin i

Cf: Fraction of life consumed by exposure to the cycles at the different stress levels.

Constant dependent on the fatigue critical detail.

Fatigue damage index is used for quantifying the damage, which is defined as the product of stress and the number of cycles operated under this stress, which is:

WNM i i i (3.11) i

Wi : Fatigue Damage index

M i : Average bending moment for bin i

Ni : Number of bending moment cycles in bin i

According to the premier discussion, the three different wind turbine blade fatigue damage index is calculated as shown in Table 3-4. The result shows the middle extend 20% and the tip extends 20% will both increase the fatigue damage about 78% compare to the original length blade.

Table 3-4 Fatigue damage index

Fatigue Damage Blade Type Increment Index[Cycle×(N×m)]

Original Length 1.54E+06 NA Middle Extend 20% 2.74E+06 78.00% Tip extend 20% 2.74E+06 77.90%

The result indicates that the innovative wind turbine blade will decrease the natural frequencies of the blade. This will give a guidance of the wind turbine manufacturer to avoid resonance when using longer wind turbine blade. The paper then uses one year recorded data from DAQ system in the prototype wind turbine to calculate the fatigue damage of the innovative blades. The result shows that this innovative blade could increase 84

the fatigue damage of the blade about 78% compare to the original length blade. It is recommended to use higher strength material to build longer wind extensible turbine blades.

3.5 Test Results

With the extension strategy defined by base structural safety considerations, which are summarized in Table 3-2, Figure 3-12 compares the corresponding power output curves of the extensible blades with that of the original blade. Both the manufacturer’s power curve and the power curve from a curve fitting of the monitoring data are plotted for comparison purposes. It is noted that the power curve from the monitoring data does not cover a high wind speed range. The predicted power curves of the extensible blade by the

BEM model with two different extension strategies are also plotted. The comparison clearly shows that the extended blade has a much higher power output at wind speeds lower than the blade’s rated wind speed of 14 m/s. There are different extents of shift in the power production curves at wind speed of 10 m/s is due to the proposed blade extension strategy that changes the extent of blade extension at 10 m/s (Table 3-3). It is assumed that the maximum output is limited to 100 kW to match the capacity of the generator (modified power curve shown in Figure 3-12). It can be seen from the figure that the power curve from the BEM model and the power curve from the monitored data are closer to each other but different from the manufacturer’s power curve. This might because the manufacturer’s power curve is measured under certain meteorology conditions which are different to the

real conditions [120]. The blade production curves of extensible blades are similar to the original blade as they are completely retracted to the original length.

Figure 3-12 Comparison of the power curves for the original length blade, tip-

extended blade and middle-extended blade.

Wind speed of three sites as described in chapter two are used to calculate the wind power output between different wind turbine types. The wind speed distribution at 37.5 m height in a typical year is shown in Figure 3-13 for each of the three sites. It can be concluded from Figure 3-13 that the wind speed distribution at the same location has slightly bias in different years, but the overall trend is similar. The wind speed at site A is more concentrated with the highest frequency at 4 m/s, and the wind speed at site C is more

distributed, varying from 0 m/s to 25 m/s. Overall, site A has the lowest mean wind speed since the site is surrounded by a few buildings with heights of up to 20 m. The wind speed of site B is slightly higher because the site is located in a rural area that most of the surrounding buildings are under 10 m in height. The offshore site C has the strongest wind speed since the terrain is flatter at the offshore location. Since the wind speed at the three selected sites was under 14 m/s for most of the time (Figure 3-13), it is expected that a turbine with extensible blades will consistently produce more energy than a regular turbine for the majority of the year. The annual energy output of the extensible and original length blade can be calculated using the corresponding power curves and the wind speed data at the three test sites.

(a)

(b)

(c)

Figure 3-13 Weibull distribution of wind speed data at 10 min intervals at (a) Site A; 88

(b) Site B; (c) Site C.

The monitored yearly wind speed data at the three locations with the low-class wind

(A and B: Class 4 and C: Class 3) are utilized to estimate the total wind energy outputs, following the validated procedures described in the earlier context. Figure 3-14 shows the histogram of the predicted average power output in 10 min intervals for each site for different years with the baseline blade and extensible blades. Overall, the comparison shows that the original baseline blade has a higher occurrence of low-energy output periods than the extended blades. In another word, the extensible blades shift the wind energy production to higher energy output than an original blade for these sites with a low class of wind resource.

(a)

(b)

(c)

Figure 3-14 Statistical distribution of 10 min of energy output for the original blade

and extensible blades at (a) Site A; (b) Site B; (c) Site C.

Table 3-5 summarizes the predicted total annual energy production by different types of blades (original versus extensible blades). The results show that the innovative extensible wind turbine blades will potentially increase the total annual wind energy production for all sites with a low class of wind. For Site A, the extensible blade that

extends at the tip will increase the power output by around 19%; the extensible blade that extends in the middle will increase the power output by 32%. For Site B, the corresponding increases in total energy production are 22% and 31% by the two types of extensible blades.

For Site C, the amount of increase in the annual energy production for tip extension and middle extension blades are around 19% and 25% respectively. The extensible blade that extends in the middle provides a larger increase in the energy output than that extends at the tip due to the larger wind carry areas; besides this, the percentage increase in energy production is more significant at site with a low class of wind (i.e., Sites A and B) than site with high class of the wind (i.e., Site C). These are a clear demonstration of the benefits of the extensible blade to boost energy production for a site with low classes of wind. In the meantime, the extension scheme is designed so that the extensible blade is protected with a similar structural safety to a regular blade.

Table 3-5 Comparison of total annual energy production by the original blade

versus extensible blades at different sites.

Energy Energy Energy Increase Output by Increase Output by Output by Year Percentage Middle Percentage Original Tip Extended (%) Extended (%) Blade (kWh) Blade (kWh) Blade (kWh)

Site A (Class 4, Onshore)

2011 80,316.25 95,692.79 19.14 106,387.10 32.46

2012 73,393.83 87,562.00 19.30 97,316.17 32.59

2013 78,277.17 92,783.17 18.53 103,071.30 31.67

2014 75,296.00 89,820.17 19.29 99,600.17 32.28

Site B (Class 4, Onshore)

2010 44,819.83 55,296.00 23.37 59,269.00 32.24

2011 70,515.67 85,525.50 21.29 92,656.17 31.40

2012 93,464.50 114,019.3 21.99 121,786.50 30.30

Site C (Class 3, Offshore)

2006 308,698.30 368,768.30 19.46 387,289.30 25.46

2007 314,257.70 372,854.80 18.65 390,657.80 24.31

2008 322,774.70 382,870.50 18.62 401,353.30 24.34

2009 305,493.20 364,615.70 19.35 382,645.30 25.25

2010 239,212.20 284,880.20 19.09 297,576.30 24.40

3.6 Conclusions

Wind farms are ideally located at locations with high-class wind. However, there are a large number of distributed wind turbines constructed at sites close to communities, with non-ideal wind conditions. This paper describes the analyses of an innovative, 94

extensible blade technology that aims to utilize wind energy in areas with low-class wind resources. The extensible blade functions by adjusting its length depending on the wind conditions (i.e., it will extend at low wind speed and retract at high wind speed). Based on the principle that the larger the sweep area, the higher the turbine energy output, dynamically adjusting the blade length helps to increase the energy output under low wind speed while mitigating safety risks under high wind speed. The computational model is developed based on the blade element momentum (BEM) theory, which determines the aerodynamic load and power output of the blade at different wind conditions. The model is firstly validated with monitored energy output data of in-service wind turbine. The validated model is subsequently used to estimate the annual energy production by the extensible blades and regular blade at three locations inland and offshore of the Lake Erie area, where yearly wind data are continuously monitored. Two types of extensible blade scheme are analyzed; i.e., extension in the middle of the blade versus extension at the tip of the blade. The extension and contraction scheme of these extensible blades are determined based on a limiting of the maximum bending moment acting on the blade, which helps ensure their structural safety. The influence of blade extension on the dynamic characteristics of blade structure is analyzed. The results show that the extensible blade

will potentially increase annual energy output up to 20% to 30% for the sites analyzed.

Besides this, the lower the wind speed, the more effective the extensible blade in increasing energy production. Overall, the results of this paper point to the promise of this innovative, extensible blade in improving the wind energy production.

CHAPTER FOUR. EXPERIMENTAL STUDY ON THE PERFORMANCE OF

SMALL HORIZONTAL AXIS WIND TURBINE WITH BIO-INSPIRED

BLADE

4.1 Overview

This research investigates the potential of bio-inspired blade technology to improve the performance in increasing wind energy output for the small horizontal axis wind turbine.

