Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13

Contents lists available at ScienceDirect

Journal of Wind Engineering & Industrial Aerodynamics

journal homepage: www.elsevier.com/locate/jweia

On the aerodynamic performance of crosswind kite power systems

Mojtaba Kheiri a, b, *, Fr
ed
eric Bourgault b, Vahid Saberi Nasrabad b, Samson Victor a
a
Concordia University, 1455 de Maisonneuve Blvd. West, Montr
eal, QC, H3G 1M8, Canada
b
New Leaf Management Ltd., 500-1177 West Hastings Street, Vancouver, BC, V6E 2K3, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: This paper generalizes the classical actuator disc theory to the application of crosswind kite power systems. Here,
Airborne wind energy for simplicity, it is assumed that the kite sweeps an annulus in the air, perpendicular to the wind direction (i.e.
Crosswind kite straight downwind configuration with the tether parallel to the wind). It is further assumed that the wind is
Induction factor uniform in space and steady in time. Expressions for the potential power output are obtained, where the effect of
Actuator disc the kite on slowing down the wind (i.e. the induction factor) is taken into account. It is shown that neglecting the
Lift mode
induction factor for a crosswind kite system, even when the factor is small (i.e. a few percentage points), may
Drag mode
Pumping kite
result in consequential overestimation of the amount of power output. Computational fluid dynamics (CFD) re-
On-board generation sults for small- and large-scale kite systems are presented to corroborate the theory. It is shown that the theoretical
model is fairly accurate in predicting the induction factor for such systems.

1. Introduction little material. So far in practice, neither ground-based generation nor

Airborne Wind Energy (AWE) concerns accessing and harnessing
high-altitude wind energy via either flying kites or aerostatic airborne
devices such as balloons, which are usually tethered to the ground. Mean
wind speed increases with altitude due to friction forces produced on the
ground surface – the so-called wind shear phenomenon. This is what
makes reaching higher altitudes appealing because wind power density is
proportional to wind speed cubed (e.g. wind twice as fast (2X) has eight
times (8X) the power). Winds at higher altitudes are also more consistent,
which helps to increase capacity factor. Various airborne concepts and
principles have been proposed and exploited to reach high altitude
winds, where electricity is typically generated either by on-board tur-
bines or by transferring mechanical power to the ground (e.g. unrolling
the tether from a drum); the interested reader is referred to Fagiano and
Milanese (2012); Ahrens et al. (2013); Cherubini et al. (2015); Schmehl
(2018).
The principle of “crosswind kite power” was first introduced in a
seminal paper by Loyd (1980). He showed that large amounts of wind
power could be harvested cheaply by means of an aerodynamically
efficient tethered wing (the kite) flying at high speed transverse to the
incoming wind direction. A crosswind kite may harvest power either in
lift mode (i.e. ground-based generation) or in drag mode (i.e. on-board
generation). The two harvesting modes are theoretically equivalent, ac-
cording to Loyd, in terms of the amount of high power generated with
on-board generation has been favoured, as indicated by experts involved
in research and development of AWE technologies; for more details see
Near Zero (2012). It is, however, their practical implementation (and
unique design challenges) which distinguishes them in terms of opera-
tion, performance, and cost of energy. In fact, very active research and
development is performed worldwide on this subject. For example, see
Bourgault et al. (2017) for an innovative drive-train concept, involving
high efficiency digital hydraulic machines coupled with
hydro-pneumatic accumulators, to be used in the ground-based genera-
tion systems.
The fundamental concept of the crosswind principle is to exploit the
glide ratio (i.e. the lift-to-drag ratio) of a kite to induce a much higher
apparent wind speed at the kite, unlike a static kite which is only sub-
jected to the incoming wind. This phenomenon, which sometimes
referred to as aerodynamic gearing, in turn increases the aerodynamic
driving forces of the system by a square factor of the apparent wind. Both
harvesting modes take advantage of this effect to extract power, i.e.
increased thrust in drag mode versus increased pulling force in lift mode.
Moreover, a crosswind kite flying a closed path in the sky can extract
wind power from a very large area (cf. a static kite or a conventional wind
turbine). Recently, Argatov et al. (2009) proposed a “refined crosswind
motion law” for lift-mode kites, including the effects of the tether drag.
They also obtained a new expression for the maximal value of the me-
chanical power, taking into account the effects of centrifugal,

* Corresponding author. Concordia University, 1455 de Maisonneuve Blvd. West, Montr
eal, QC, H3G 1M8, Canada.
E-mail address: mojtaba.kheiri@concordia.ca (M. Kheiri).

https://doi.org/10.1016/j.jweia.2018.08.006
Received 18 April 2018; Received in revised form 7 August 2018; Accepted 9 August 2018

0167-6105/© 2018 Elsevier Ltd. All rights reserved.

