Airborne Wind Energy

Technology Review and Feasibility in Germany

Seminar Paper for Sustainable Energy Systems Faculty of Mechanical Engineering Technical University of Munich

Supervisors Johne, Philipp Hetterich, Barbara Chair of Energy Systems

Authors Drexler, Christoph Hofmann, Alexander Kiss, Balínt

Handed in Munich, 05. July 2017 Abstract

As a new generation of wind energy systems, AWESs (Airborne Wind Energy Systems) have the potential to grow competitive to their conventional ancestors within the upcoming decade. An overview of the state of the art of AWESs has been presented. For the feasibility analysis of AWESs in Germany, a detailed wind analysis of a three dimensional grid of 80 data points above Germany has been conducted. Long-term NWM (Numerical Weather Model) data over 38 years provided by the NCEP (National Centers for Environmental Prediction) has been analysed to determine the wind probability distributions at elevated altitudes. Besides other data, these distributions and available performance curves have been used to calcu- late the evaluation criteria AEEY (Annual Electrical Energy Yield) and CF (Capacity Factor). Together with the additional criteria LCOE (Levelised Costs of Electricity), MP (Material Per- formance), and REP (Rated Electrical Power) a quantitative cost utility analysis according to Zangemeister has been conducted. This analysis has shown that AWESs look promising and could become an attractive alternative to traditional wind energy systems.

2 Table of Contents

1 Introduction ...... 5 1.1 Initial situation and motivation...... 5 1.2 Objective ...... 6 1.3 Approach...... 7 2 Fundamentals ...... 8 2.1 Wind modelling...... 8 2.1.1 Low altitude wind...... 8 2.1.2 High altitude wind ...... 9 2.2 Airborne Wind Energy Systems ...... 12 2.2.1 Classiﬁcation ...... 12 2.2.2 Maximum theoretical yield of AWE...... 13 3 State of the art...... 15 4 Wind conditions in Germany and data ...... 21 5 Cost Utility Analysis...... 25 5.1 Zangemeister Analysis ...... 25 5.2 Criteria deﬁnition ...... 26 5.2.1 LCOE ...... 26 5.2.2 CF...... 26 5.2.3 AEEY...... 26 5.2.4 MP ...... 27 5.2.5 REP...... 27 6 Conclusion and outlook...... 31

Appendices...... 34 A Typical function values ...... 35 B Interpolated wind distributions of investigated systems ...... 36 C Performance curves of investigated prototypes ...... 40 D Complete overview of existing AWES ...... 43

3 Acronyms

A AEEY - Annual Electrical Energy Yield...... 2, 3, 27, 31, 32 AEP - Annual Energy Production...... 3, 26 API - Application Programming Interface...... 3 AWES - Airborne Wind Energy System 2, 3, 6–9, 12, 13, 15, 17, 22, 23, 25, 27–29, 31–33

C CCGT - Combined Cycle Gas Turbine...... 3 CF - Capacity Factor...... 2, 3, 26, 27, 31 CFD - Computaional Fluid Dynamic...... 3, 19 CRF - Capital Recovery Factor...... 3, 26 CUA - Cost Utility Analysis ...... 3, 27, 29

D DPKVHLC - Drag Power Kite with Very High Lift Coefﬁcient...... 3, 19, 29

I ICC - Initial Capital Costs...... 3, 26, 28, 32

L LCOE - Levelised Costs of Electricity...... 2, 3, 5, 6, 20, 26, 29, 31

M MP - Material Performance...... 2, 3, 27, 29, 31

N NCEP - National Centers for Environmental Prediction ...... 2, 3, 11 NWM - Numerical Weather Model...... 2, 3, 9, 11

O OMC - Operation and Maintanance Cost...... 3, 26

R RA - Ranalyses ...... 3, 21–23, 27 REP - Rated Electrical Power...... 2, 3, 26, 27, 29, 31, 32 RRE - Resulting Rated Energy ...... 3

4 1. Introduction

1.1. Initial situation and motivation

First, like solar, wind power is one of the few renewable energy resources that is in principle large enough to satisfy all of humanity’s energy needs.

Ahrens et al. 2013, p. 3

Energy production from wind is already competitive to conventional power plants burning fossil fuels or nuclear power production. Figure 1 shows the LCOE for selected energy production methods. At perfect onshore locations modern wind energy systems are able to produce electricity at a lower price than black coal power plants. Average onshore wind energy systems have LCOE between 0.045 and 0.107 e/kWh. Despite their in average higher util- ization rate those of offshore wind energy systems are considerably higher ranging between 0.119 and 0.194 e/kWh due to high investment costs. In comparison, brown coal power plants have the lowest LCOE ranging between 0.038 and 0.053 e/kWh (Kost et al. 2013).

Figure 1 LCOE comparison of different technologies (Data from Kost et al. 2013)

In 2016 renewable energy made up for over 30% of Germany’s gross electricity consumption whereof 12.4 % were produced from wind energy alone (Durstewitz et al. 2017). At the end of 2016 global installed wind power capacity hit 486,749 MW. Of the total 161,330 MW installed in Europe, around 50,018 MW were installed in Germany alone (GWEC, 2017). Typical total investment costs of wind energy systems in Germany range between 1.39 and 1.71 e/kWh depending on hub high, performance class and wind conditions Wallasch et al. 2013.

5 The main investment costs have a large share of the total investment costs with up to 78%. They refer to transport, installation and mainly the wind energy system itself which consists of tower, nacelle and rotor. Additional investment costs include the foundation, the grid connection, the infrastructure as well as planning and design. The foundation alone aggregates to about 4% of the total investment costs (Wallasch et al. 2013).

Collectively the investment costs required for the material of the main components are re- sponsible for the majority of the total investment costs and thereby for the LCOE of conventional wind power. The carbon dioxide footprint of conventional wind power ranges between

30-45 gCO2 / kWh. 90 % is due to the high amount of material cost (Wagner et al. 2007). Furthermore conventional wind energy systems are limited to special hub heights of up to 140 m maximum and not able to harvest winds of higher altitudes which are typically stronger and more consistent. This is one reason for limitations regarding the utilisation rate and the necessity of energy storage as well as restrictions regarding location selection and grid ex- pansion. A new approach to create electricity from wind energy are the so called AWESs. In contrast to conventional ground based and towered wind turbines AWESs are ﬂying freely in the air using soft kites or reinforced wings.