The high lift low Reynolds number airfoil S1223 was chosen in this research, and the wind tunnel test was conducted in the Control & Energy Systems Center at Case Western

Reserve University established by Professor Mario Garcia Sanz’s group. Seven types of wind turbine blade, including traditional blade and bio-inspired blade with different wavelength and amplitudes of tubercles, are studied in this research. The experiments were performed with the 3D printed rotor blade models in the wind tunnel to obtain the maximum power coefficient. The result shows that the blade with a shorter wavelength and larger wavelength of tubercles has better performance in increasing the maximum power coefficient. Additionally, the tubercles can delay the stall significantly comparing to the reference blade. The rotor blade models were also tested to produce power under different wind speed during a 200 seconds test period. Finally, a series of numerical simulation were

performed to study the mechanism of the tubercles in effecting the wind turbine blades.

The results indicate that the tubercles can increase the power performance of the blade at low Reynolds number, and the energy output is enhanced up to 4.44% comparing with the traditional blade model in the test period by simulating the real wind turbine system.

4.2 Introduction

Although the tubercle technology has already been studied by a significant number of researchers as introduced in chapter one, the previous literature shows that most studies are carried out only on a few of the NACA 4 and 6 series airfoils since its shape is most similar to the cross-section of the humpback whale flipper [121]. The systematic and quantitative investigations on the effects on the rotor blades of the small-scale HAWT system are still needed. Studies should be carried out on the effect of tubercles on low

Reynolds number small-scale wind turbine airfoils. In addition, the tubercles in the former studies are uniformly distributed. However, it can be observed that the amplitude and the wavelength of the tubercles on the flipper of the whale flippers are not evenly distributed in practical. Furthermore, most of the former studies, which included wind tunnel tests and numerical simulations, only tested one section of the blade with no twist and a uniform chord length. The section of the blade is tested to plot the lift and drag coefficient at 98

different attack angles. The power efficiency increment and thus the power output cannot be extrapolated directly. In some studies carried out in the wind tunnel, the torque increment, and thus the power coefficient was monitored by using the blade model with uniform chord length and no twist angles. The real wind turbine blade, however, is twisted from the tip to the hub of the blade. This is because the angular speed of each blade element is different when the blade is rotating at working conditions. The resultant speed of each blade element is therefore toward the different direction. The twist leads the resultant blade to act at the optimized attack angle, which has the best power coefficient.

In this study, the high lift low Reynolds number airfoil S1223, which is suitable for small wind turbines, is selected in this research, and the wind tunnel test was conducted in the Control & Energy Systems Center at Case Western Reserve University. Seven models of wind turbine blade, including reference blade without tubercles and bio-inspired blade with different wavelength and amplitudes of tubercles are studied. The blade with variable wavelength and amplitude tubercles inspired by real humpback whale are also tested. The blade models tested in this research are designed using the open source blade design software QBlade [122], which is based on the Blade Element Momentum (BEM) theory.

The models are optimized with variable chord length and twist angle to simulate the blade

of the real wind turbine. The power coefficient curve of each model is plotted and compared with the reference model. The energy output of each model is also tested under varies wind speed in a 200 seconds period to simulate the real working condition of the wind turbine.

The numerical simulation based on the commercial CFD software COMSOL® are also conducted to investigate the mechanism of the tubercles on the leading edge of the blade.

This study provides a systematic experimental study in investigating the tubercles on the leading edge of the wind turbine blade for small wind turbine. The goal of this research is to study the potential of bio-inspired blade technology to improve the performance in increasing wind energy output for the small horizontal axis wind turbine.

4.3 Setup of Wind Tunnel Tests

4.3.1 Wind Tunnel Components

The wind tunnel tests were performed in the Low Reynolds number wind tunnel in the Control and Energy Systems Center at Case Western Reserve University (CWRU-

CESC) [122]. A schematic of the test package is depicted in Figure 4-1. The wind tunnel is an effective method for investigating the aerodynamic performance of the wind turbine.

This method tests small-scale models under higher wind speeds in order to ensure a satisfactory Reynolds number during the tests. The real wind turbine are usually costly and 100

are constrained by the manufacture design standard; thus, a wind tunnel test provides reliable experimental results for concept proofs and numerical modeling calibrations. Wind tunnel tests have previously been applied in investigating the performance of wind turbine blade in many studies [123-125] and proofed to be reliable.

Figure 4-1 Lab scale wind tunnel schematic

As shown in Figure 4-1, the inlet of the wind tunnel consists of honeycomb, which aims to align and smooth the airflow before it enters the contraction section. The honeycomb can also reduce the turbulence level of the airflow. The contraction at the inlet

101

ensures a uniform flow velocity profile with a minimum boundary layer thickness [126].

The contraction gently accelerates the air from the honeycomb section and guide it into the test section. Turbulence intensity of the airflow is further reduced in this process. The test section area is 1200 mm (long) ×800 mm (high) × 700 mm (wide), offering the wind speed range of 3 m/s to 20 m/s [127]. The diffuser is used to extract the air out of the wind tunnel and gradually slow down the wind speed. Moreover, it can reduce the dynamic pressure, i.e. kinetic energy and increase the static pressure. The angle included by the diffuser walls is generally limited up to a small degree, which leads to no flow separation inside the diffuser section. The wind source is produced by a 3-phase variable speed motor, controlled by the Variable Frequency Drive (VFD. It is controlled by the computer through the commercial software MATLAB (MATLAB® 2017a). The detail of this wind tunnel can be refer to Dr. Fa Wang’s Thesis [122].

4.3.2 Wind Tunnel Control System

Figure 4-2 schematically depicts the control system of the wind turbine in this study.

The control system was developed by Professor Mario Garcia Sanz’s group and the detail can be refer to [128]. The wind turbine system includes a gearbox, an electrical generator, a power electronics circuit, and a controller with the Maximum Power Point Tracking 102

(MPPT) algorithm, which is widely used in wind energy systems; its main goal is to maximize the energy capture under all wind conditions [129].

The wind turbine system is controlled by a Supervisory Control and Data Acquisition

(SCADA) system，which contains a central real-time control microprocessor and an external computer with MATLAB. When the wind speed is low, the wind turbine is operated below the rated power. The SCADA system will then use the MPPT algorithm to change the electrical torque of the generator in order to achieve the maximum power coefficient. The detail of this control system can be refer to the previous studies by Wang et al. [127].

Figure 4-2 Block diagram of wind turbine system (adjust from [127]) 103

4.4 Geometrical Design of Blade Models

4.4.1 Model Description

The tested wind turbine model is a three-bladed horizontal axis wind turbine

(HAWT) as shown in Figure 4-2. The blade model tested in this research is designed in

the commercial design software Rhinoceros® 5 and fabricated by the 3D printer

Ultimaker® 2+ as shown in Figure 4-3 (a) . The nozzle size of the 3D printer is chosen

as 0.4mm, and the highest resolution is 20 micron. The material is polylactic acid (PLA),

which provides good surface quality (Figure 4-3(b)). S1223 airfoil was selected in this

research as the prototype airfoil as shown in Figure 4-4. S1223 is a high lift airfoil with

max thickness 12.1% at 19.8% chord of the blade. The lift and drag coefficients of the

S1223 airfoil was tested through the wind tunnel test in the University of Illinois open

return wind subsonic wind tunnel [130]. The blade has good lift and drag characteristics

at low Reynolds number, and is perfect for small wind turbines [131]. The length of the

blade is 140 mm, and the chord length of the blade varies from 20 mm to 35 mm along

the blade. Wavelength and amplitude of the tubercles are the two parameters that

control the shape of the tubercle. According to previous studies [132-135], both

amplitude and wavelength of the tubercles on the blades affects aerodynamic

104

performances of the blades. The previous studies conducted on the NACA 634-021 airfoil, which resembled the whale flipper, shown that tubercles with small amplitude and large wavelength had better aerodynamic performance [136]. However, in another study performed by Hansen et al. [137], the airfoils NACA 0021 demonstrated the smallest amplitude and smallest wavelength tubercle showed superior performances.

Meanwhile, for the study conducted on the delta wing [138], the authors found the low amplitude and low wavelength tubercles were the optimum. In another study conducted by Selig et al. [139], the performance of the flat plate airfoil had a significant increase for the larger amplitude and smaller wavelength model. These former researches indicate that the influence of the wavelength and amplitude varies with airfoil shape.

The low Reynolds airfoil S1223 has not been studied previously. Therefore, it is necessary to study the influence of different amplitude and wavelength to the power performance on the airfoil S1223.

105

(a)

(b)

Figure 4-3 (a) Ultimaker® 2+ 3D printer; (b) 3D printing material

106

Figure 4-4 S1223 foil section

Figure 4-5 Wind turbine Models

107

(a)

(b) 108

(c)

(d) 109

(e)

(f) 110

(g)

Figure 4-6 Rotor blade models

In this study, the baseline type of S1223 airfoil (Figure 4-5(a)) is tested.

Moreover, as shown in Figure 4-5, six bio-inspired blades with different shapes were designed, and they have sinusoidal protuberances at the leading edge with different amplitudes and wavelength. A parameter c is introduced to describe the shape of the tubercles on the blade, which represents the chord length of the blade cross-section.