M. Kheiri et al.
Fig. 1. Capture area of a conventional HAWT (left) versus an AWE crosswind
kite system (center). The arrows to the right illustrate a typical wind velocity
gradient. Almost half of the power generated by a HAWT comes from the last
one-third of the blades (Bazilevs et al., 2011), but this requires a large tower
structure to sustain the loads. The crosswind kite only needs a light tether, can
cover a larger area and can reach more powerful winds at higher altitudes.
gravitational and frictional forces. In a subsequent paper, Argatov et al.
(2011) extended the refined crosswind motion law to include effects of
the kite's control and gravity.
Simplistically, a crosswind kite power system (CKPS) parallels a
horizontal-axis wind turbine (HAWT), where the trajectories traced by
the kite in the sky are reminiscent of the turbine blade tip (see Fig. 1). For
a HAWT, approximately half of the power is generated by the last one-
third of the blade (Bazilevs et al., 2011). To capture the same wind
power as a HAWT, a CKPS does not require a massive hub and nacelle, a
large tower structure and reinforced foundation. Instead, it requires only
a single lighter blade (i.e. the kite) and lightweight tether(s) (made of
ultra-high-molecular-weight polyethylene, for example). Furthermore,
reaching higher-altitude winds or increasing capture area for CKPSs in-
curs significantly lower costs. These characteristics make CKPSs very
attractive in terms of the low cost of energy produced. Despite significant
advantages in employing CKPSs, and AWE systems in general, there are
many technical challenges that must be addressed if these systems to
become commercially viable. For example, there are many safety con-
cerns: operating in the proximity to populated areas, or motorways or
power lines, and the possibility of the occurrence of lightning strikes, just
to name a few; see Lunney et al. (2017) for a detailed discussion. The
reader is also referred to Bruinzeel et al. (2018) for an interesting review
of ecological impact of AWE technology.
The pursuit of Crosswind Kite Power technology has only been made
possible in recent years due to the advancements in electronics and
control systems, light structural materials, tether technology, as well as
2
Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13
the emergence of fully autonomous aerial vehicles. For a comprehensive
review of airborne wind technologies and a list of some of the main
players in the field from both academia and industry, the reader is
referred to Ahrens et al. (2013); Cherubini et al. (2015).
It is well-known that there is a theoretical limit to the amount of
power that can be extracted from a freestream via an energy extracting
device. This limit is commonly referred to as the Betz-Joukowsky limit
and may be derived from the actuator disc theory (Okulov and van Kuik,
2009). The actuator disc may be the first and simplest representation of a
rotor, and, in general, of any energy extracting region in the flow. The
actuator disc theory was developed in the late 1800s and early 1900s by
some prominent figures in fluid mechanics, such as Rankine, Froude,
Joukowsky and Betz. According to the theory (Wilson and Lissaman,
1974), a rotor is represented by a permeable disc over which the load is
distributed uniformly. Using the continuity and linear axial momentum
equations, expressions for the axial force acting on the disc (i.e. thrust)
and power extracted from flow are found. These expressions are essential
for preliminary design, performance prediction and load calculations of
real turbine rotors.
Based on the actuator disc theory, the power coefficient, defined as
the ratio of power extracted from flow to that available in an area equal to
the disc's, is Cp ¼ 4að1  aÞ2 , where a is called the induction or inter-
ference factor. The induction factor is a measure of the influence of the
disc on the flow, and it may be correlated to the capability of the disc to
harvest power from flow. In other words, the disc extracts power by
slowing down the flow, and the induction factor serves as an indicator of
flow deceleration. It follows from the above expression that the
maximum value of Cp is 16/27, and it is achieved when a ¼ 1=3. This
means that power extraction from flow is maximized when flow is
decelerated in the vicinity of the disc to 1  a ¼ 2=3 of the freestream
velocity. This shows that power extraction does not increase mono-
tonically with the amount of flow deceleration, but rather it reaches a
limit before it starts decreasing. It is noted that for a  1=2, the theory is
no longer valid as the wake velocity, given by the theory, becomes zero or
even negative, meaning that the flow direction is reversed.
As explained earlier in this section, a CKPS functions similarly to a
conventional wind turbine, and it seems reasonable to use the actuator
disc theory for performance prediction of the kite system. However, some
researchers have expressed reservations about applying the Betz-
Joukowsky limit to crosswind kite systems. For example, Loyd (1980)
states that: “The criteria for the efficiency of a kite or its turbine are
somewhat different from those used by Betz.” He then neglects the kite
induction effect of slowing the wind, arguing that, power is maximized
when induction is minimized (i.e. for a small kite area compared to the
capture area) and that the actuator disc efficiency of the kite is only a few
percentage points. Archer (2013) confirms that the power coefficient of
AWE systems is currently unknown, but she doubts the relevance of the
Betz-Joukowsky limit for AWE systems, claiming that the concept of a
disc-like swept area is not applicable. Also, Costello et al. (2015) argue
Fig. 2. A schematic showing an actuator disc (solid line) of
area A exposed to wind flow velocity v∞ , (from the left) and
moving downwind at constant speed vd . CV1 (dotted black
line) represents the control volume enclosing the actuator disc
and also coinciding with the streamtube, while CV2 (dashed
red line) is the control volume enclosing only the disc; both
control volumes move with the disc at constant speed vd ;
(v∞  vd ) is the velocity of flow entering CV1 through inlet
cross-section area Ai ; vr is the (relative) flow velocity at the
actuator disc, and (vo  vd ) is the velocity of flow leaving CV1
through outlet area Ao . (For interpretation of the references to
colour in this figure legend, the reader is referred to the Web
version of this article.)
M. Kheiri et al. Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13

Table 1
Small-scale kite system parameters.
Nominal powerz  305 KW R (gyration radius) 79.9 m
Airfoil type Clark-Y As 8788 m2
Ak 23.5 m2 v∞ 10.54 m/s
b (wing span) 17.5 m vc 80.31 m/s
AR (wing aspect ratio) 13 e 0

z Calculated according to Loyd (i.e. no induction) subject to wind velocity v∞ and

at the optimal CL ðCL =CD Þ2 .

Table 2
Large-scale kite system parameters.
Nominal powerz  5.5 MW R (gyration radius) 123.3 m
Airfoil type Clark-Y As 41785 m2
Ak 200.7 m2 v∞ 12.50 m/s
b (wing span) 53.94 m vc 91.03 m/s
AR (wing aspect ratio) 14.5 e 1/3

z Calculated according to Loyd (i.e. no induction) subject to wind velocity v∞ and

Fig. 3. A schematic showing an annular flow area dAs swept by a differential
element dAk of a kite (i.e. dAk ¼ c  dr, c being the chord length), located at a
distance r from the axis of rotation.
at the optimal CL ðCL =CD Þ2 .
power by taking into account the airfoil-airmass interaction. In addition,
most recently, Leuthold et al. (2017) examined the relevance of axial and
lateral induction factors to the modelling of lift mode multiple-kite
airborne wind energy systems (MAWESs). They concluded that the
axial induction factor is relevant and should be considered in the power
optimization of a MAWES, while the lateral induction factor can be
neglected.
The primary objective of this paper is to obtain, in a systematic way,
the theoretical potential output power by a crosswind kite, taking into
account the effect of flow retardation by the kite. To achieve this goal, an
extended actuator disc theory is developed, where the effects of disc
motion at a constant speed are considered. The extended theory is then
applied to a CKPS. For the sake of simplicity, a crosswind kite in the
straight downwind configuration (i.e. the kite sweeps an annulus
perpendicular to the wind direction) with the tether aligned with the
wind direction is considered. It is further assumed that the kite is sub-
jected to a flow uniform in space and steady in time. To the best of the
authors’ knowledge, this paper in the manner adopted here is the first of
its kind.
The rest of the paper is organized as follows: in Section 2, the classical
actuator disc theory is extended to the case of a moving disc. Expressions
for the induction factor and extracted power in both lift and drag modes
are obtained in Sections 3–5. In Section 6, some results from CFD sim-
ulations for a small- and a large-scale kite system are discussed, which are
used to support the theory. Finally, in Section 7, the importance of
including the induction factor in calculations of power for CKPSs will be
discussed through some numerical results.