AWESs might have the potential to reach higher altitudes, harvesting more wind energy and enable to open up new locations for wind power generation. AWESs can be automatically adapted to the height of maximum wind speeds, thereby providing a more consistent power production. On the other hand AWESs might need signiﬁcant less investment costs for material compared to conventional plants. Both arguments combined could lead to lower levelised costs of electricity and thus pushing the implementation of renewable energies (Archer et al. 2014). Simulations of a 350 kW AWES carried out by a research group of the TU Delft in 2014 resulted in LCOE of 0.06 e/kWh already (Grete, 2014). AWES technology and research are still in the early stages of development and further LCOE decline can be expected due to material and design improvements.

1.2. Objective

The primary objective of this paper is to provide an overview of AWESs for generating power from wind energy and a feasibility study for an implementation in Germany. In preparation data about planned and existing systems and their characteristics as well as data about Ger- man wind conditions is collected. Meanwhile necessary requirements are defined. Sub- sequently relevant system characteristics are prioritized and evaluated within a cost-utility analysis according to the defined requirements. Building on the research and knowledge gained within this paper the feasibility of an implementation of AWESs in Germany is dis- cussed in a final step.

6 1.3. Approach

To meet the objective targeted theoretical and technological basics are imparted within chapter 2, “Fundamentals”. Thereby the paper primarily focusses on wind modelling, the general AWESs classification and operation principles as well as the physical potential and the limits of AWESs. Chapter 3, “State of the art”, categorises and structures existing and future AWES and the institutions developing them. Relevant findings of related papers or theses are introduced. Chapter 4, “Wind conditions in Germany and data”, analyses the local situation by combining precise simulation and results found in literature. Based both on the individual system characteristics displayed in chapter 3 and findings about the local wind conditions of chapter 4, precise criteria for a successful AWES implementation are derived in chapter 5, “Cost Utility Analysis”. Using the defined criteria, the systems are evaluated following a quantitative approach according to Zangemeister. The study focussed on an implementation in Germany. The final chapter 6, “Conclusion and outlook”, summarises the paper, contem- plates the findings and suggests further fields of research and development.

7 2. Fundamentals

2.1. Wind modelling

To evaluate the theoretical performance of AWES the wind conditions have to be known. Only very rarely, wind speed measurements for high altitudes are available. Contrary to near ground altitudes, no constant measurements are possible. In some cases, high altitude winds are investigated using weather balloons. To predict the performance and thus, proﬁtability of AWESs, prediction of high altitude wind speeds is inevitable. This section focuses on the methods and simpliﬁcations which are used to generate estimates.

2.1.1. Low altitude wind Wind speed follows a Weibull distribution. Mean wind speed generally increases with vertical distance from the ground z. This trend is only disturbed by turbulences and local variations. Two common formula for describing the vertical wind speed proﬁle exist. Both only model the ﬂuid dynamic effects. Atmospheric phenomena of the earth like Coriolis force and pressure extrema are not accounted for (Ahrens et al. 2013, p. 90).

The ﬁrst law is the logarithmic expression in equation 2.1. The wind speed value at reference altitude vw(zref ) is usually measured at heights around 10 metres. Using logarithmic expres- sions, the actual altitude z is then related to the surface roughness z0. The surface roughness is a ground characteristic, which describes the size and amount of obstacles protruding the area’s surface. The taller the obstacles, the bigger z0 (Ahrens et al. 2013, p. 90). Typical values for z0 are given in table 2.

log z z0 vw(z) = vw(zref ) · (2.1) log zref z0

The second law is a power law (2.2). Similar to the logarithmic expression it relies on the measurement of a reference wind speed and an estimation of the surface roughness, α.

Similar to the surface roughness z0, the friction coefﬁcient α is proportional to the size of the protruding obstacles. Values for different surfaces are given in table 3. The formulation in equation 2.2 has shown better correlation to empirical values (Ahrens et al. 2013, p. 90).

z α vw(z) = vw(zref ) · (2.2) zref

Both approaches are only valid within a boundary layer up to 500 metres above ground

8 (Ahrens et al. 2013, p. 90). The logarithmic law is superior to the power within regions closest to the ground with z ∈ [0m, 20m]. In the intermediate region from 20 m to 100 m, both models perform similarly. At altitudes higher than 100 metres, the power law has proven to be more reliable Stegner et al. 2017. The range covered with these simpliﬁcations is not sufﬁcient. To predict wind conditions above altitudes of 500 metres, scientists rely on NWMs. An overview of the modelling approach is given in section 2.1.2.

2.1.2. High altitude wind To model the high altitude winds, NWM are used. These models describe properties of air, like speed and temperature. For this holistic description, four governing equations are made use of.

Continuity equation

The continuity equation describes the mass conversion of ﬂuids. It couples the density of air

ρ with the wind velocity vector vw (Ahrens et al. 2013, p. 83):

  vw,x   1 dρ   · = −∇v , ; , v =   (2.3) w w vw,y ρ dt       vw,z

The term ∇vw is called air divergence. It quantifies the source of a vector field, here the wind velocity. If the air divergence is positive, the air density on the left hand side must be negative, or decrease. If a positive change, or increase in air density is present, the velocity field has a negative divergence. The divergence is null if the fluid is incompressible. For most AWESs, the air is treated as an incompressible fluid (Ahrens et al. 2013, p. 82-84).

Thermodynamic equation

The ﬁrst law of thermodynamics states the conservation of energy (Ahrens et al. 2013, p. 84):

dT dα c + p = J (2.4) v dt dt H

−1 With the speciﬁc heat cv, α = ρ , the pressure p, and the mass-speciﬁc rate of heating

JH are related. The latter is composed of heat due to radiation, conduction, and latent heat release. It does not include heat due to friction. Even though no velocity is explicitly included in equation 2.4, it is valid for moving particles (Ahrens et al. 2013, p. 84).