Model R1 and R2 represent blades with a wavelength of 0.5c and amplitudes of 0.05c and 0.1c. Model R3 and R4 are blades with a wavelength of 2.5c and amplitudes of

111

0.05c and 0.1c, respectively. These ranges correspond to the morphology found on the leading edge of humpback whales’ flippers which can be referred to the study of Johari et al. [136]. It is observed from the real humpback flippers that the tubercles size are not uniformly distributed as shown in Figure 1-5. The amplitude of the flippers seems randomly distributed. Two ununiformed blade shapes R5 & R6 are inspired by real humpback whale flipper, and they are fabricated in this research for comparisons.

Model R5 is a blade with a wavelength of 0.5c and amplitudes of both 0.05c and 0.1c interval. Model R6 is a blade with a wavelength of 0.25c and interval amplitudes of both 0.05 and 0.1. Table 4-1 lists parameters of the seven models tested in this study.

Table 4-1 Characteristics of the test models

Model number Wavelength Amplitude

R0 NA NA

R1 0.5c 0.05c

R2 0.5c 0.1c

R3 0.25c 0.05c

R4 0.25c 0.1c

R5 0.5c 0.05c &0.1c

R6 0.25c 0.05c &0.1c

112

4.4.2 Blade Design

In the procedure of turbine blade designs, the drag force of the blade is reduced by twisting the blade along its length. Twisting the blade changes the inflow angle along the blade. With the combined effects of twisting and tapering of the blade along the blade length, the efficiency of the blade increases while the drag force reduces. The twist angle of the blade in this study is designed using the open source design software of wind turbine blade QBlade [140]. The software utilizes the Blade Element Momentum (BEM) theory to calculate the airfoil parameters for each cross-section of the blade. BEM theory is composed of two different theories, i.e., blade element theory and momentum theory [104].

Blade element theory assumes that blades can be divided into small elements that act independently of the surrounding elements and operate aerodynamically as two- dimensional airfoils. The characteristics of blade responses (drag and life on each element) are determined by the angle of attack of incoming wind, which is the angle between the center reference line of the geometry and the relative incoming flow. The momentum theory assumes that the loss of air pressure or generation of turning momentum in the airfoil blade element is caused by the work done by the incoming airflow [105]. The BEM theory couples these two theories together and calculates the total lift and momentum via an iterative process [106]. Li and Yu [10] introduced the of BEM theory in detail with a flowchart. The twisting blade profiles about the quarter chord are obtained by optimizing the airfoil parameters as shown in Figure 4-7.

113

Figure 4-7 The blade twisting profiles about the quarter chord

4.5 Experimental Setups

The theoretical power output of a wind turbine is described in Eq. (2.1). The equation shows that, at a certain inflow wind speed and air density (which are primarily decided by the climate condition and the topology of a particular site), the power output of a wind turbine is dependent upon its power coefficient and the rotor swept area. The power coefficient is decided by the mechanical structure of the rotor, with the theoretical maximum given by the Benz limit. Incremental improvements of the power coefficient are continually being sought by the engineers by changing the design of the blade. The rotor swept area is decided by the length of the blade. The power coefficient of a rotor varies with the tip speed ratio (TSR), which is the ratio of rotor tip speed to free wind speed as described in Eq. (4.1), and it is only a maximum for a unique tip speed ratio [141]. This is 114

because most of the wind will pass undisturbed through the openings between the blades with only litter power extraction if the rotor of the wind turbine turns too slowly. However, the rotating blades will act a solid wall obstructing the wind flow thus reducing the power coefficient if the rotor turns too fast. Wind turbines must thus be designed to operate at their optimal wind tip speed ratio in order to extract as much power as possible from the wind stream.

speed of rotor tip R   (4.1) wind speed V

In which R is rotor radius (m), and Ω is the rotational speed (rpm). It shows clearly from the equation that it is possible to maintain the maximum power coefficient over a range of wind speeds by adjusting the rotational speed of the blade. When the turbine operates below its rated power, the methods that achieve maximum power are known as maximum power point tracking (MPPT) strategies [142]. The main control objective when using a variable speed wind turbine in the modern wind farms is to maximize the power generation at every wind speed. In this study, the blade was tested using the maximum power point tracking (MPPT) under variable wind speeds.

Two series of experiments were conducted in this research. The first group of experiments is conducted under varies the rotational speed of the blade to display the power

115

coefficient as a function of tip speed ratio (TSR). This is because the maximum power coefficient of each blade design depends on the particular rotor airfoil profile. During the tests, the income wind velocity of the tunnel was fixed, and the rotational speed was varied to achieve the different TSR. Afterwards, the TSR at maximum power coefficients of each model was input into the MPPT algorithm to run the second experiment. The second experiment aimed to test the power output of the seven models under varies wind speed in a period of 200s. The wind speed began at 7.5 m/s for the first 80s, and then the wind speed increased to 8.3 m/s until 140s. Finally, it dropped to 8 m/s until 200s. The variation of the wind speed is depicted in Figure 4-8. The purpose of the second experiment was to simulate the power performance between different models under varies wind speed conditions. The compaction of the power output between different models clearly illustrated the function of the tubercles for the wind turbine under real working conditions.

116

Figure 4-8 Wind speed variation

4.6 Experimental Results

4.6.1 Power Coefficient Curve

In this section, the effect of the leading edge tubercles to the power coefficient is analyzed. Each turbine models were tested by method described in section 4. The experiments were conducted at the wind speeds of 8 m/s, and the correlated Reynolds number is 7.7 ×104. The experiments aim to test the effects of tubercles with different amplitudes and wavelengths (R1-R6) by comparing with the reference model R0. The test

117

results are plotted in Figure 4-9. The Cp curves marked in red is the power coefficient curve of the reference blade without any tubercles, and the black lines represent the models with the leading-edge tubercle. The blue line, which represents the maximum value of the

reference model, is also plotted for comparison. The trend of the Cp curve for all models is similar that the power coefficient increases at lower TSR and drops significantly after the critical value of TSR. The effects of leading edge on the resulting power coefficient values

Cp can be clearly observed from the figure. By comparing with the reference model, the

bio-inspired blades with tubercle lead to better Cp values for most TSR. It can be concluded that the tubercles play an important role in increasing power coefficient values, especially at the higher TSR. For lower TSR, however, the power coefficient of some blades with tubercles are even smaller than the reference blade as shown in Figure 4-10.

In the upper left figure, the Cp of R0 reaches the maximum value of 0.236 at TSR equals 2.83; the R1 model achieves the maximum value at 0.240 with a small delay when

TSR equals 2.9. Similar to the results of the upper left figure, the maximum power coefficients of all models with tubercle are enhanced compared to the reference model. On

the other hand, the TSR values are all delayed except for the R1 model. The maximum Cp values and the correlated TSR are summarized in Table 4-2. By comparing R1and R2, R3

118

and R4, respectively, it can be observed that the larger amplitude has better performance in increasing the power coefficient. Therefore, it is concluded that the maximum power coefficient increase with the amplitude. This phenomenon is more obvious when comparing R1, R2, and R5; R3, R4, and R6. On the other hand, the shorter wavelength seems better than the longer wavelength by comparing R1 and R3. The maximum power coefficients of R2 and R4 are the same, which does not show any benefit of the difference in the wavelength. However, it can be observed from Figure 4-10 that the power coefficients of R4 are higher than that of R2 at larger TSR. The same situation can be found

on R5 and R6 with the same maximum power coefficient but higher Cp at larger TSR. This means the larger amplitude can delay the TSR. The two ununiformed blade shapes inspired by real humpback whale flipper follows the same effects of the amplitude and wavelength.

The maximum power coefficient of R5 is in-between R1 and R2, while the maximum power coefficient of R6 is in-between R3 and R4. The first wind tunnel test indicates that the tubercles can both increase the maximum power coefficient and delay the TSR compare to the reference blade. After all, a conclusion can be drawn that the maximum power coefficient increased more for the models with a larger amplitude and shorter wavelengths.

119

Table 4-2 Maximum power coefficient of the blade models

Maximum Cp Model TSR at maximum Cp Maximum Cp Increment (%) R0 2.83 0.236 NA R1 2.83 0.239 1.27 R2 2.90 0.240 1.69 R3 2.85 0.238 0.85 R4 2.90 0.240 1.69 R5 2.85 0.239 1.27 R6 2.92 0.239 1.27

(a)

120

(b)

(d)

(e) 122

(f)

Figure 4-9 Power coefficient for the rotor models

123

Figure 4-10 Power coefficient increment

4.6.2 Power Output

The TSR at the maximum power coefficient of each model can be achieved from the first experiment. In this section, the TSR is input in the MPPT algorithm as described in section 2 to maintain the power coefficient of each blade in a range near the maximum power coefficient. This experiment gives us the ability to simulate the real wind turbine working conditions, and it aims to simulate the power output performance between different models. Figure 4-11 plotted the variation of the power coefficient for each model

124

during the 200s test. The red dots is the power coefficient of the reference blade, and the blue dots is the power coefficient of blades with tubercles. The variation of the power coefficient is due to the system vibration, wake turbulence, mechanical load, etc., and the

MPPT system needs to adjust the TSR to control the power coefficient of the wind turbine

to reach the maximum Cp. It shows from Figure 4-11 that the MPPT system works well as the power coefficient was controlled over a contain range around the maximum power coefficient. The disperse dots at 80 s and 140s are caused by the change of the wind speed as described in Figure 4-8. Overall, the figure illustrates that the power coefficient of the entire blade with tubercles are higher than that of the reference blade, which is consistent with the first experiment described in section 4.6.1.