2. Extended actuator disc theory

Fig. 4. Velocity vectors and aerodynamic forces acting on an airfoil section of a
crosswind kite: va , vc , and Vrel represent, respectively, the axial, lateral, and total
relative fluid-kite velocities; L, D, and F are, respectively, the lift, drag, and total
aerodynamic forces; α is the angle of attack, θ is the pitch angle, γ is the angle
between F and the x-axis, and ϕ ¼ θ þ α is the angle between Vrel and the y-axis.
that the Betz-Joukowsky limit “cannot meaningfully be applied to kites”
as the area swept by the kite is generally very large and the kite would
only remove a small fraction of the available wind energy.
Nevertheless, there have been a few attempts to include the effects of
the induction factor in the modelling of CKPSs. Zanon et al. (2014)
explored ‘the impact of the airfoil-airmass interaction on the extracted
power’ of a single- and dual-airfoil drag mode systems. For a single
large-scale kite ( 14 MW), they obtained 27% reduction in the extracted
In this section, the classical actuator disc theory is extended to the
case of a disc (i.e. rotor) moving at constant speed vd , with respect to an
inertial coordinate system, in the flow direction (see Fig. 2).1 The moti-
vation is to apply the extended theory to a crosswind kite which, in
general, may translate downwind while it is also moving crosswind (e.g.
spiraling). This is characteristic of a CKPS in lift mode during the power
harvesting or reel-out phase of a pumping cycle.
Consider a control volume, i.e. CV1, enclosing an actuator disc of area
A, which also coincides with the streamtube formed around the actuator
1
The axial momentum theory has been applied to a wind turbine-driven
vehicle by Sørensen (2016). The present derivation is, however, with greater
elaboration.

3
M. Kheiri et al.
Fig. 6. Axial induction factor, a, at different dimensionless radial distances from
the center of rotation, r=R, in the vicinity of the small-scale kite; r and R are,
respectively, the radial distance and radius of gyration. The kite extends from r=
R ¼ 0:912 to r=R ¼ 1:088.
disc. It is assumed that CV1 is moving with the disc at speed vd . From the
continuity equation, we may write
m_ ¼ ρAi ðv∞  vd Þ ¼ ρAvr ¼ ρAo ðvo  vd Þ; (1)
where m_ represents the mass flow rate through the control volume, ρ is
the mass density of fluid flow, v∞ and vo are, respectively, the absolute
flow velocities at sufficiently far upstream (i.e. inlet of CV1) and down-
stream (i.e. outlet of CV1) of the disc; Ai and Ao are, respectively, the flow
area at the inlet and outlet of the control volume, and vr is the relative
flow velocity at the actuator disc.
By applying the linear momentum equation to CV1, we can obtain the
4
Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13
Fig. 5. CFD simulation setup for the large-scale kite:
(a) solution domain is 19  6  6 km (L  W  H); it
contains 3.8M nodes and 5.8M cells with the
maximum skewness of 0.82; boundary conditions are:
velocity inlet, pressure outlet, symmetry over the top,
bottom and sides of the domain, and no-slip over all
the faces of the kite, (b) the rotating domain (with the
radius of 180 m); a non-conformal domain is created
between the stationary and rotating domains, (c)
mesh around the kite airfoil section (Clark Y airfoil,
chord: 3.72 m, maximum thickness: 0.43 m, No. of
elements around the airfoil: 260, bias factor at the L.E.
and T.E.: 2, No. of inflation layers around the airfoil:
22, the growth ratio: 1.2), and (d) solution domain
mesh (top-view) with the ‘sphere of influence’
downstream of the kite.
Fig. 7. Axial induction factor, a, at different dimensionless radial distances from
the center of rotation, r=R, in the vicinity of the large-scale kite; r and R are,
respectively, the radial distance and radius of gyration. The kite extends from r=
R ¼ 0:797 to r=R ¼ 1:203.
axial force, T, acting on the flow, the reaction of which is applied to the
disc. That is
T ¼ ρAo vo ðvo  vd Þ  ρAi v∞ ðv∞  vd Þ ¼ mðv
_ o  v∞ Þ; (2)
in which equation (1) has been utilized; also, it has been assumed that the
control volume is under a uniform ambient pressure p∞ , and thus zero net
pressure force acting on the boundaries of the control volume.
Furthermore, the linear momentum equation may be applied to the
control volume enclosing only the disc, i.e. CV2:
T ¼ ðpd  pdþ ÞA; (3)
M. Kheiri et al. Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13

Table 3
Induction factor (a) and thrust (T) values, obtained computationally and analytically, for the large-scale kite system at two different rotational speeds, i.e. vc ¼ 91.03 m/s
and vc ¼ 79:6 m/s.
CFD Theory Relative error %

vc ¼ 91:03 vc ¼ 79:6 vc ¼ 91:03 vc ¼ 79:6 vc ¼ 91:03 vc ¼ 79:6

a 0.126 0.113 0.119 0.095 5.6 15.9

T (KN) 759.421 630.281 683.596 586.368 10.0 7.0

Fig. 8. Normalized lift power as a function of the ratio of reel-out speed to

Fig. 10. Induction factor in lift mode as a function of the ratio of reel-out speed
freestream velocity, e, for different values of the solidity factor, σ . For all the
to freestream velocity, e, for different values of the solidity factor, σ . For all the
curves, CL ¼ 1:0 and ðCL =CD Þ ¼ 10.
curves, CL ¼ 1:0 and ðCL =CD Þ ¼ 10.

Fig. 9. Lift power overestimation as a function of the ratio of reel-out speed to Fig. 11. Normalized drag power as a function of the ratio of thrust of on-board
freestream velocity, e, for different values of the solidity factor, σ . For all the turbines to kite drag, κ, for different values of the solidity factor, σ . For all the
curves, CL ¼ 1:0 and ðCL =CD Þ ¼ 10. curves, CL ¼ 1:0 and ðCL =CD Þ ¼ 10.

where it has been assumed that the flow velocity does not change across equations for these streamlines may be written as:
the actuator disc, only pressure does; pd and pdþ represent, respectively, 1 1
the static pressure just before and after the disc. p∞ þ ρðv∞  vd Þ2 ¼ pd þ ρv2r ; (4)
2 2
Consider a streamline extending from the inlet to the actuator disc
and another one from the disc to the outlet, and apply Bernoulli's equa- 1 1
tion to these streamlines. As shown in Appendix A, the Bernoulli pdþ þ ρv2r ¼ po þ ρðvo  vd Þ2 ; (5)
2 2

5
M. Kheiri et al. Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13

By equating equations (2) and (7) and letting m_ ¼ ρAvr (refer to

equation (1)), one may obtain

1
vr ¼ ½ðv∞  vd Þ þ ðvo  vd Þ; (8)
2

which means that the relative flow velocity at the actuator disc is the
average of the inlet and outlet relative flow velocities.
It appears reasonable to define the induction or interference factor a
in connection to vr , as

vr  ðv∞  vd Þð1  aÞ: (9)