9 Momentum equation

Various effects are present in the rotating coordinate system describing the earth. Not all of which arise from the fact that the earth is rotating, but interact with each other in the form of ∗ density-speciﬁc pressures pi (Ahrens et al. 2013, p. 84):

∂vw X ∗ + (vm · ∇)vm = p (2.5) ∂t i i

The effects follow to (Ahrens et al. 2013, p. 84):

1. Pressure gradient force: Air volumes are pushed into the negative direction of the pressure gradient, from regions with higher to regions with lower pressure.

2. Coriolis force: Air particles are accelerated to their right in the northern hemisphere and to the left in the southern hemisphere.

3. Gravity force: Consists of the true gravitational force, which pulls particles towards the surface of the earth, and the centrifugal force.

4. Turbulence flux divergence: Eddies in the air exchange heat and momentum. The turbulence flux divergence quantifies this phenomenon.

5. Viscous diffusion: Representing general dampening effects, which describe how faster particles transfer momentum to slower particles, ultimately the ground.

Equation of state

The equation of states is a constitutive equation and correlates the pressure p to the density ρ, the universal gas constant R, and the temperature T (Ahrens et al. 2013, p. 85):

p = ρRT (2.6)

10 NWMs are used to solve this set of equations and describe the movement of air Ahrens et al. 2013, p. 86. The NCEP provides multiple datasets with different properties National Centers for Environmental Prediction, 2016. Later on, this work utilised data from the reanalysis project (http://reanalyses.org/). It is a combination of measured and simulated values for weather prediction for the prediction of state of earth’s atmosphere. The data is available over many decades, which allows the calculation of reliable mean values for wind speed. Even though the whole globe is covered, the downside of the reanalysis datasets is the low resolution of only 2.5◦ in both, latitude and longitude. Figure 2 shows the mean wind speeds at different altitudes over Germany in January.

1000mb/111m 925mb/763m 850mb/1458m 700mb/3013m 10

4 wind speed m/s speed wind 2

0 55 15 50 10

deg latitude (N) 45 5 deg longitude (E)

Figure 2 Mean wind velocities over Germany. Data from NOAA/ESRL/PSD, 2017

11 2.2. Airborne Wind Energy Systems

2.2.1. Classification AWESs can be categorised by their modus operandi. The first mode is called drag mode. Electric generators are placed onto the tethered wing and equipped with propellers to harvest the wind energy. An electric cable is used to transmit the power to the ground station, as shown in figure 3a.

(a) Drag mode. (b) Lift mode.

Figure 3 Modus operandi (Cherubini et al. 2015).

The second mode, lift mode, is also referred to as ground-based power generation. By pla- cing the generators on the ground, no electric cable is necessary, which results in lighter wings (3b). During operation, the tether is periodically rolled onto (recovery phase) and un- rolled from (generation phase) a drum which is connected to the generator. Alternatively, the movement of the kite be used to move wagons or rotate platforms on the ground, as shown in ﬁgure 4 (Ahrens et al. 2013, p. 7-12; Cherubini et al. 2015, p. 5-6).

Figure 4 Lift mode with moving ground station (left), circular rail system (middle) and linear rail system (right) (Cherubini et al. 2015).

12 2.2.2. Maximum theoretical yield of AWE The lift generated by a kite suspended in mid-air is proportional to the density of the ﬂuid

ρ surrounding it, its reference area A and corresponding lift coefﬁcient CL, and the true airspeed va.

1 F = ρAC v2 (2.7) L 2 L a

Loyd (1980) deducted a relation for the power P which can be generated using a tethered airfoil with idealized assumptions, while operating in drag or in lift mode (eq. 2.8). Instead of the true airspeed, the expression incorporates the actual wind speed vw and the drag coefﬁcient

CD. The influence of the wind speed increased even more compared to the equation for the lift force (2.7), with a power of three. The gliding number, which is described by the ratio of lift to drag coefficient has a significant influence as well, with a power of two. From this equation, two main strategies for increasing the power which can be harvested with AWESs become clear: Increasing operational height, where higher wind speeds are apparent, or increasing the gliding number.

2 2 3 CL P = ρAvwCL (2.8) 27 CD

Cosine losses

Not all of the power inherent in the wind can be extracted by AWESs. Besides losses due to mechanical and electrical efﬁciencies under 100%, the geometrical set-up of AWESs makes it impossible to do so. The aerodynamic force Fa holding the kite suspended in mid-air extracts a certain amount of power PW ind from the horizontally moving air. It is quantiﬁed by the projection of the aerodynamic force onto the direction of the wind:

1 F = ρAv2C (2.9) a 2 R

PW ind = vw · Fa · cos γ (2.10)

The smaller the operating angle between the tether and the ground is, the smaller the cosine losses. Optimized AWESs typically ﬂy at angles less than 30◦, resulting in geometrical cosine losses of around 13%. This effect is further increased by the weight of the wing and tether, which results in a more vertical alignment of the aerodynamic force Fa. However, this effect is usually negligible for most systems and only becomes relevant for systems with a high weight -to-power ratio at low wind speeds. (Ahrens et al. 2013, p. 6-7)

13 Drag losses

The drag losses of an aerodynamic system can be quantiﬁed using the drag coefﬁcient CD (Ahrens et al. 2013, p. 7).

1 P = v · F = v · ρAC v2 (2.11) loss,D a D a 2 D a

14 3. State of the art

Several different companies, research groups and projects from various countries can be found in the ﬁeld of AWES. A detailed overview of all systems and companies considered and contacted for this paper can be found in tables 31 and 32 in the appendix. Due to lack of complete sets of information, only some of the systems could be investigated in detail. An overview of the considered systems is given in table 1 and in this section.