125

(a)

(b) 126

(c)

(d) 127

(e)

(f) 128

Figure 4-11 Power coefficient in time domain

Figure 4-12 displays the variation of the measured mechanical power in 200 seconds test time. The power curve marked in red is the curve of the reference blade without any tubercles and the blue lines represent the models with the leading-edge tubercle. The power variation flows the same trend as the wind speed variations depicted in Figure 4-8, and the power output increases with the wind speed within the rated power of the generator.

The noise of the curve is due to the variation of the power coefficient as shown in Figure

4-11. Overall, it can be observed that the mechanical power of all the models with tubercles is above the mechanical power line of the reference model for most of the time. The total energy output of all models is summarized in table 4-3. It shows that all models with tubercles on the leading edge could increase the energy efficiency under the 200s period with variable wind speed. Model R4 has the highest energy output of 374.93 J with an increment of 4.44% compared to the reference model. The second optimized model is R6 with an increment of 4.38% of the total energy output. By comparing R1, R2, and R5; R3,

R4, and R6, respectively, it can be observed that the blade with larger amplitude has better energy output performance. On the other hand, the blades with shorter wavelength have better performance than those with longer wavelength by comparing R2 and R3, R2 and

129

R4, R5 and R6, respectively, which correspond with the results of the first experiment.

(a)

130

(b)

(d)

(e) 132

(f)

Figure 4-12 Mechanical power output in time domain

Table 4-3 Total energy output

Model Total energy output (J) Increment R0 359 NA R1 367 2.23% R2 367.47 2.36% R3 371.96 3.61% R4 374.93 4.44% R5 367.48 2.36% R6 374.74 4.38%

133

4.7 Numerical Simulation and Discussions

4.7.1 Introduction

In order to study the mechanism of the tubercles in influencing the wind turbine blade, a numerical simulation was carried out in this section. This investigation is conducted with the commercial computational fluid dynamic (CFD) software COMSOL®

5.2 by solving the discretized Reynolds Averaged Navier-Stokes equations. The K-epsilon

(k-ε) turbulence model is selected to simulate the wind flow in this study. The k-ε turbulence model mainly solves two variables, the turbulence kinetic energy (k) and the rate of dissipation of turbulence kinetic energy (ε). The k-ε model is commonly used for industrial applications since it has low memory requirements and the easy converge feature.

It also performs well for external flow around complex geometry problems.

4.7.2 Validation

For validation purpose, the simulation method is first validated with the former wind tunnel test. The validation is undertaken of an wind tunnel test carried out by

Gregorek et al [143]. The experiment was carried out in Ohio State University to test the aerodynamic performance of NACA 0021 airfoil. The Reynolds number is 106 according

134

to the following equation.

Re  cU 1 (4.2)

The parameters are chosen as the same in the wind tunnel test. The inflow wind speed U is 8.15m/s, air density is 1.225kg/m3, dynamic viscosity is 1.7965×105 (kg/m×s), and the chord length is 1.8m. A 2D computational domain was created and the mesh model is shown in Figure 4-13. Meshing strategy is a very important part of the numerical simulation whereby high quality meshes significantly improve computational accuracy and convergence rate [144]. In this research, the simulation model is meshed with finer elements near the blade model and coarser elements away from the blade model to increase the efficiency and maintain accuracy at the same time. The lift coefficient is compared with experimental study of [145].

135

Figure 4-13 Computational grid

The flow field of the airfoil model at different attack angle is plat in Figure 4-14.

For the attack angles 0° to 13°，the flow remains fully attach to the airfoil. When the attack angle increase to 14°, the vertex flow region develops at the trailing edge due to an adverse pressure formed at this region. This stall effect shows clearly in the lift coefficient.

L CL  1 2 (4.3) UA 2

where L is lift force,  is air density, U is inflow wind speed, and A is a projective area of the blade. The numerical simulation lift coefficient is compared with experimental study carried out by Rostamzadeh [145] as shown in Figure 4-15.

136

Figure 4-14 Velocity field at different attack angle

Figure 4-15 The comparison of simulation lift coefficient and experimental lift 137

coefficient

It also shows clearly in Figure 4-16 that the pressure coefficient of the airfoil in the simulation model is very similar to the pressure coefficient of the airfoil measured in the experiment According to Walter and Stuart’s study [146].

pp C   p 1 (4.4) U 2 2

In which Cp is the pressure coefficient, p is the static pressure at the point which pressure coefficient is being evaluated, p∞ is the static pressure in the freestream,  is air density, U is inflow wind speed.

138

(a)

(b) 139

(c)

Figure 4-16 Airfoil pressure coefficient at different attack angle (a) α=0; (b) α=6; (c)

α=12

4.7.3 Simulation Model of S1223 Airfoil

Meshing strategy is a very important part of the numerical simulation whereby high quality meshes significantly improve the computational accuracy and the convergence rate

[144]. In this research, the simulation model is meshed with finer elements near the blade model and coarser elements away from the blade model to increase the efficiency and maintain accuracy simultaneously. The mesh result is shown in Figure 4-17. The chord

140

length of the reference blade is 0.14m, and the aspect ratio is 3; the Reynolds number of the blade is the same as the previous wind tunnel test, and the amplitude and wavelength of the tubercles are the same as the models in the wind tunnel test. The models in the CFD simulation do not have twist angles since the inflow wind direction is constant and there is no rotation of the model.

Figure 4-17 Computational domain and meshed grid

141

The pressure distribution and the pressure gradient lines near the blade models are plotted in Figure 4-18 to study the mechanism of the tubercles. Two cut planes are used to present the pressure distribution near the blade surface. The first cut plan is at the margin of the leading edge, and the second cut plan is at the 0.1 times chord. By comparing the pressure distribution between the reference blade and the bio-inspired blades at cut plan 1, it shows that the tubercles alter the pressure distribution on the blade surface. By comparing the cut plan 2 of R0, R1, and R3, it can be observed that the pressure on the surface of the blades of the bio-inspired blade are different from that of the reference blade at 0.1 times chord. This indicates that the tubercles can redirect the flow at the surface of the blade and thus delays the separation of the boundary layer of the flow. This will then led to a gradual tendency of the TSR when rotation under working condition. Computational Fluid

Dynamics demonstrates that the tubercles on the leading edge could improve the pressure distribution on the surface of the wind turbine blade. This numerical analysis illustrate the basic mechanism of the test results and serve as a supplementary to the experiments.

142

(a)

(b)

143

(c)

(d)

144

(e)

(f)

145

(g)

Figure 4-18 Pressure distribution and the pressure gradient lines near the blade

surface of the models

4.7.4 Modal Analysis

A good design for reducing vibration is to separate the natural frequencies of the structure from the harmonics of rotor speed. The modal analysis of the bio-inspired blade helps us understand how the natural frequencies change, thus avoid resonance when the large

146

amplitudes of vibration could damage the wind turbine. The software COMSOL® is used to calculate the un-damped modal characteristics of the blade. The blade is considered as a cantilever beam with blade root fixed. The model that described in the last section is used here.

Table 4-4 presents the results of modal analysis with the first four modes. Overall, the tubercles could increase the natural frequencies of the blade. From the results of the modal analysis, the dominant vibration mode for baseline blade R0 has a natural frequency of

9.13 Hz. The first natural frequency of the blades are all increased except the R4 blade.

The natural vibration mode shapes are shown in Figure 4-19.

Table 4-4 Modal Frequency

Model Shape R0 (Hz) R1(Hz) R2(Hz) R3(Hz) R4(Hz) R5(Hz) R6(Hz) 1 9.13 9.81 9.32 9.74 9.11 9.37 9.25 2 55.23 58.49 56.29 58.1 55.07 56.27 55.61 3 58.23 61.01 57.64 60.54 56.3 58.05 57.04 4 66.72 67.89 65 67.53 63.87 65.44 64.71

147

(a) R0

(b) R1

148

(d) R3

149

(e) R4

(f) R5

150

(g) R6

Figure 4-19 The first four modal shapes of the blades

4.8 Conclusions

Currently, small-scale wind turbines are gaining more attention due to the fact that rising energy costs have driven the consumer to seek alternative solutions of independent power supply [58]. Increasing the energy harvesting efficiency of the small-scale wind turbine blade has become a popular topic. In this study, the bio-inspired wind turbine blade with tubercles on the leading edge has been proposed, and its power output performance has been investigated by means of a series of wind tunnel tests conducted in the Control &

Energy Systems Center at Case Western Reserve University. Seven groups of wind turbine

151

blades were studied in this research, which include one reference blade without any tubercles and six bio-inspired blades with different wavelength and amplitudes of tubercles.