Equation (9) means that retardation/deceleration happens to the
relative incoming flow velocity/wind speed ðv∞  vd Þ. Using equations
(8) and (9) and letting vd ¼ ev∞ , one may obtain

vo ¼ v∞ ½1  2að1  eÞ: (10)

Thus, the axial force or thrust (refer to equation (7)) may be re-written
as
 
1
T¼ ρAv2∞ 4að1  aÞð1  eÞ2 : (11)
Fig. 12. Drag power overestimation as a function of the ratio of thrust of on- 2
board turbines to kite drag, κ, for different values of the solidity factor, σ . For
all the curves, CL ¼ 1:0 and ðCL =CD Þ ¼ 10. It is interesting to see from equation (11) that the thrust of a moving
actuator disc is the same as a stationary actuator disc scaled by a factor of
ð1  eÞ2 . In other words, to find the thrust of an actuator disc moving with
the constant speed of vd ¼ ev∞ , one may simply replace v∞ in the thrust
formula for a stationary disc by v∞  vd ¼ ð1  eÞv∞ , i.e. the relative
freestream velocity. However, one should note the difference between
the induction factor for a stationary disc, defined as vr  v∞ ð1  aÞ, and
that for a moving disc, defined in equation (9).
Using equations (9) and (11), the power extracted by the actuator disc
from flow may be obtained as
 
1
P ¼ Tvr ¼ ρAv3∞ 4að1  aÞ2 ð1  eÞ3 : (12)
2
Similarly, one may conclude that the power for an actuator disc
moving at constant speed vd ¼ ev∞ may be obtained from the formula for
a stationary disc, provided that v∞ is replaced by ð1  eÞv∞ .
The thrust coefficient CT and the power coefficient CP for a moving
disc/rotor may be defined as follows:
1 
ρAv2∞ 4að1  aÞð1  eÞ2
CT ¼ 2
¼ 4að1  aÞð1  eÞ2 ; (13)
1
2
ρAv2∞

1 
ρAv3∞ 4að1  aÞ2 ð1  eÞ3
CP ¼ 2
¼ 4að1  aÞ2 ð1  eÞ3 : (14)
Fig. 13. Induction factor in drag mode as a function of the ratio of thrust of on- 1
2
ρAv3∞
board turbines to kite drag, κ, for different values of the solidity factor, σ . For all
the curves, CL ¼ 1:0 and ðCL =CD Þ ¼ 10. By letting e ¼ 0 in equations (13) and (14), one can recover expres-
sions for CT and CP based on the classical actuator disc (i.e. for a sta-
tionary disc).
which show that relative flow velocities appear in the dynamic pressure
Finally, the efficiency of the actuator disc may be defined as the ratio
terms.
of extractable power to the wind power available to the disc (refer to
Using equations (4) and (5) and the fact that po ¼ p∞ , we can obtain
Wilson and Lissaman, 1974):
ðpd  pdþ Þ as2
1 
1
 ρAv3 4að1  aÞ2 ð1  eÞ3
pd  pdþ ¼ ρ ðv∞  vd Þ2  ðvo  vd Þ2 : (6) ηAD ¼ 1 ∞ 
2
¼ 4að1  aÞ: (15)
2 2
ρAv3∞ ð1  aÞð1  eÞ3
Substituting equation (6) into equation (3) yields
3. Induction factor for a crosswind kite
1

T ¼ ρA ðv∞  vd Þ2  ðvo  vd Þ2 : (7)
2 The annulus flow derivation presented in this section follows similar
steps as in Wilson and Lissaman (1974). For simplicity, it is assumed that
the induction factor in the crosswind (or lateral) direction is negligible
0
2 (i.e. a ¼ 0). Normally, the axial induction factor a is found at any radial
According to the actuator disc theory, it is assumed that the outlet is suffi-
ciently far from the disc such that the static pressure is recovered and reaches
distance r from the center of rotation. This is performed by equating the
p∞ . axial force generated in the annular element of width dr to the axial force

6
M. Kheiri et al. Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13

Fig. 14. A schematic showing the effect of a moving

disc on the flow field. The top figure shows the situ-
ation at time t, while the bottom one corresponds to
time t þ Δt. Points A and B are fixed in space, while
points ① to ④ are moving with the disc; in fact, point
① is where the flow velocity is v∞ ; point ④ is where
the flow velocity is vo , and points ② and ③ are just
before and after the disc; Xi ði ¼ 1; 2Þ are inertial co-
ordinate systems, while xi ði ¼ 1; 2Þ are moving with
the disc. Also, ℓ1 and ℓ2 represent the extent of the
streamtube, respectively, upstream and downstream of
the disc; Δs is the distance travelled during the period
Δt; lastly, v represents the absolute flow velocity at the
disc.

axial and crosswind components of Vrel . Also, θ is the pitch (or twist)
angle, α is the angle of attack, γ is the angle between F and the x-axis, and
ϕ ¼ θ þ α is the angle between the relative flow velocity, Vrel , and the y-
axis.
Equating the total axial force generated in the annular area As swept
by the kite (refer to equation (11)) to the total axial aerodynamic force
acting on the kite yields

1 1
ρAk Vrel
2
ðCL cos ϕ þ CD sin ϕÞ ¼ ρAs v2∞ 4að1  aÞð1  eÞ2 ; (16)
2 2

where Ak is the kite planform area; CL and CD are the lift and drag co-
efficients of the kite.4
Using Fig. 4, one may obtain γ as
   
va CD
γ ¼ tan1  tan1 ; (17)
vc CL

which confirms that γ ’ 0 when ðva =vc Þ; ðCD =CL Þ << 1. If this condition
holds, one may write
Fig. 15. Optimal values of the ratio of thrust of on-board turbines to kite drag,
κ , as a function of C. v a CD
tanϕ ¼ ’ ; (18)
vc CL
predicted by the blade element aerodynamic considerations (see Fig. 3).
Here, for the sake of simplicity, instead of finding local axial induction or alternatively,
factors, an average induction factor is obtained, assuming that the ve-  
locity of the kite in the crosswind direction is constant along its span.3 A CL
vc ¼ ð1  aÞð1  eÞv∞ ; (19)
more elaborate theory considering the lateral induction factor and radial CD
distribution of the axial induction factor is outside the scope of this paper
in which va has been replaced with v∞ ð1  aÞð1  eÞ as it is the relative
and will be the subject of a future publication.
flow velocity in the axial direction in the vicinity of the kite (refer to
Fig. 4 shows an airfoil section of a kite with the aerodynamic forces
equation (9)).
acting on it: L, and D represent, respectively, lift and drag forces, and F is
Equation (16) may be re-written as
the resultant force. For simplicity, we consider a straight downwind
configuration, where the kite sweeps an area perpendicular to the wind  
1   C 1
direction while also reeling out (i.e. translating downwind) at a constant ρAk v2a þ v2c CL þ CD D ’ ρAs v2∞ 4að1  aÞð1  eÞ2 ; (20)
2 CL 2
speed of vd ¼ ev∞ . As shown in the figure, Vrel is the relative flow velocity
in the vicinity of the airfoil section, and va and vc are, respectively, the in which cos ϕ ’ 1 and sinϕ ’ tanϕ ’ ðCD =CL Þ have been utilized.
The left-hand side of equation (20) may be further simplified,

3
Further assumptions should be made for obtaining the average axial in-
4
duction factor; for instance, the airfoil sections along the span should be In the present formulation, in general, CD also includes the normalized drag
identical. coefficient of the tether, in addition to the drag coefficient of the kite.