EnerKíte

The EnerKíte GmbH was founded in 2010 in Berlin with the goal to define new wind energy using innovative kites. In 2012 a small mobile AWES, the EK30 (fig. 5a), was developed to test, develop and demonstrate the principle of the technology. In 2013 the EK30 performed multiple autonomous flights, each lasting several days. Flying at altitudes of about 200 m, the kite was with- standing wind speeds of up to 30 m/s. The test, which was supervised by the Fraunhofen IWES, validated the system controls and confirmed former simulations. To date the EK30 has already completed several hundred hours of operation in typical environments and conditions. (The Eco Solutions, 2017; Enerkite GmbH, 2017)

At the moment EnerKíte works on developing the cheap, mobile and fully autonomous EK200 with a rated power output of 100 kW (ﬁg. 5b). The system operates in lift mode, using a ﬁxed ground base and a 30 m2 hard wing. The kite is mounted on a telescopic crane. Rotating this crane allows the system to take off and land, even with zero wind speed at ground level. (Enerkite GmbH, 2017)

(a) EK30 (b) EK200

Figure 5 Hard kite systems by EnerKíte (Enerkite GmbH, 2017)

15 KitePower BV

KitePower BV is a start-up at the TU Delft. It is working on a fully autonomous commercial lift mode system with a ﬁxed ground station. It uses a soft kite, has a rated power output of 100 kW and is available from 2018 on. KitePower is one of the members of the Horizon 2020 Framework Programme of the European Union, which funds research in novel ﬁelds and hopes to accelerate innovation and its cultural acceptance. The company offers a holistic service. Besides including the actual power generation setup, complete off-grid service with batteries is available. The social and structural integration is considered through partnerships and trainings with local communities and partners. (Kitepower, 2017a,[b])

NTS Energy and Transport Systems GmbH

NTS Energy and Transport Systems GmbH was founded in Berlin in 2006 (The Eco Solutions, 2017). NTS uses soft kites in lift mode operation with moving ground station. The Kites pull train like power units on a round rail. Since 2012 the system is tested on a 400 m straight test rail system. Fully automated energy production on the test rail over 8 h was achieved. Fully automated controlling tests of the kites lasted several weeks. NTS is working on a round rail track of 5 km in length with 24 power units. Each unit is pulled by a kite of up to 400 m2, rated at 1 MW. A conceptual sketch of a power unit is shown in ﬁgure 6. The main system components rely on rail and train technologies which have existed for several decades with proven reliability and optimised efﬁciencies. (Kitegen, 2017; Ahrens, 2017)

According to (Ahrens, 2017) 4000 - 6000 full load hours and a electricity production of about 120 GWh per year are feasible on this track in Germany. The area of the kites and their height of operation has to be adjusted depending on the location. The system is able to achieve an LCOE of 2 - 4 ct/kWh. The data is conﬁrmed by a survey conducted by the Fraunhofer IWES. Also political and legal realisation is seen as indeed possible according to external evaluations (Ahrens, 2017).

Figure 6 Power unit by NTS (X-Wind, 2017).

16 KiteGen

KiteGen was founded in 2007 in Turin with the ultimate goal to realise a 1 GW AWES. During their founding year they successfully tested the ﬁrst prototype Kite Steering Unit 1. With an operational height of 800 m, the rated system power was determined to 5 kW average and 40 kW peak. (The Eco Solutions, 2017)

Since 2012 KiteGen is working on the STEM which has a rated output power of 3 MW and is shown in ﬁgure 7. Similar to the EK200 by EnerKíte the STEM is operated in lift mode. The ground base consists of a rotating crane to start and land the Kite. Unlike EnerKíte, STEM uses a soft kite. (Kitegen, 2017)

Figure 7 STEM by KiteGen (Kitegen, 2017).

A more advanced concept is the ARIA100, shown in ﬁgure 8. It operates on lift mode using a moving ground station, formed like a carousel. A series of 52 kites circle at an altitude of about 800 m to rotate a carousel. (Kitegen, 2017)

Figure 8 The ARIA100 system (Kitegen, 2017).

17 Makani

Makani was founded in 2006 and received funding as part of Google.org’s “Renewable En- ergy cheaper than Coal initiative”. Later on, Google bought Makani. It is one of the few companies researching drag mode operated systems. The M600 system utilises eight onboard wind generators with a total rated power of 600 kW (ﬁg. 9). Energy is produced by circling in the air. For take-off and landing the M600 uses the electric generators as motors to hover like an octocopter. After the system has ascended to operating altitude, it switches to power generation mode. Makani promotes the M600 as a fail safe system. In critical weather conditions, or even in the highly improbable scenario of the connecting tether ripping apart, the system is able to switch into hovering mode and ﬂy safely back to the landing zone. (The Eco Solutions, 2017; Makani Power Inc., n.d.)

Figure 9 Makani’s M600 system (Makani Power Inc., 2017).

18 12

TUM DPKVHLC (Drag Power Kite with Very High Lift Coefﬁcient)

Research is conducted at the Technical University of Munich considering a novelty wing design. By implementing multi-part wings, the lift coefﬁcient of the kite can be optimised (see ﬁgure 10). Additionally, the lift is increased by a biplane arrangement. Currently, a prototype is developed and built. Using genetic algorithms, the design optimisation resulted in a system rated at 6.9 MW. (Bauer et al. 2017).

Figure 10 Extract from CFD (Computaional Fluid Dynamic) simulations showing a cross-section of the multi-part wing design (Bauer et al. 2017).

Fig. 7. CFD results for the designed airfoil in monoplane configuration (top) and biplane configuration (bottom): (unstructured) mesh (left), velocity field (middle) and pressure coefficient field (right). All space coordinates are in m and all speed values are in m/s.