The blades were designed based on the BEM theory with twist angles to optimize the angle of attack along the length of the blade, and they were fabricated using a 3D printer. Two series of wind tunnel test were conducted. In addition, a numerical simulation model was proposed to study the mechanism of the leading edge tubercles at low Reynolds number conditions. The following conclusions can be drawn from this study:

(1) The tubercles on the leading edge of the blade can increase the maximum power

coefficient of the small horizontal wind turbine airfoil at low Reynolds number. Figure

4-8 shows that the maximum power coefficients for all the bio-inspired blade are

enhanced regardless of the difference in size of the amplitude and the wavelength. It

can also be observed that the leading-edge tubercles can improve the performance of

the turbine for almost all ranges of the TSR.

(2) By comparing the six groups of the bio-inspired blade in the first wind tunnel test, both

the wavelength and the amplitude have the obvious influences of the power coefficient

of the bio-inspired blade. In particular, the blade with larger amplitude tubercles has

better performance in increasing the maximum power coefficient; the blade with a

152

shorter wavelength of leading-edge tubercles can increase the power coefficient of the

blade at higher TSR. Overall, the optimal model in the research is R4 with a larger

amplitude and a shorter wavelength.

(3) The second series of wind tunnel test was conducted in 200s period with varying wind

speeds. The results indicated that the power output for the models with tubercles is

higher at all time when compared with the power output of the reference model. The

bio-inspired blade can increase the total energy output up to 4.44% compared to the

reference model under low Reynolds number.

(4) The mechanism of the tubercles was studied by using a numerical simulation model.

This turbulent flow in this research is chosen as the K-epsilon (k-ε) turbulence model.

The pressure distribution on the blade near the tubercles indicates that the tubercles can

change the pressure distribution near the surface of the blade. The boundary layer is

then separated based on the different size of the tubercles. This can ultimately lead to

an increment of the power coefficient at higher TSR. The numerical simulation in this

research mainly qualitatively studied the mechanism of the tubercles on the leading

edge as supplementary to the experiment results, and more explicit numerical analysis

is expected to present in the near future.

153

CHAPTER FIVE. SUMMARIES AND FUTURE WORK

5.1 Summaries on This Research

This dissertation mainly contains four parts. It started with an introduction of the development of modern wind energy. The global energy capacity installation, the onshore and offshore wind capacity and the distributed wind turbines are all introduced based on the most recent statistical data. Especially, the wind energy development in the United

States and several key projects are introduced and summarized.

The second part mainly introduced the wind resources in the Cleveland area. This part evaluated the implementation potential of LiDAR system for wind resource assessment. This mobile instrument can simultaneously measure the wind speed and wind direction for five heights up to 150m of height above ground level. A site in northeast

Cleveland are evaluated using LiDAR technology and the results are compared with wind turbine energy output. It proved that the LiDAR monitored data is accurate in predicting energy otuptu of wind turbines.

The third part of the thesis describes the feasibility analysis of an innovative, extensible blade technology. The blade aims to significantly improve the energy production

154

of a wind turbine, particularly at locations with unfavorable wind conditions. The extensible blade functions by adjusting its length depending on the wind conditions (i.e., it will extend at low wind speed and retract at high wind speed). Based on the principle that the larger the sweep area, the higher the turbine energy output, dynamically adjusting the blade length helps to increase the energy output under low wind speed while mitigating safety risks under high wind speed.

A computational model is developed based on the blade element momentum (BEM) theory, which determines the aerodynamic load and power output of the blade at different wind conditions. In this study, the bio-inspired wind turbine blade with tubercles on the leading edge was proposed, and its power output performance was investigated by means of a series of wind tunnel tests conducted in the Control & Energy Systems Center at Case

Western Reserve University. Seven groups of wind turbine blades are studied in this research, which include one reference blade without any tubercles and six bio-inspired blades with different wavelength and amplitudes of tubercles. Two series of wind tunnel test were conducted. The first test aimed to find the maximum power coefficient by plotting the power coefficient as a function of tip speed ratio (TSR). The effects of different amplitude and wavelengths of the tubercles are depicted based on the wind tunnel tests. In

155

addition, the effects of the tubercles in delaying the maximum TSR of the rotors were summarized. The second experiment aims to simulate the power performance of the bio- inspired blade to simulate the working condition of the real wind turbine. The power output of the seven models were tested under varies wind speeds in 200s period. Finally, a numerical simulation model was proposed to study the mechanism of the leading edge tubercles at low Reynolds number condition.

5.2 Future Work

This study proposed two innovative smart blade model in increasing the wind energy efficiency. The experimental tests, theatrical calculation, and numerical simulation are used to study and analyze the models. Despite the innovative efforts and progress that have been made, there are still several challenges that need to be solve in the next step.

Therefore, there are still server research of interests, which can be further studied. The future research interests can be concluded as follows:

(1) From macroscopic aspect, the wind energy capacity will continually increase in the U.S.

and around the world. A significant amount of the wind turbines will be built as

distributed wind turbines to integrate into the smart electricity grid. Increasing the

156

efficiency of the wind turbine under unfavorable wind conditions will be one of the

most important research focus for wind engineering.

(2) This study aims to provide practical prediction of performance of the extensible blades.

For this reason, the well-established BEM theory that is accepted by the industry is

used for the analyses. The performance of the BEM model is firstly validated with the

monitored energy production data from the conventional 100kW turbine blade. The

validated BEM theory is then used to predict the performance of extended blade. In the

next step of the study, a sophisticated CFD model will be built and validated with the

theoretical calculation. The sophisticated CFD model can be used to predict different

size of wind turbine under different wind conditions with lower cost.

(3) This work evaluated the benefits of the extensible blade in energy production, but the

mechanism for control has not been presented. We have investigated into a few

strategies for the implementation of the extensible blade mechanism. Mechanisms

considered include that used in extensible umbrella, by use of a linear actuator and

locking mechanism; telescope blade etc. These involves control algorithms and is an

area we continue to look into.

(4) The performance of the bio-inspired blade are mainly based on the experimental study.

157

The fundamental phenomena of the tubercles on the leading of the wind turbine blade

has been studied with statistic model based on CFD software. The flow and wake under

rotating condition of the bio-inspired blade, which requires more computational

capacity has not been studied in this research. Besides, the size and scale of the

tubercles in this study are based on the real tubercles on the flippers of the whale. This

could be further optimized in the future study.

(5) The bio-inspired tubercles on the leading edge presented in this study cannot only be

applied on the wind turbine blades, but also other structures in the civil engineering

field. For instance, the development of the concrete 3D printing technology makes it

possible to build the abnormal shape of bridge pier with lower cost. The tubercles might

have the ability to mitigate the bridge scour because there are similar point between the

hydrodynamics and the aerodynamics. The similar analysis procedures can be used on

the bridge scour problems

158

REFERENCES

[1] P.W. Carlin, A.S. Laxson, E. Muljadi. The history and state of the art of variable‐speed wind turbine technology. Wind Energy. 6 (2003) 129-59.

[2] T. Burton, D. Sharpe, N. Jenkins, E. Bossanyi. Wind energy handbook. John Wiley & Sons2001.

[3] X. Wang, X. Zeng, X. Yang, J. Li. Feasibility study of offshore wind turbines with hybrid monopile foundation based on centrifuge modeling. Applied Energy. 209 (2018) 127-39.

[4] X. Wang, X. Zeng, J. Li, X. Yang. Lateral bearing capacity of hybrid monopile-friction wheel foundation for offshore wind turbines by centrifuge modelling. Ocean Engineering. 148 (2018) 182-92.

[5] X. Wang, X. Yang, X. Zeng. Seismic centrifuge modelling of suction bucket foundation for offshore wind turbine. Renewable Energy. 114 (2017) 1013-22.

[6] J. Koh, E. Ng. Downwind offshore wind turbines: Opportunities, trends and technical challenges. Renewable and Sustainable Energy Reviews. 54 (2016) 797-808.

[7] X. Wang, X. Yang, X. Zeng. Centrifuge modeling of lateral bearing behavior of offshore wind turbine with suction bucket foundation in sand. Ocean Engineering. 139 (2017) 140- 51.

[8] X. Wang, X. Zeng, J. Li, X. Yang, H. Wang. A review on recent advancements of substructures for offshore wind turbines. Energy Conversion and Management. 158 (2018) 103-19.