7
M. Kheiri et al. Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13

assuming that ðva =vc Þ << 1 and the fact that CD ðCD =CL Þ << CL , as 5. Drag mode power

1 1
ρAk v2c CL ’ ρAs v2∞ 4að1  aÞð1  eÞ2 : (21) In drag mode, wind power is not extracted by pulling a load down-
2 2 wind; rather, power is produced by loading the kite with additional drag:
By substituting equation (19) into equation (21), the following on-board turbines carried by the kite use the high relative wind speed to
equation yields generate power. In this section, the extended form of actuator disc theory
   2 is used to obtain the expression for power extracted by the kite in drag
a 1 Ak CL mode (i.e. with on-board generation).
’ CL : (22)
1  a 4 As CD It is assumed that the on-board turbines have a total thrust, Tt , which
is κ times the drag of the kite (i.e. Tt ¼ κD). The power extracted from
Equation (22) gives the average axial induction factor for a kite with a
flow by the on-board turbines then may be written as
planform area of Ak , swept area of As and aerodynamic coefficients CL
and CD . The area ratio Ak =As may be called the solidity factor of the kite, in 1
accordance with the terminology used for wind turbines. It can easily be PD ¼ Tt vc ¼ κDvc ’ ρAk v3c κCD ; (26)
2
concluded from equation (22) that increasing either the solidity factor, CL
and/or ðCL =CD Þ will increase the induction factor. It should be reminded where Vrel ’ vc has been used in the expression for the kite drag, D.
that this equation is valid for cases where ðva =vc Þ; ðCD =CL Þ << 1.5 Re-writing equations (19) and (22) for a kite in drag mode, where e ¼
0, results in the following equations
4. Lift mode power    
CL 1 CL
vc ¼ ð1  aÞv∞ ¼ ð1  aÞv∞ ; (27)
In this section, starting from the expression found for the axial force CD þ κCD 1 þ κ CD
or thrust acting on the moving actuator disc (i.e. equation (11)), one may
   2    2  2
find the expression for the potential power output from a kite in lift mode a 1 Ak CL 1 Ak CL 1
’ CL ¼ CL : (28)
(during the harvesting phase of a pumping cycle),6 that is 1  a 4 As CD þ κCD 4 As CD 1þκ
  Equation (28) shows that the induction factor for the kite in drag
1
PL ¼ Tvd ¼ ρAs v3∞ 4að1  aÞð1  eÞ2 e; (23) mode, in addition to being dependent on the solidity factor and on the lift
2
and drag coefficients, is also a function of κ (cf. equation (22)).
where As represents the area (e.g. annulus) swept by the kite, in a similar Substituting equation (27) into (26) yields the following generalized
fashion that A is normally used for the rotor disc area of a wind turbine. equation for drag mode power:
Recall here that the kite translates downwind with the speed of vd ¼ ev∞ .    2
Using equations (22) and (23), the lift mode power may be expressed 1 C κ
PD ¼ ρAk v3∞ CL L ð1  aÞ3 : (29)
in terms of kite's planform area, Ak , and its aerodynamic characteristics: 2 CD ð1 þ κÞ3

  "    2 # Setting a ¼ 0 in equation (29) recovers Loyd's expression for the

1 1 Ak CL
PL ¼ ρAs v∞ 4
3
CL ð1  aÞ ð1  aÞð1  eÞ2 e power output in drag mode. Hence, for a negligible induction factor, it
2 4 As CD follows that the maximum PD is achieved when κ ¼ 1=2.
   2
1 C A closed-form solution for the maximum drag mode power when
¼ ρAk v3∞ CL L ð1  aÞ2 ð1  eÞ2 e: (24) induction is non-negligible cannot be found, in contrast to the lift mode.
2 CD
This is mainly because the induction factor itself is also a function of κ,
Notice that setting a ¼ 0 in equation (24) recovers Loyd's expression which adds to the complexity of the derivation; see equation (28).
for the lift mode power. However, the maximum drag mode power may be obtained numerically,
If we assume that a and e are independent, it follows that PL is and it can be shown that the value of κ leading to the maximum power
maximized when e ¼ 1=3. Thus, the maximum lift mode power may be may be obtained from the only positive real root of the following cubic
obtained as equation (see Appendix B for the derivation of the equation, as well as the
   2 graph showing the variation of optimal k as a function of C):
4 1 C
PL;max ¼ ρAk v3∞ CL L ð1  aÞ2 : (25)
27 2 CD 2κ 3  3κ 2 þ 4Cκ þ C þ 1 ¼ 0; (30)
Equation (25) shows that neglecting a small induction factor of a few    2
percent in power calculations may result into significant overestimation where C ¼ 1 Ak CL
4 As CL CD .
of the maximum power output in lift mode; for example, neglecting a 5%
induction factor results in nearly 10% power overestimation, which is The drag mode power may also be expressed in terms of the swept
appreciable. area, As , by combining equations (28) and (29), as follows
 
1 κ
PD ¼ ρAs v3∞ 4að1  aÞ2 : (31)
2 1þκ

6. CFD simulations

5
A more accurate form of equation (22) may be obtained by retaining Some CFD simulations results are presented in this section in order to
   2  2 !2 support the theory developed in Section 3 for obtaining the induction
CD ðCD =CL Þ and v2a in equation (20), that is 1a
a
¼ 14 AAks CL CCDL 1 þ CCDL .
factor of a CKPS. Two kite systems of different scales were used in the
However, equation (22) is preferred for its simpler form. simulations: a system with the nominal power of 305 KW, that we call the
6
To obtain the net average power over a pumping cycle, PC , one has to also small-scale system, and another system with the nominal power of
take into account the power required for the retraction or reel-in phase, Pri , as 5.5 MW, that is called the large-scale system from now on. The nominal
well as the time required for each phase, in a weighted sum as follows: PC ¼ power values were calculated based on Loyd (1980)'s equations where no
ðPL tL þ Pri tri Þ=ðtL þ tri Þ, where tL and tri are the reel-out and reel-in time period, induction is considered and CL ðCL =CD Þ has the optimal value. The design
respectively.