F. Proposed Planform Design of a Drag Power Kite lift reduction is induced at or close to the wing tips, it could Fig. 9 shows the proposed planform design of the utility- even help to form a more elliptical lift distribution and thus scale biplane kite with 40 m wing span: The two main wings could increase the Oswald efficiency factor for the considered are connected through vertical wings with symmetric airfoils. simple rectangular wings without washout. The rotors are attached to the eight outer joints of the vertical The kite’s power and tether force are mainly limited by wings and the main wings. Therefore, the downwashes of the the kite’s lift, and the kite’s lift is mainly controlled by the rotors do not affect the tail, which is crucial particularly during flaperons instead of changing the angle of attack. As a reduced hovering. As the rotors are connected close to the ends of the lift considerably also reduces the drag, the kite’s speed might wings, they can not only actuate a large moment, but can also increase to too high values, cf. (9). To limit/control the airspeed compensate partly the wing tip vortexes. Moreover, the vertical in this case, the drag can be increased through the vertical wings at the wing tips function as winglets and also form a box wings by introducing an angle of sideslip controlled by the wing together with the main wings. Therefore, an increased rudders, or by breaks mounted to the vertical wings. Oswald efficiency factor e could be expected. The kite has a The proposed design can be seen as a merger of Makani large tail to control the angle of attack and angle of sideslip Power’s/Google’s and Joby Energy’s design (cf. [6] with [31]): and to compensate the relatively strong moments imposed by The differences to Makani Power’s/Google’s design are the the main wings for changed angles of attack or flaperon angles. use of more than two elements [32] for a very high-lift multi- Moreover, the tail compensates the moments imposed by the element airfoil, and the biplane configuration. The difference to tether, as the tether connection does not coincide with the Joby Energy’s design [31] is the use of a high-lift multi-element center of mass. The elevator is behind and above the center of airfoil instead of a reflex airfoil which cannot achieve high lift mass and can be fully rotated. Therefore, the elevator can also coefficients, but requires no tail to compensate the aerodynamic help to control the pitch during hovering, similar to Makani moments for changes of angle of attack or flaperon angles. Power’s/Google’s kite [30]. The rotors are in front of the main wings andConfidential thus disturb and reduce the airflow which leads to IV. CONCLUSIONS &OUTLOOK a lower lift of these wing sections behind the rotors. However, In this study, the hypothesis was made, that maximizing a as the induction factor of the rotors of a drag power kite is drag power kite’s lift coefficient to or close to its physically rather low (cf. [2, Chap. 28.2.5]) and the affected wing section feasible maximum, also maximizes its power, energy yield as is rather small, the lift reduction is small. Moreover, as this well as maximum allowed costs and profit margin. For that, a Table 1 Prototype data used for detailed investigation (Enerkite GmbH, 2017; Kitegen, 2017; Kitepower, 2017[b]; X-Wind, 2017; Bauer et al. 2017; Makani Power Inc., n.d.).

Mode of Airborne Company Prototype Rated Fail safe LCOE Material Automati- Total Operation System output [Cent/kWh] Weight [t] sation investment power [1k e] [MW]

Drag Hard Makani M600 0.6 fail save to - 31.3 - - Mode wing hover, onboard navigation system

TUM DPKVHLC 4.1 fail save to 5 - full 2,566 hover

Lift Mode Hard wing EnerKite EK200 0.1 rotation, 7 - 12 12.5 full - ﬁxed landing ground station Soft Kite KiteGen STEM 6.8 rotation, 5 - 6 21.3-64 full 1,800 landing

Soft Kite KitePower - 0.1 - 15 10 - -

Lift Mode Numerous KiteGen ARIA100 100 pulled by 2.6 25,400 - 94,000 moving soft kites carousel, ground landing station NTS Round rail 24 pulled by 2 - 4 1,500 full 29,500 20 train, landing 4. Wind conditions in Germany and data

Two steps are necessary, before the RA (Ranalyses) data can be used in this work. Velocity information is available for u-wind and v-wind. The u component is positive for air movement from west to east, while the v component is positive for air movement from south to north (European Centre for Medium-Range Weather Forecasts, 2017). No values for vertical air movement are given. Therefore, the calculation of the absolute wind speed is given by a two-dimensional Theorem of Pythagoras:

p 2 2 vw = vu + vv (4.1)

The second step concerns the transformation from pressure levels in mb to meters of altitude, which are also given in ﬁgure 2. The wind data is available at pressure levels from p ∈ [10 mb; 1, 000 mb]. Following the table from SensorsONE Ltd, 2017, the altitude z can be determined. This procedure is prone to errors, due to the fact, that the conversion table relies on standard conditions:

• Reference pressure:

pref = p(z = 0m) = 1013.25 mb (4.2)

• Reference temperatures:

◦ Tref,1 = T (z = 0 m) = 15 C (4.3) ◦ Tref,2 = T (z = 11000 m) = T (z = 20000 m) = −56.5 C (4.4)

• Thermal gradient:

 ◦  − C ∀ ∈  0.0065 m h [0 m; 11, 000 m]   ◦ ∆T (z) = 0 C ∀h ∈ [11000 m; 20, 000 m] (4.5)  m    ◦C  0.001 m ∀h ∈ [20, 000 m; 32, 000 m]

2 • m Standard acceleration g = 9.81 s

• J Universal gas constant R = 8.31 K·mol

• −2 kg Molar mass of air mmol = 2.9 · 10 mol

21 In order to evaluate the performance of AWESs, probability distributions of present wind speeds is necessary. The following procedure is based on the RA data NOAA/ESRL/PSD, 2017. A grid of four latitudinal values (47.5 N, 50 N, 52.5 N, and 55 N) and five longitudinal values (5 E, 7.5 E, 10 E, 12.5 E, and 15 E) was chosen, as shown in figure 11a. This area encloses Germany. At each grid point data for multiple pressure levels was available. The four pressure levels used in this paper cover a span of the operational heights of all investigated systems: 111 m, 763 m, 1,458 m, and 3,013 m. These altitude values were determined using the conversion table introduced above. The final 3D grid consists of 80 points.

(a) The investigated geographic area cover whole Germany. The (b) Detailed map of Bavaria (Bayerisches grid resolution is 2.5◦lon x 2.5 ◦lat (From maps.google.com). Staatsministerium für Wirtschaft Infrastruktur Verkehr und Technologie, 2010).