[9] M.S. Adaramola, M. Agelin-Chaab, S.S. Paul. Assessment of wind power generation along the coast of Ghana. Energy Conversion and Management. 77 (2014) 61-9.

159

[10] J. Li, X. Yu. Analyses of the Extensible Blade in Improving Wind Energy Production at Sites with Low-Class Wind Resource. Energies. 10 (2017) 1295.

[11] G.W.E.C. (GWEC). Global Wind Report Annual Market Update. 2016.

[12] Ü. Sağlam. A two-stage data envelopment analysis model for efficiency assessments of 39 state’s wind power in the United States. Energy Conversion and Management. 146 (2017) 52-67.

[13] R. Wiser, E. Lantz, T. Mai, J. Zayas, E. DeMeo, E. Eugeni, et al. Wind vision: A new era for wind power in the United States. The Electricity Journal. 28 (2015) 120-32.

[14] J. Li, X. Wang, Y. Tang. Modelling and hydrodynamic analysis of tension leg platform floating support structure for offshore wind turbine. 50th AIAA/ASME/SAE/ASEE Joint Propulsion Conference2014. p. 3414.

[15] W. Musial, S. Butterfield, B. Ram. Energy from offshore wind. Offshore technology conference. Offshore Technology Conference2006.

[16] X. Wang, X. Yang, X. Zeng. Lateral response of improved suction bucket foundation for offshore wind turbine in centrifuge modelling. Ocean Engineering. 141 (2017) 295-307.

[17] A.D. Mekonnen, P.V. Gorsevski. A web-based participatory GIS (PGIS) for offshore wind farm suitability within Lake Erie, Ohio. Renewable and Sustainable Energy Reviews. 41 (2015) 162-77.

[18] M. Ashtine, R. Bello, K. Higuchi. Assessment of wind energy potential over Ontario and Great Lakes using the NARR data: 1980–2012. Renewable and Sustainable Energy Reviews. 56 (2016) 272-82.

[19] A.C. Chiang, G.A. Keoleian, M.R. Moore, J.C. Kelly. Investment cost and view damage cost of siting an offshore wind farm: A spatial analysis of Lake Michigan. Renewable Energy. 96 (2016) 966-76.

[20] P.S. Grewal, P.S. Grewal. Can cities become self-reliant in energy? A technological scenario analysis for Cleveland, Ohio. Cities. 31 (2013) 404-11.

160

[21] International Electrotechnical Commission, TC88-MT, IEC 61400-3: Wind Turbines– Part 1: Design Requirements. Geneva, Switzerland, 2005.

[22] K. Chen, M. Song, X. Zhang. The iteration method for tower height matching in wind farm design. Journal of Wind Engineering and Industrial Aerodynamics. 132 (2014) 37-48.

[23] S. Calvert, R. Thresher, S. Hock, A. Laxson, B. Smith. US department of energy wind energy research program for low wind speed technology of the future. Journal of Solar Energy Engineering. 124 (2002) 455-8.

[24] M. Robinson, P. Veers. US national laboratory research supporting low wind speed technology. Transactions-American Society of Mechanical Engineers Journal of Solar Energy Engineering. 124 (2002) 458-9.

[25] J. Zhang, Z. Zhou, Y. Lei. Design and research of high-performance low-speed wind turbine blades. World Non-Grid-Connected Wind Power and Energy Conference (WNWEC 2009) Nanjing, China, 24-26 September 2009. pp. 1-5.

[26] J. Li, X. Yu. Feasibility Analyses of Using Renewable Wind Energy for Landslide Mitigation. Geotechnical Frontiers 2017. pp. 382-9.

[27] J. Li, X.B. Yu. Environmental Benign Electrokinetics for Landslide Mitigation. Geoenvironmental Practices and Sustainability. Springer2017. pp. 21-30.

[28] A.C. Orrell, N.A. Foster, J.S. Homer, S.L. Morris. 2016 Distributed Wind Market Report. Pacific Northwest National Laboratory (PNNL), Richland, WA (US)2017.

[29] R. Barnes, E. Morozov, K. Shankar. Improved methodology for design of low wind speed specific wind turbine blades. Composite Structures. 119 (2015) 677-84.

[30] C. Wichser, K. Klink. Low wind speed turbines and wind power potential in Minnesota, USA. Renewable Energy. 33 (2008) 1749-58.

[31] M.M. Alam, S. Rehman, J.P. Meyer, L.M. Al-Hadhrami. Review of 600–2500kW sized wind turbines and optimization of hub height for maximum wind energy yield realization. Renewable and Sustainable Energy Reviews. 15 (2011) 3839-49.

161

[32] J. Lee, D.R. Kim, K.-S. Lee. Optimum hub height of a wind turbine for maximizing annual net profit. Energy Conversion and Management. 100 (2015) 90-6.

[33] S. Rehman, L.M. Al-Hadhrami, M.M. Alam, J.P. Meyer. Empirical correlation between hub height and local wind shear exponent for different sizes of wind turbines. Sustainable Energy Technologies and Assessments. 4 (2013) 45-51.

[34] A. Pellegrino, C. Meskell. Vortex shedding from a wind turbine blade section at high angles of attack. Journal of Wind Engineering and Industrial Aerodynamics. 121 (2013) 131-7.

[35] T. No, J.-E. Kim, J. Moon, S. Kim. Modeling, control, and simulation of dual rotor wind turbine generator system. Renewable Energy. 34 (2009) 2124-32.

[36] E. Farahani, N. Hosseinzadeh, M. Ektesabi. Comparison of fault-ride-through capability of dual and single-rotor wind turbines. Renewable Energy. 48 (2012) 473-81.

[37] D. Huang, G. Wu. Preliminary study on the aerodynamic characteristics of an adaptive reconfigurable airfoil. Aerospace Science and Technology. 27 (2013) 44-8.

[38] S. Bhuyan, A. Biswas. Investigations on self-starting and performance characteristics of simple H and hybrid H-Savonius vertical axis wind rotors. Energy Conversion and Management. 87 (2014) 859-67.

[39] C.L. Bottasso, A. Croce, F. Gualdoni, P. Montinari. Load mitigation for wind turbines by a passive aeroelastic device. Journal of Wind Engineering and Industrial Aerodynamics. 148 (2016) 57-69.

[40] P. Lissaman. Low-Reynolds-number airfoils. Annual Review of Fluid Mechanics. 15 (1983) 223-39.

[41] Y. Wang, X. Sun, X. Dong, B. Zhu, D. Huang, Z. Zheng. Numerical investigation on aerodynamic performance of a novel vertical axis wind turbine with adaptive blades. Energy Conversion and Management. 108 (2016) 275-86.

[42] P.S. Veers, T.D. Ashwill, H.J. Sutherland, D.L. Laird, D.W. Lobitz, D.A. Griffin, et al.

162

Trends in the design, manufacture and evaluation of wind turbine blades. Wind Energy. 6 (2003) 245-59.

[43] G.S. Bir. Computerized method for preliminary structural design of composite wind turbine blades. Journal of Solar Energy Engineering. 123 (2001) 372-81.

[44] K. Jackson, M. Zuteck, C. Van Dam, K. Standish, D. Berry. Innovative design approaches for large wind turbine blades. Wind Energy. 8 (2005) 141-71.

[45] D.A. Griffin. Blade system design studies volume I: composite technologies for large wind turbine blades. Paper No SAND-1879, Sandia National Laboratories. Paper No. SAND-1879, Sandia National Laboratories, Livermore, CA, USA, 2002.

[46] D.A. Griffin. WindPACT Turbine design scaling studies technical area 1œComposite blades for 80-to 120-meter rotor. Technical report, National Renewable Energy Laboratory. Technical report, National Renewable Energy Laboratory, Livermore, CA, USA, 2001.

[47] M. Jureczko, M. Pawlak, A. Mężyk. Optimisation of wind turbine blades. Journal of Materials Processing Technology. 167 (2005) 463-71.

[48] Y. Luo, L. Yuan, J. Li, J. Wang. Boundary layer drag reduction research hypotheses derived from bio-inspired surface and recent advanced applications. Micron. 79 (2015) 59- 73.

[49] A. Naumann, M. Morsbach, C. Kramer. The conditions of separation and vortex formation past cylinders. AGARD Conf Proc1966. pp. 539-74.

[50] M. Tanner. A method for reducing the base drag of wings with blunt trailing edge. The Aeronautical Quarterly. 23 (1972) 15-23.

[51] F.E. Fish, J.M. Battle. Hydrodynamic design of the humpback whale flipper. Journal of Morphology. 225 (1995) 51-60.

[52] E.A. Van Nierop, S. Alben, M.P. Brenner. How bumps on whale flippers delay stall: an aerodynamic model. Physical review letters. 100 (2008) 054502.

163

[53] D. Serson, J. Meneghini. Numerical Study of Wings With Wavy Leading and Trailing Edges. Procedia IUTAM. 14 (2015) 563-9.