8
M. Kheiri et al.
parameters for these kites are given in Table 1 and 2, respectively. The
kites were modelled as straight wings in the straight downwind config-
uration to conform to the assumptions made in Section 3.
In the present paper, an incompressible, unsteady, Reynolds-averaged
Navier-Stokes (URANS) model coupled with the Standard k  ε turbu-
lence model (with y þ ¼ 30) was solved numerically. The simulations
were performed using ANSYS Fluent R16.2 (Academic) on a computer
equipped with 12 processors and 96 GB of RAM. For the 5.5 MW system,
the simulations took 25 days during which the kite completed 27 circular
revolutions. The solution time step was considered to be equivalent to 1
deg of rotation. For example, for the large-scale system, the time step is
equal to 0.0236 s since the rotational speed of the kite is 7.05 rpm and the
gyration radius is R ¼ 123.3 m (see Table 2). The solution is said to have
converged and reached post-transient conditions if two criteria are
satisfied. The first criterion is that the residual values should fall below
105 . The second criterion is met by monitoring the average streamwise
flow velocity over multiple concentric rings across the wake width at
several points from the rotor (e.g. at x ¼ 5R and x ¼ 8R). The solution is
assumed to have converged when for any two successive revolutions of
the kite, the average difference between the velocities at equivalent time
steps within the period is less than 1%.
The rationale to use the k  ε turbulence model is its reasonable ac-
curacy, fast convergence and relatively inexpensive computational time
for modelling attached turbulent flows over the kite surface. This is
supported by several studies found in the literature on similar systems;
see for example, Barthelmie et al. (2007); Thumthae and Chitsomboon
(2009); Avila et al. (2013). But one should note the limitations of the k 
ε model, the major one being it cannot accurately predict the complex
separated flows near the surface. For the present case studies, however,
fully turbulent attached flows are expected, which may fairly be pre-
dicted via the Standard k  ε model, as also indicated by Rhie and Chow
(1983); Wolfe and Ochs (1997). Nevertheless, more CFD studies using,
for example, URANS (with the k  ω SST turbulence model) and LES flow
solvers are desirable in order to verify the present CFD results. These
studies are currently underway, but their results will be deferred to a
future publication.
Fig. 5(a) and (b) show the solution domain setup for the large-scale
kite. Similar settings (but with a different domain size and number of
cells and nodes) have been used for the CFD simulation of the small-scale
kite. The solution domain is 19  6  6 km (L  W  H) in size. Boundary
conditions are: uniform constant flow velocity at the inlet (or velocity
inlet), zero gauge pressure at the outlet (or pressure outlet), symmetry over
the top, bottom and sides of the domain, and no-slip over all the faces of
the kite. The outlet is placed 15 km from the center of the rotating
domain to avoid adverse pressure gradients. The SIMPLE algorithm is
used for pressure-velocity coupling.
The mesh quality parameters were set according to the criteria
described in the user guide of Ansys Inc. (2009). The number of nodes
and elements in the solution domain are 3.8 and 5.8 million, respectively,
with the maximum element skewness of 0.82 (this is well below 0.95 that
is the limit value recommended by ANSYS). A hybrid mesh technique is
used for the computational domain. Hexahedral elements are created
around the kite to efficiently capture the flow. The first layer height is
defined by the y þ conditions, and inflation layers are used with the
growth rate of 1.2. The airfoil is split into segments to impose higher
mesh density around the leading and trailing edges. There are 260 ele-
ments around the airfoil that are swept across the kite surface to ensure
uniform inflation layers and consistent mesh quality; see Fig. 5(c). The
kite and its attached mesh are contained in a rotating domain (with the
radius of 180 m) which is itself contained in the larger flow volume. A
non-conformal domain is created between the stationary and rotating
domains, and the sliding mesh technique is employed at all
non-conformal interfaces, which allows the nodes to move rigidly at
given dynamic mesh zone. The sphere of influence is used at the down-
stream side of the kite to effectively capture the developed wake, as also
9
Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13
seen in Fig. 5(d).
Numerical results for axial induction factor a in the vicinity of a kite
for different radial locations along the kite span were obtained following
the method proposed by Johansen and Sørensen (2004) for wind tur-
bines. Fig. 6 shows such a plot for the small-scale kite. The
post-processing calculations were performed for a region extending from
a distance of one gyration radius upstream (  R) to one gyration radius
downstream ( þ R) of the kite, and the kite was divided into 5 equal
segments spanwise. The average induction factor over the kite was found
to be a ¼ 0:046. Fig. 7 shows a similar plot but for the large-scale system.
In this case, the kite was divided into 9 unequal segments spanwise. The
induction factor, as expected (given this kite's higher solidity factor), is
much higher – the average induction factor over the large-scale kite is
0.126.
The analytical induction factor can be obtained iteratively using ex-
pressions presented in Section 3. This is provided that values of CL versus
angle of attack, α, are available. Here, experimental CL  α values given
in (Silverstein, 1935, Table XVIII) for a wing with Clark Y airfoil section,
aspect ratio AR ¼ 6, and the Reynolds number Re ¼ 6:12  106 have
been used.7 Using equations (19) and (22), one can easily find the
following equation:
   2
1 Ak vc
a2  a þ CL ¼ 0: (32)
4 As ð1  eÞv∞
Knowing values of ðAk =As Þ, ðvc =v∞ Þ and e, and starting with a guess
value for CL and using equation (32), the induction factor a is obtained. In
the second step, with the obtained value of a and equation (18), the angle
of attack (assuming zero twist, i.e. θ ¼ 0) is obtained. In the third step,
the tabulated data is interpolated with the obtained value of α to get a
new value of CL . These steps with the new value of CL should be repeated
until converged values are obtained for CL , a, and α.
The analytical value of the induction factor obtained with the pro-
cedure described above for the small- and large-scale kites are 0.037 and
0.119, respectively. The relative error between the CFD results and
analytical predictions are 20% and 6% for the small- and large-scale
kites, respectively. If, instead, CL  α values for a Clark Y airfoil (i.e. an
infinite aspect ratio wing) are used,8 the induction factor for the small-
and large-scale kites are found to be 0.047 and 0.150, corresponding to
2% and 19% relative error, respectively. It may then be concluded that if
CL  α values for the exact geometry of the kites and flow conditions
were available for use in the analytical calculations, the relative error
between the predictions and the CFD results would be lower than 20%.
Further results have been obtained for the large-scale system. They
include the induction factor and thrust (or axial load) for a different
rotational speed vc ¼ 79:6 m/s. Table 3 summarizes the results obtained
analytically and computationally for vc ¼ 91:03 m/s and vc ¼ 79:6 m/s.
A fairly good quantitative agreement between CFD and analytical results
can be observed from the table. One may also notice that the power
values obtained by multiplying thrust values in Table 3 by the reel-out
speed (i.e. vd ¼ ev∞ ’ 4:17 m/s) are different from 5.5 MW, which is
the optimal power based on Loyd's theory. This is because the kite
rotating at either vc ¼ 91:03 m/s or vc ¼ 79:6 m/s does not correspond to
the optimal condition, and thus the output power is less than the optimal
value. Theoretically, for the given system, the kite should fly at the speed
of vc ¼ 127:5 m/s to extract the maximum power.
By carefully examining the wake flow region of the kites in the so-
7
The average Reynolds number based on chord length for the small-scale kite
is approximately 7  106 , and that for the large-scale kite is around 2:2  107 .
Experimental values for CL and CD given in Silverstein (1935) show very little
sensitivity to Re beyond Re ¼ 6:12  106 . Thus, the experimental values at Re ¼
6:12  106 are used for calculation of the induction factor for both kites.
8
The values were obtained at Re ¼ 6  106 using XFLR5 v6.39 program, in
the absence of experimental values for a high-aspect ratio wing or an airfoil at
such high Re.
M. Kheiri et al.
lution domain (not shown here), it is confirmed that a highly-turbulent,
low-speed flow is developed downstream the kites. This is particularly
crucial for layout design and optimization of kite farms; for more details,
see Kheiri et al. (2017). It should be noted that, the reliability of these
results needs to be assessed using experimental measurements. However,
such measurements, to the best of the authors' knowledge, are not
currently publicly available for crosswind kite power systems. On the
other hand, several programs (or toolkits) that are widely used for wake
flow modelling of conventional wind turbines, such as ECN's WAKE-
FARM (Schepers, 2003), and NUTA (Magnusson et al., 1996) have
implemented the k  ε turbulence model for the flow calculations (for
more details, please see Barthelmie et al., 2007). Thus, because of the
fundamental similarities between crosswind kites and conventional wind
turbines, the present modelling approach may be viewed acceptable.
Nonetheless, as previously mentioned in this section, more numerical
studies using URANS (with the k  ω SST turbulence model) and LES
flow solvers are presently underway and will be presented in a future
publication.
7. Numerical results
In this section, some selected numerical results are presented for
crosswind kites in both lift and drag modes. More specifically, variation
of the normalized power, defined as P=ð1=2ρAk v3∞ Þ, power over-
estimation, i.e. ðPLoyd  PÞ=PLoyd  100, PLoyd being the power predicted
by Loyd (1980)'s expressions, and the induction factor are given as a
function of: (a) the ratio of reel-out speed to freestream velocity for a lift
mode kite, and (b) the ratio of on-board turbines thrust to kite drag for a
drag mode kite.
7.1. Crosswind kite in lift mode
Fig. 8 shows variation of the normalized power in lift mode,
PL =ð1=2ρAk v3∞ Þ, as a function of the ratio of reel-out speed to freestream
velocity, e, for different values of the solidity factor typical of crosswind
kite systems (i.e. σ ¼ Ak =As ¼ 0.001 to 0.01). It is assumed that CL ¼ 1:0
and ðCL =CD Þ ¼ 10 (note that CD also includes the tether drag) for all the
curves in this figure and in all subsequent figures. For comparison pur-
poses, the normalized power curves for σ ¼ 0 and σ ¼ 0:0345 were also
shown in the figure. A solidity factor of zero corresponds to Loyd's
formulation (i.e. an infinite capture area), and σ ¼ 0:0345 was used in
Burton et al. (2011) as the solidity factor for a typical, modern,
three-blade conventional wind turbine.
As seen from Fig. 8, by increasing the solidity factor, the normalized
power decreases. This is to be expected as a higher solidity factor cor-
responds to a smaller swept area for a given kite area, meaning that less
wind power is available for extraction. As seen from the figure, the peak
power for a given kite is decreased by 33% when the solidity factor is
increased from 0.001 to 0.01. Also, the peak power occurs at e ¼ 1=3,
and it is independent of the solidity factor. Moreover, note how much less
power is generated as the solidity of the crosswind kite approaches that of
a conventional wind turbine.
Fig. 9 shows how much overestimation is made in lift power calcu-
lation if one uses Loyd (1980)'s formulation (i.e. neglecting the induction
factor) instead of the formulation presented in this paper. For example, a
realistic solidity factor for a large pumping kite system may be approxi-
mately 0.005 (refer to Table 2), and the corresponding power over-
estimation for such a kite would be approximately 21%. As seen from the
figure, the overestimation is strongly dependent on the solidity factor: it
becomes more significant when the solidity factor is increased. This is
specially important because it is expected that the solidity factor values of
future utility-scale kite power systems to be much higher than the small,
experimental-scale systems which have so far been demonstrated.
Conversely, the usual low-solidity factors of those experimental systems
lead to low induction factor values, and therefore, the errors made in
10
Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13
power predictions are small when the induction effects are neglected.
The errors are, in most cases, likely to fall below the noise levels of the
complex instrumentation required. As also seen from the figure, the de-
gree of overestimation is independent of the ratio of reel-out speed to
freestream velocity.
Fig. 10 shows variation of the induction factor in lift mode, aL , as a
function of the ratio of reel-out speed to freestream velocity, e, for
different values of the solidity factor, σ . As seen from the figure, except
for very small values of σ , the induction factor is not negligible. For
example, for σ ¼ 0:005, the induction factor is about 0.11, corresponding
to approximately 21% power overestimation. As also seen, the induction
factor in lift mode is independent of e,9 but it increases as the solidity
factor is increased. The zero induction factor corresponds to Loyd
(1980)'s.
7.2. Crosswind kite in drag mode
Fig. 11 shows variation of the normalized power in drag mode,
PD =ð1=2ρAk v3∞ Þ, as a function of the ratio of thrust of on-board turbines to
kite drag, κ, for a typical range of solidity factors, i.e. σ ¼ Ak =As ¼ 0.001
to 0.01. Here again, curves for σ ¼ 0:0 (Loyd's formulation) and σ ¼
0:0345 (a modern HAWT) are shown for comparison purposes.
As seen from the figure, similarly to the lift-mode power, the
normalized power in drag mode decreases as the solidity factor is
increased. The peak power is decreased by approximately 22% when the
solidity factor is increased from 0.001 to 0.01. In contrast to the lift-mode
power, the peak power occurs at a different κ as the solidity factor is
varied. This is because the induction factor for a drag-mode kite is also
dependent on κ, as also previously explained in Section 5. As seen from
the figure, the peak shifts towards a higher κ as σ is increased. For
example, for zero solidity factor, the peak power occurs at κ ¼ 1=2, while
for σ ¼ 0:01, it occurs at κ ’ 0:66.
Fig. 12 shows how much overestimation is made in drag power
calculation if one uses Loyd's formulation (i.e. neglecting the induction
factor) instead of the formulation presented in this paper. As seen, the
amount of overestimation is dependent on both the ratio of thrust of on-
board turbines to kite drag and the solidity factor: it becomes larger when
the solidity factor is increased and the thrust to drag ratio is decreased.
For example, for a small solidity factor σ ¼ 0:001, drag power over-
estimation is relatively modest and varies between 2% (at κ ¼ 1) and 7%
(at κ ¼ 0:01). For a larger solidity σ ¼ 0:01, that error becomes very
consequential as it varies between 17% (at κ ¼ 1) and 48% (at κ ¼ 0:01).
Fig. 13 shows variation of the induction factor in drag mode, aD , as a
function of the ratio of thrust of on-board turbines to kite drag, κ, for
different values of the solidity factor, i.e. σ ¼ Ak =As ¼ 0.001 to 0.01. As
expected from equation (28), the induction factor for a drag-mode kite
increases, and may become quite significant, as σ increases and/or κ
decreases.
As seen from Fig. 8 and Fig. 11, the peak normalized power in lift
mode for a kite with σ ¼ 0 (i.e. infinite capture area) is approximately
14.81 (assuming CL ¼ 1:0 and ðCL =CD Þ ¼ 10) which is the same as in
drag mode. This agrees well with Loyd (1980)'s results. However, for a
non-zero solidity factor (thus a non-zero induction factor) system, the
maximum drag power looks greater than the maximum lift power. One
may then quickly jump into conclusion that the drag-mode power gen-
eration is more efficient than the lift-mode power generation. However,
this conclusion may be wrong as a kite in lift mode normally has a much
smaller solidity factor (as it sweeps a large area that moves up in altitude
during the power stroke) compared to a kite with the same rated power in
drag mode – the same kites in lift and drag modes have power curves
9
This is obviously as a result of the assumption made in Section 4. In reality, it
is expected that the induction factor to be also dependent on e. Discovering this
correlation is, however, beyond the scope of the present paper and is deferred to
a future publication.
M. Kheiri et al. Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13