Figure 11 Geographic area covered by the wind data used in this paper. Black dots show the grid points. The blue dot shows the location, which was used for system comparison in chapter 5, “Cost Utility Analysis”. The green and orange mark the locations used for north-south and east-west trend-analsis. The dots correspond do the colour mapping in ﬁgures 13 and 14).

Sufﬁcient amounts of data are necessary for a statistically correct evaluation of wind speed probabilities. The data provided on reanalyses.org dates back until 1 Jan 1979. Most current values are available until 31 Dec 2016. Each day, four values are available, which leads to a total of 49,672 data samples at 80 individual grid points.

The Weibull distributions were determined using MATLAB routines. At each grid point, the wind speeds are discretised (rounded) to values from zero to 24. The result are probability distributions, like the ones shown in ﬁgure 12. The variation over different altitudes becomes clearly visible, when comparing the distribution at 111 m altitude, at 518 m and at 3,013 m. In general the mean wind speed shifts to higher values with rising altitude, the probability distributions ﬂattens out. The higher, the more likely are higher wind speeds.

22 20 50N 10E, 111 m 18 50N 10E, 518 m 50N 10E, 3013 m 16

8 probability in % in probability

0 0 5 10 15 20 25 wind speed in m/s

Figure 12 Comparison of the Weibull wind speed distributions at the chosen location for different altitudes.

The operational altitudes of the investigated AWESs range between the 111 m and 3,013 m altitudes of the RA data set. Linear interpolation was used to determine the intermediate values of the individual operational altitudes. The given altitude of 518 m represents the arithmetic average of all operational altitudes of AWESs examined in this paper and is referred to as average altitude in the following. For the cost utility analysis in chapter 5 one grid location was chosen. Its position is 50◦N 10◦E and marked blue in ﬁgure 11. The wind speed range in ﬁgure 11b shows, that the chosen location in general is a good representation of the average wind speed in Bavaria at an altitude of 140 m.

Figure 13 displays the shift of the probability distribution if moving on average altitude from 50◦N 7.5◦E in the west of our chosen location to 50◦N 12.5◦E in the east. No clear shift is recognizable. The maxima of all three distributions are at about 4.5 m/s. When comparing the green and orange graph which represent the most western respectively eastern location two major trends can be identiﬁed.

The green graph ha a lower maximum corresponding to a lower probability of wind speeds between 3 and 6 m/s. At higher wind speeds of 8 to 12 m/s the green graph (more western) reaches higher values of probability than the orange one (more eastern). Overall a the graphs show slight shift to higher wind speeds when going west.

23 20 50N 7.5E, 518 m 18 50N 10E, 518 m 50N 12.5E, 518 m 16

8 probability in % in probability

0 0 5 10 15 20 25 wind speed in m/s

Figure 13 East to west comparison at average altitude.

The last ﬁgure 14 displays the shift of the probability distribution at a more northern point (52.5◦N 10◦E) and a more southern point (47.5◦N 10◦E). The wind speed clearly shifts to higher values when moving north to the coastline of Germany. The probability distributions ﬂattens out. This simulation explains why the north of Germany is in general more attractive for wind energy. Wind speed decreases when moving south shifting the maximum of the distribution to a lower value. It has to be mentioned, that the southern location (47.5◦N 10◦E) is placed in the alpine foothills. The mountainous terrain has a huge impact on the wind speed.

20 47.5N 10E, 518 m 18 50N 10E, 518 m 52.5N 10E, 518 m 16

8 probability in % in probability

0 0 5 10 15 20 25 wind speed in m/s

Figure 14 Nord to South comparison at average altitude.

24 5. Cost Utility Analysis

5.1. Zangemeister Analysis

This section contains the quantitative cost utility analysis according to Zangemeister. It aims to determine a ranking of different solution alternatives , in our case AWES systems. For a successful application of the analysis the two following requirements have to be fulﬁlled:

• Several solution alternatives which differ in the quality of their attributes

• A variety of criteria which has to be considered

The Zangemeister approach is used if not only measurable and ﬁnancial criteria have to be considered. It is also possible to subjectively weigh the different criteria. The procedure works as follows (Walter, 2017):

1. Deploy the target amount matrix (see ﬁg. 15) by listing the alternative solutions row- wise and the evaluation criteria column-wise

2. Rank the alternative solutions based on the values within each column1

3. Each column is weighted subjectively, the sum of all weight factors gj is one

4. Multiply each rank nij with the corresponding weight factor to generate weighted values

5. Adding up the weighted values in each row results in the ﬁnal utility value Ni for each alternative solution i X Ni = gj · nij (5.1) i

The lower the ﬁnal utility value of each AWES, the more applicable it is at the chosen location.

1 The classical cost utility analysis according to Zangemeister does not allow two alternatives with the same ranking within one evaluation criteria (Walter, 2017).

25 5.2. Criteria deﬁnition

This paper focuses on the following six criteria to further further evaluate and compare the selected systems:

1. LCOE: Levelised Costs of Electricity

2. CF: Capacity Factor

3. AEEY: Annual Electrical Energy Yield

4. MP: Material Performance

5. REP: Rated Electrical Power

5.2.1. LCOE The LCOE is deﬁned as the total costs over lifetime divided by the AEP (Annual Energy Production). The total costs over lifetime can be calculated using Initial ICC (Initial Capital Costs), CRF (Capital Recovery Factor) and OMC (Operation and Maintanance Cost) and corresponds to the sum of investment, operational, maintenance and fuel expenditures. (Ahrens et al. 2013, p. 280)

ICC × CRF + OMC LCOE = (5.2) AEP

Only if new technologies are able to supply energy at lower cost than the existing market success is guaranteed. The LCOE is the ﬁrst and fundamental criteria used in this paper and therefore weighed highest with the factor 0.40.

5.2.2. CF The CF describes the actual workload suitability of a system at a given location. It allows the comparison of systems which differ in size and REP. It is follow to (Bauer et al. 2017):

yr AEEY · 8,760h CF = (5.3) REP

The CF prioritised second with a value of 0.25.