[54] P. Watts, F. Fish. The influence of passive, leading edge tubercles on wing performance. Proc Twelfth Intl Symp Unmanned Untethered Submers Technol. Auton. Undersea Syst. Inst. Durham New Hampshire2001.

[55] D. Miklosovic, M. Murray, L. Howle, F. Fish. Leading-edge tubercles delay stall on humpback whale (Megaptera novaeangliae) flippers. Physics of Fluids (1994-present). 16 (2004) L39-L42.

[56] C.-J. Bai, Y.-Y. Lin, S.-Y. Lin, W.-C. Wang. Computational fluid dynamics analysis of the vertical axis wind turbine blade with tubercle leading edge. Journal of Renewable and Sustainable Energy. 7 (2015) 033124.

[57] R.K. Zhang, V.D.J.Z. Wu. Aerodynamic characteristics of wind turbine blades with a sinusoidal leading edge. Wind Energy. 15 (2012) 407-24.

[58] G.-Y. Huang, Y. Shiah, C.-J. Bai, W. Chong. Experimental study of the protuberance effect on the blade performance of a small horizontal axis wind turbine. Journal of Wind Engineering and Industrial Aerodynamics. 147 (2015) 202-11.

[59] C. Fant, B. Gunturu, A. Schlosser. Characterizing wind power resource reliability in southern Africa. Applied Energy. 161 (2016) 565-73.

[60] H. Saleh, A.A.E.-A. Aly, S. Abdel-Hady. Assessment of different methods used to estimate Weibull distribution parameters for wind speed in Zafarana wind farm, Suez Gulf, Egypt. Energy. 44 (2012) 710-9.

[61] S.A. Akdağ, A. Dinler. A new method to estimate Weibull parameters for wind energy applications. Energy Conversion and Management. 50 (2009) 1761-6.

[62] O. Ohunakin, M. Adaramola, O. Oyewola. Wind energy evaluation for electricity generation using WECS in seven selected locations in Nigeria. Applied Energy. 88 (2011) 3197-206.

164

[63] S. Akdağ, H. Bagiorgas, G. Mihalakakou. Use of two-component Weibull mixtures in the analysis of wind speed in the Eastern Mediterranean. Applied Energy. 87 (2010) 2566- 73.

[64] C. Justus, W. Hargraves, A. Mikhail, D. Graber. Methods for estimating wind speed frequency distributions. Journal of Applied meteorology. 17 (1978) 350-3.

[65] E.K. Akpinar, S. Akpinar. An assessment on seasonal analysis of wind energy characteristics and wind turbine characteristics. Energy Conversion and Management. 46 (2005) 1848-67.

[66] F. Chellali, A. Khellaf, A. Belouchrani, R. Khanniche. A comparison between wind speed distributions derived from the maximum entropy principle and Weibull distribution. Case of study; six regions of Algeria. Renewable and Sustainable Energy Reviews. 16 (2012) 379-85.

[67] P.A.C. Rocha, R.C. de Sousa, C.F. de Andrade, M.E.V. da Silva. Comparison of seven numerical methods for determining Weibull parameters for wind energy generation in the northeast region of Brazil. Applied Energy. 89 (2012) 395-400.

[68] J. Seguro, T. Lambert. Modern estimation of the parameters of the Weibull wind speed distribution for wind energy analysis. Journal of Wind Engineering and Industrial Aerodynamics. 85 (2000) 75-84.

[69] J.A. Carta, P. Ramirez, S. Velazquez. A review of wind speed probability distributions used in wind energy analysis: Case studies in the Canary Islands. Renewable and Sustainable Energy Reviews. 13 (2009) 933-55.

[70] J.P. Hennessey Jr. Some aspects of wind power statistics. Journal of Applied meteorology. 16 (1977) 119-28.

[71] A. Peña, C.B. Hasager, S.E. Gryning, M. Courtney, I. Antoniou, T. Mikkelsen. Offshore wind profiling using light detection and ranging measurements. Wind Energy. 12 (2009) 105-24.

[72] E.K. Akpinar, S. Akpinar. Determination of the wind energy potential for Maden- 165

Elazig, Turkey. Energy Conversion and Management. 45 (2004) 2901-14.

[73] Ž. Đurišić, J. Mikulović. Assessment of the wind energy resource in the South Banat region, Serbia. Renewable and Sustainable Energy Reviews. 16 (2012) 3014-23.

[74] S. Rehman, N.M. Al-Abbadi. Wind shear coefficient, turbulence intensity and wind power potential assessment for Dhulom, Saudi Arabia. Renewable Energy. 33 (2008) 2653- 60.

[75] A. Scientific. Wind resource assessment handbook: Fundamentals for conducting a successful monitoring program. AWS Scientific Inc, New York, TAT-5-15283-01. (1997).

[76] F.A. Jowder. Wind power analysis and site matching of wind turbine generators in Kingdom of Bahrain. Applied Energy. 86 (2009) 538-45.

[77] G. Hellmann. Über die Bewegung der Luft in den untersten Schichten der Atmosphäre. Kgl. Akademie der Wissenschaften [G.] Reimer1914.

[78] M. Lydia, S.S. Kumar, A.I. Selvakumar, G.E.P. Kumar. Linear and non-linear autoregressive models for short-term wind speed forecasting. Energy Conversion and Management. 112 (2016) 115-24.

[79] J. Counihan. Adiabatic atmospheric boundary layers: a review and analysis of data from the period 1880–1972. Atmospheric Environment (1967). 9 (1975) 871-905.

[80] G. Van den Berg. Wind turbine power and sound in relation to atmospheric stability. Wind Energy. 11 (2008) 151-69.

[81] ASCE. Minimum design loads for buildings and other structures. Amer Society of Civil Engineers2005.

[82] O. Jaramillo, M. Borja. Wind speed analysis in La Ventosa, Mexico: a bimodal probability distribution case. Renewable Energy. 29 (2004) 1613-30.

[83] R. Farrugia. The wind shear exponent in a Mediterranean island climate. Renewable Energy. 28 (2003) 647-53.

166

[84] S. Rehman, N.M. Al-Abbadi. Wind shear coefficients and their effect on energy production. Energy Conversion and Management. 46 (2005) 2578-91.

[85] NREL. Ohio annual average wind speed estimates at 100m height. National Renewable Energy Lab2007.

[86] J. Li, X. Yu. Model and Procedures for Reliable Near Term Wind Energy Production Forecast. Wind Engineering. 39 (2015) 595-608.

[87] D. Matthiesen. Wind Energy Research & Commercialization (WERC) Center. figshare. 2016.

[88] C.-Y. Hsuan, Y.-S. Tasi, J.-H. Ke, R.A. Prahmana, K.-J. Chen, T.-H. Lin. Validation and measurements of floating LiDAR for nearshore wind resource assessment application. Energy Procedia. 61 (2014) 1699-702.

[89] D.A. Smith, M. Harris, A.S. Coffey, T. Mikkelsen, H.E. Jørgensen, J. Mann, et al. Wind lidar evaluation at the Danish wind test site in Høvsøre. Wind Energy. 9 (2006) 87- 93.

[90] C.B. Hasager, A. Peña, T. Mikkelsen, M.S. Courtney, I. Antoniou, S.-E. Gryning, et al. 12MW Horns Rev Experiment. Risø National Laboratory2007.

[91] H.E. Jørgensen, T. Mikkelsen, J. Mann, D. Bryce, A. Coffey, M. Harris, et al. Site wind field determination using a CW Doppler LIDAR-comparison with cup anemometers at Risø. Delft University of Technology Proceedings. (2004).

[92] C.R. Standridge, D. Zeitler, E. Nordman, T.A. Boezaart, J. Edmonson, Y. Nieves, et al. Laser Wind Sensor Performance Validation with an Existing Gage. (2013).

[93] NREL. Classes of wind power density at 10 m and 50 m. 2014.

[94] H. Bagiorgas, G. Mihalakakou, D. Matthopoulos. A statistical analysis of wind speed distributions in the area of Western Greece. International Journal of Green Energy. 5 (2008) 120-37.

167

[95] C. Steele, S. Dorling, R. von Glasow, J. Bacon. Modelling sea‐breeze climatologies and interactions on coasts in the southern North Sea: implications for offshore wind energy. Quarterly Journal of the Royal Meteorological Society. 141 (2015) 1821-35.

[96] B. Bailey. Wind resources for offshore wind farms: characteristics and assessment. Offshore Wind Farms: Technologies, Design and Operation. (2016) 29.

[97] R. Sharma, U. Madawala. The concept of a smart wind turbine system. Renewable Energy. 39 (2012) 403-10.

[98] M. Imraan, R.N. Sharma, R.G. Flay. Wind tunnel testing of a wind turbine with telescopic blades: The influence of blade extension. Energy. 53 (2013) 22-32.