associated with different values of σ . Furthermore, the average induction
factor for a lift-mode kite may be reduced when taking into account that
no wind power is extracted during the retraction phase of a pumping
cycle. Moreover, the on-board turbines in the drag-mode power gener-
ation normally see a lower relative flow velocity due to their induction
effects in the crosswind direction. These effects were not, however, taken
into account in the present theory, which could result in a slight over-
estimation of drag-mode power.
8. Concluding remarks
In this paper, the actuator disc theory was extended and applied to
crosswind kite power systems in the straight downwind configuration. In
contrast to previous studies in which the effect of kite on slowing the
incoming flow has been neglected, in the present paper, the effect of axial
induction factor was considered. Expressions for induction factor and
potential power output in both lift and drag modes were obtained. It was
shown that for the induction factor of zero, the present expressions for
power are reduced to those obtained by Loyd (1980). CFD simulations
were performed for a small- and a large-scale kite power system to
corroborate the theory. The agreement between theoretical and CFD
results was found to be fairly good. Moreover, using the theoretical
model, some numerical results were obtained. The results showed that
the extracted power is strongly dependent on the solidity factor of the
kite. It was shown that the maximum lift mode power occurs when the
reel-out speed is one third of freestream velocity (e ¼ 1=3),
independently from the solidity factor. In contrast, the maximum drag
mode power occurs at different values of the ratio of on-board turbines
thrust to kite drag, κ, as the solidity factor is varied. It was also shown that
the induction factor for a crosswind kite system is not always negligible
in contrast to what is commonly assumed. Moreover, it was found that
neglecting a small induction factor may result in a significant over-
estimation of power. It is noted that the results presented in this paper are
valid strictly speaking for CKPSs in the straight downwind configuration.
Further research is planned to extend the present theory to the more
realistic configuration of the CKPS where a non-zero tether elevation
angle results in inclined trajectories. In addition, some CFD studies are
already underway using URANS (with the k  ω SST turbulence model)
and LES flow solvers.
Acknowledgments
This research was conducted at New Leaf Management Ltd. The au-
thors are grateful to Prof. Marius Paraschivoiu of Concordia University
for providing them with the CFD computational facilities and Prof.
Curran Crawford of the University of Victoria for his useful comments on
the earlier version of the paper presented at the ISWTP 2017 in Montr
eal.
Prof. M. Mahdi Salehi of Sharif University of Technology is also appre-
ciated for useful discussions on CFD simulations. The first author is also
grateful to the Faculty of Engineering and Computer Science of Con-
cordia University for a start-up research grant.