5.2.3. AEEY To fulﬁl the world’s need for energy a certain amount of energy is required. Wind energy systems are rated by the maximum possible amount of energy they can produce, given the wind conditions at their location. Therefore the individual performance curve of each system

Pel(vw) is multiplied by location and altitude speciﬁc the Weibull distribution pw(vw) from

26 chapter 4, “Wind conditions in Germany and data”. (Bauer et al. 2017)

8, 760 h Z ∞ AEEY = · pw(vw)Pel(vw)dvw (5.4) yr 0

The probability distributions generated from the RA simulation data are discrete functions. Each integer-valued wind speed within the interval [0 m/s; 24 m/s] is paired with a probability. MATLAB’s trapz function is used to approximate the integral of the numerical data points for the power curves. The discrete form of equation 5.4 follows to:

24 8, 760 h X AEEY = · p (v )trapz(P (v )) (5.5) yr w i el w i=0

The weight of the AEEY was set to 0.2, which corresponds to the third place.

5.2.4. MP Possible material savings are one of the main opportunities of AWESs compared to traditional wind turbines. The amount of material used per system is proportional to the CO2 footprint.

The MP divides the total system weight msystem by the REP in order to analyse how much material is used for each generated kW. Using this criteria different AWES can be compared among themselves and with other technologies in their use of resources.

msystem MP = (5.6) REP

The weight of the MP was set to 0.1.

5.2.5. REP Possible maximum power output under full load is relevant if sufﬁcient wind speeds and peak energy demands overlap. A higher peak performance is rated positive. REP is the last criteria considered in the analysis with a weight of 0.05.

The authors of this paper are aware of the interaction between CF, AEEY and REP. Never- theless, each individual criteria illustrates a relevant system behaviour as outlined above.

All weight values are referring to the ﬁrst CUA (Cost Utility Analysis). For the second and third CUA, the values weight of each dropped criteria were added to all other criteria in equal parts.

27 Alternative criteria REP [kW] AEEY [kWh/y] CP [%] LCOE MP [ct/kWh] [kg/kW]

Makani M600 600 2.36E+06 44.84% - 52.17 TUM DPKVHLC 4100 2.36E+07 65.68% 5 - EnerKíte EK 200 100 5.20E+05 59.32% 12 125 KiteGen STEM 2000 9.26E+06 52.86% 6 32 KitePower 100 4.95E+05 56.49% 15 100 KiteGen Carousel 100000 3.89E+08 44.45% 2.6 254 NTS Rail 24 24000 - - 4 62.4

Figure 15 Target amount matrix.

Additional criteria

There are additional criteria which could not be included in this paper. Partly because of the limited temporal scope, partly because reliable data was not always available.

ICC AWESs are a relatively new technology. To increase the probability to ﬁnd investors and realise a given project the initial total investment costs can be crucial besides already listed criteria. Therefore lower initial investment costs are rated positive in this paper.

System safety A complex range of possible evaluating criteria for AWES can be used to rate system safety. Possible factors are system redundancy, dimensioning of the safety factor of crucial components, fail safety, additional safety systems, lightning protection, the deﬁnition of restriction zones for human activities in the proximity of the ground station, and no-ﬂy zones above them (Bauer, 2015). Not enough reliable data could be found in literature and the authors have not enough experience to judge the different technologies2.

Level of automatisation To compete with current energy production technologies AWES must have a high grade of automation. Fully automated operation is desired. Again, advanced system knowledge and expertise are required to make reliable statements2.

2 Further information can be found in (Bauer, 2015; Bauer et al. 2017).

28 As shown in the utility value matrix in ﬁgure 16 three different analyses have been conducted based on the available data shown in the target amount matrix 15.

1. The ﬁrst CUA compares four different systems based on all ﬁve evaluation criteria. The Carousel system directly followed by STEM, both by the company KiteGen, have the lowest and therefore best utility values of 2.00 respectively 2.05. KitePower is ranked third with a value of 2.70 directly followed by the EK 200 with a value of 2.75. The REP and low LCOE placed both KiteGen systems at good ranks.

2. Within the second CUA ﬁve different systems are compared using four evaluation criteria. MP is not available for the AWES TUM DPKVHLC. The Carousel and STEM by KiteGen still have the lowest utility values of 2.10 and 2.58. The TUM DPKVHLC is ranked third with an utility value of 2.98. EK200 by Enerkíte and KitePower switched ranking positions compared to the ﬁrst CUA, with values of 3.64 and 3.71.

3. Within the third CUA, six different systems are compared based on three evaluation criteria. LCOE data was not available for the AWES Makani M600. TUM DPKVHLC now ranks ﬁrst, followed by both KiteGen systems. The biggest difference in cost utility value is present between the EK200 and the KitePower system. Makani’s M600 ranks last.

29 Alternative criteria REP [kW] AEEY [kWh/y] CP [%] LCOE MP [ct/kWh] [kg/kW] CUA 1 EnerKíte EK 200 3.5 3 1 3 3 KiteGen STEM 2 2 3 2 1 KitePower 3.5 4 2 4 2 KiteGen Carousel 1 1 4 1 4 subjective weighting 0.050 0.200 0.250 0.400 0.100 Σ 1.0 Utility value multiplication value A EnerKíte EK 200 0.18 0.60 0.25 1.20 0.30 2.35 KiteGen STEM 0.10 0.40 0.75 0.80 0.10 2.05 KitePower 0.18 0.80 0.50 1.60 0.20 3.10 KiteGen Carousel 0.05 0.20 1.00 0.40 0.40 2.00

CUA 2 TUM DPKVHLC 2 2 1 2 EnerKíte EK 200 4.5 4 2 4 KiteGen STEM 3 3 4 3 KitePower 4.5 5 3 5 KiteGen Carousel 1 1 5 1 subjective weighting 0.075 0.225 0.275 0.425 Σ 1.0 Utility value multiplication value B TUM DPKVHLC 0.15 0.45 0.28 0.85 1.73 EnerKíte EK 200 0.34 0.90 0.55 1.70 3.49 KiteGen STEM 0.23 0.68 1.10 1.28 3.28 KitePower 0.34 1.13 0.83 2.13 4.41 KiteGen Carousel 0.08 0.23 1.38 0.43 2.10

CUA 3 Makani M600 4 4 5 TUM DPKVHLC 2 2 1 EnerKíte EK 200 5.5 5 2 KiteGen STEM 3 3 4 KitePower 5.5 6 3 KiteGen Carousel 1 1 6 subjective weighting 0.217 0.367 0.417 Σ 1.0 Utility value multiplication value C Makani M600 0.87 1.47 2.08 4.42 TUM DPKVHLC 0.43 0.73 0.42 1.58 EnerKíte EK 200 1.19 1.83 0.83 3.86 KiteGen STEM 0.65 1.10 1.67 3.42 KitePower 1.19 2.20 1.25 4.64 KiteGen Carousel 0.22 0.37 2.50 3.08

Figure 16 Utility value matrix.