[99] T.J. McCoy, D.A. Griffin. Control of rotor geometry and aerodynamics: Retractable blades and advanced concepts. Wind Engineering. 32 (2008) 13-26.

[100] J.M. Jonkman, S. Butterfield, W. Musial, G. Scott. Definition of a 5-MW reference wind turbine for offshore system development. National Renewable Energy Laboratory (NREL) Golden, CO, USA, 2009.

[101] J.F. Manwell, J.G. McGowan, A.L. Rogers. Wind energy explained: theory, design and application. John Wiley & Sons2010.

[102] S. Mathew. Wind energy: fundamentals, resource analysis and economics. Springer2006.

[103] H. Glauert. Airplane propellers. Aerodynamic theory. Springer, Berlin, Heidelberg, 1935. pp. 169-360.

[104] J. Leishman. Principles of Helicopter Aerodynamics, CambridgeUniv. Press, New. (2000).

[105] P.J. Moriarty, A.C. Hansen. AeroDyn theory manual. National Renewable Energy Laboratory Golden, Colorado, USA2005.

[106] E. Kulunk. Aerodynamics of wind turbines. Fundamental and Advanced Topics in

168

Wind Power. (1970).

[107] N. Veritas. Guidelines for design of wind turbines. Det Norske Veritas: Wind Energy Department, Ris ̜National Laboratory: Oslo, Norway2002.

[108] A. Sedaghat, M.E.H. Assad, M. Gaith. Aerodynamics performance of continuously variable speed horizontal axis wind turbine with optimal blades. Energy. 77 (2014) 752-9.

[109] N. Tenguria, N. Mittal, S. Ahmed. Investigation of blade performance of horizontal axis wind turbine based on blade element momentum theory (BEMT) using NACA airfoils. International Journal of engineering, science and technology. 2 (2010).

[110] E. Kulunk. Aerodynamics of Wind Turbines, Fundamental and Advanced Topics in Wind Power, Dr. Rupp Carriveau (Ed.), ISBN: 978-953-307-508-2. InTech2011.

[111] P. Pourazarm, L. Caracoglia, M. Lackner, Y. Modarres-Sadeghi. Stochastic analysis of flow-induced dynamic instabilities of wind turbine blades. Journal of Wind Engineering and Industrial Aerodynamics. 137 (2015) 37-45.

[112] N. Chaudhari. Dynamic Characteristics of Wind Turbine Blade. International Journal of Engineering Research and Technology. ESRSA Publications2014.

[113] Y. Cui, L. Shi, F. Zhao. Modal analysis of wind turbine blade made of composite laminated plates. Power and Energy Engineering Conference (APPEEC), Chengdu, China, 28-31 March 2010. pp. 1-4.

[114] M. Matsuiski, T. Endo. Fatigue of Metals Subjected to Varying Stress (The Fatigue Life of Materials Subjected to Random Strains). Kyushu District Meeting, Japan Soc Mech Engrs, Kitakyushu-Shi1968.

[115] S.N. Clarke, D.W. Goodpasture, R.M. Bennett, J.H. Deatherage, E.G. Burdette. Effect of cycle-counting methods on effective stress range and number of stress cycles for fatigue-prone details. Transportation Research Record: Journal of the Transportation Research Board. 1740 (2000) 49-60.

[116] S.D. Downing, D. Socie. Simple rainflow counting algorithms. International Journal

169

of Fatigue. 4 (1982) 31-40.

[117] A. Niesłony. Determination of fragments of multiaxial service loading strongly influencing the fatigue of machine components. Mechanical Systems and Signal Processing. 23 (2009) 2712-21.

[118] Matlab. Matlab. Availiable at . 2016.

[119] M.A. Miner. Cumulative damage in fatigue. Journal of applied mechanics. 12 (1945) 159-64.

[120] E. Taslimi-Renani, M. Modiri-Delshad, M.F.M. Elias, N.A. Rahim. Development of an enhanced parametric model for wind turbine power curve. Applied Energy. 177 (2016) 544-52.

[121] S. Aftab, N. Razak, A.M. Rafie, K. Ahmad. Mimicking the humpback whale: An aerodynamic perspective. Progress in Aerospace Sciences. 84 (2016) 48-69.

[122] F. Wang. Wind Farm Control for Optimal Power Generation and Fatigue Reduction: Strategies and Experimentation in Wind Tunnel. Case Western Reserve University2017.

[123] I. Bayati, M. Belloli, L. Bernini, A. Zasso. Aerodynamic design methodology for wind tunnel tests of wind turbine rotors. Journal of Wind Engineering and Industrial Aerodynamics. 167 (2017) 217-27.

[124] F.-B. Hsiao, C.-J. Bai, W.-T. Chong. The performance test of three different horizontal axis wind turbine (HAWT) blade shapes using experimental and numerical methods. Energies. 6 (2013) 2784-803.

[125] Q.a. Li, Y. Kamada, T. Maeda, J. Murata, N. Yusuke. Effect of turbulence on power performance of a Horizontal Axis Wind Turbine in yawed and no-yawed flow conditions. Energy. 109 (2016) 703-11.

[126] M. Zhang, G. Wang, J. Xu. Experimental study of flow separation control on a low- Re airfoil using leading-edge protuberance method. Experiments in fluids. 55 (2014) 1710.

170

[127] F. Wang, L. Wheeler, M. Garcia-Sanz. Wind Turbine Power Optimization: Experimental Validation of Extremum Seeking and Perturb/Observe Strategies. ASME 2016 Dynamic Systems and Control Conference. American Society of Mechanical Engineers2016. pp. V002T21A5-VT21A5.

[128] F. Wang, M. Garcia-Sanz. Wind farm cooperative control for optimal power generation. Wind Engineering. (2018 (In Press)).

[129] M. Garcia-Sanz, C.H. Houpis. Wind energy systems: control engineering design. CRC press2012.

[130] M.S. Selig, J.J. Guglielmo. High-lift low Reynolds number airfoil design. Journal of aircraft. 34 (1997) 72-9.

[131] S.A. Oller Aramayo, L.G. Nallim, S.H. Oller Martínez. The usability of the Selig S1223 profile airfoil as a high lift hydrofoil for hydrokinetic application. Journal of Applied Fluid Mechanics. 9 (2016) 537-42.

[132] J. Guerreiro, J. Sousa. Low-Reynolds-number effects in passive stall control using sinusoidal leading edges. AIAA journal. 50 (2012) 461-9.

[133] D. Custodio, C. Henoch, H. Johari. Aerodynamic characteristics of finite span wings with leading-edge protuberances. AIAA journal. (2015).

[134] M. Zhang, G. Wang, J. Xu. Aerodynamic control of low-Reynolds-number airfoil with leading-edge protuberances. AIAA journal. (2013).

[135] I. Zverkov, B. Zanin, V. Kozlov. Disturbances growth in boundary layers on classical and wavy surface wing. AIAA journal. 46 (2008) 3149-58.

[136] H. Johari, C.W. Henoch, D. Custodio, A. Levshin. Effects of leading-edge protuberances on airfoil performance. AIAA journal. 45 (2007) 2634-42.

[137] K. Hansen, R. Kelso, B. Dally. The effect of leading edge tubercle geometry on the performance of different airfoils. World. (2009).

171

[138] T. Goruney, D. Rockwell. Flow past a delta wing with a sinusoidal leading edge: near-surface topology and flow structure. Experiments in fluids. 47 (2009) 321-31.

[139] M.S. Selig, R.W. Deters, G.A. Williamson. Wind tunnel testing airfoils at low Reynolds numbers. AIAA Paper. 875 (2011).

[140] D. Marten, J. Wendler, G. Pechlivanoglou, C. Nayeri, C. Paschereit. QBlade: An open source tool for design and simulation of horizontal and vertical axis wind turbines. IJETAE. 3 (2013) 264-9.

[141] T. Burton, N. Jenkins, D. Sharpe, E. Bossanyi. Wind energy handbook. John Wiley & Sons2011.

[142] H.G. Jeong, R.H. Seung, K.B. Lee. An improved maximum power point tracking method for wind power systems. Energies. 5 (2012) 1339-54.

[143] M. Gregorek, M. Hoffmann, M. Berchak. Steady State and Oscillatory Aerodynamic Characteristics of a NACA 0021 Airfoil: Data Report. Ohio State University, Columbus. (1989).

[144] B. Hand, G. Kelly, A. Cashman. Numerical simulation of a vertical axis wind turbine airfoil experiencing dynamic stall at high Reynolds numbers. Computers & Fluids. 149 (2017) 12-30.

[145] N. Rostamzadeh, R. Kelso, B. Dally, K. Hansen. The effect of undulating leading- edge modifications on NACA 0021 airfoil characteristics. Physics of Fluids. 25 (2013) 117101.

[146] W.P. Wolfe, S.S. Ochs. Predicting aerodynamic characteristic of typical wind turbine airfoils using CFD. Sandia National Labs., Albuquerque, NM (United States)1997.

172