30 6. Conclusion and outlook

The EK200 and the KitePower Systems both have far lower REP than all other systems but outstanding CFs of 59.32 and 56.49 % respectively. In comparison the average CP of all systems examined is 50.76 %. With the presented weights, these systems could not achieve better rankings. When comparing MPs, the KiteGen STEM system is clearly ahead with a value of only 32 kg/kW. The average of all systems is 104 kg/kW. The LCOEs of the systems range between 0.026 and 0.120 e/kWh with an average of only 0.069 e/kWh. This puts the relatively new system of AWESs on one level with conventional wind energy systems.

The AEEYs ranged between 495 and 520 MWh/yr for the smaller KitePower and EnerKíte systems. Those calculated values for the given location are very close to the communicated yield values by the companies of 500 and 570 MWh/yr. An average a household consumes 4 MWh/yr.

The AEEYs of the mid large systems lie between 2.36 GWh/yr for the Makani M600 and 9.26 GWh/yr for the KiteGen STEM system. In comparison the wind power station Munich Fröttmaning has a total energy output of 2 GWh/yr and an REP 1,500 kW. With its total weight of 912 t its MPs results in 608 kg/kW. As shown in the last chapter, the MPs of Makani M600 and KiteGen STEM are 52.17 and 32 respectively. The AWESs achieve a higher but comparable yearly energy output while having a MP of only up to 8.5 % of the wind power station. (Stadtwerke München Services Energie und Wasser GmbH, 2007)

The AEEYs of the large systems ranges between 28.1 GWh/yr for the simulation of the TUM DPKVHLC and 389.00 GWh/yr for the KiteGen Carousel. The NTS rail system, which could not be examined further due to the missing performance curve, claims to be able to produce 120 GWh/yr in energy according to detailed simulations and measurement data. One of the biggest onshore wind parks in Germany, the “Altentreptower Windpark” produces around 480 GWh/yr. Enough energy to power around 4,000 homes. With this comparison, it becomes apparent that large AWESs are indeed able to theoretically compete with traditional wind energy systems. (Klimaretter.info, 2013)

31 Even though, the inﬂuence of the ICC was not investigated in detail in this work, they shall be noted here. When related to the REP, the investment cost per kW power can be determined. Compared to traditional wind energy systems, which rate between 1,464 and 1,714 e/kW, AWESs are clearly superior (Wallasch et al. 2013). The values of some systems are:

• KiteGen STEM 900 e/kW

• TUM 877 e/kW

• KiteGen Carousel 940 e/kW

• NTS 1,229 e/kW

As mentioned before, the scope of this work did not allow further investigation or data analysis. Often, complete data sets were not available. The calculations give an estimation of the applicability of AWESs in Germany. However, they cannot replace a full operation site analysis. To generate reliable statements, long-term wind observations are necessary.

The data presented in this work can further be complemented and evaluated over the whole geographic area. Figure 17 shows the AEEY trend of the Enerkíte EK 200. Similar analysis could be done with other criteria to give a holistic assessment of the applicability of AWESs in Germany.

EnerKite EK 200 (190 m)

105 6

2 AEEY in kWh/yr in AEEY 1

0 55 15 50 10

deg latitude (N) 45 5 deg longitude (E)

Figure 17 Possible analysis as a follow up to this work. The AEEY can be calculated for all grid points and allow trends for the applicability in whole Germany.

32 In an interview from 2015, the CEO of Kite Power Systems Limited noted some interesting facts about the future of AWESs and their competitiveness to offshore wind parks. These further emphasize the huge potential which lies within this technology. (Maiki & Maciejowska, 2015)

AWESs use less than 20% of the material which is needed for conventional systems, numbers we could verify within this paper. This lightweight design signiﬁcantly decreases the torque on the foundation, which allows the installation at offshore locations with greater depths than what is possible for conventional systems. (Maiki & Maciejowska, 2015)

In current research more and more of the initial challenges are tackled and overcome. That trend is noticed by investors. Increasing demand and available founding catalyse the maturing of the technology. One of the leading investors in new technologies with humanitarian, Bill Gates, accredited the technology “a 10 per cent chance it’s the magic solution” and expressed his regret that not more governments fund it. (Adams & Thornhill, 2015; Maiki & Maciejowska, 2015)

33 Appendices

34 A. Typical function values

Roughness length

Table 2 Typical values for the roughness length z0 (Masters, 2004, p. 320).

Terrain Roughness length z0 in m

Water surface 0.0002

Open areas with a few windbreaks 0.03

Farm land with some windbreaks more than 1 km apart 0.1

Urban districts and farm land with many windbreaks 0.4

Dense urban or forest 1.6

Friction coefﬁcient

Table 3 Typical values for the friction coefﬁcient α (Masters, 2004, p. 320).

Terrain Friction coefﬁcient α

Smooth hard ground, calm water 0.10

Tall grass on level ground 0.15

High crops, hedges and shrubs 0.2

Wooded countryside, many trees 0.25

Small town with trees and shrubs 0.30

Large city with tall buildings 0.40

35 B. Interpolated wind distributions of investigated systems

Makani M600

15 z = 225 m

10 probability in % in probability