Appendix A. Bernoulli's equation for flow in contact with a moving actuator disc

Referring to Fig. 14, the top sub-figure shows a moving actuator disc at time t with an ‘inlet’ streamline extending between points ① and ② and an
‘outlet’ streamline from point ③ to ④. The bottom sub-figure shows the same system at time t þ Δt. Points ① and ④ are located, respectively, suf-
ficiently far upstream and downstream of the disc, where the (absolute) flow velocities are, respectively, v∞ and vo . Points ② and ③ are located just in
front and back of the actuator disc and are moving with the disc at its speed, vd ; the (absolute) flow velocity at these points is denoted by v. It is assumed
that the velocity of the flow changes linearly between points ① and ② and between points ③ and ④.10 Two different types of coordinate system are
used: (a) fixed/inertial coordinate systems represented by Xi ; i ¼ 1; 2, and (b) moving coordinate systems represented by xi ; i ¼ 1; 2; the indices 1 and 2
refer, respectively, to the points on the left- and right-hand sides of the disc.
Applying unsteady Bernoulli's equation in inertial frame to the streamline between points ① and ②, and that between points ③ and ④ leads to (refer
to White, 2008):
Z  
2
∂v 1  
ρ ds þ pd  p∞ þ ρ v2  v2∞ ¼ 0; (A.1)
1 ∂t 2

Z  
4
∂v 1  
ρ ds þ po  pdþ þ ρ v2o  v2 ¼ 0; (A.2)
3 ∂t 2

where pd and pdþ are static pressures at points ② and ③, respectively.
One may find the flow velocities at arbitrary points A and B, shown in Fig. 14, as
 
v  v∞
vf 1 ðx1 ; tÞ ¼ v∞ þ x1 ; 0
x1
ℓ1 (A.3)
ℓ1
 
vo  v
vf 2 ðx2 ; tÞ ¼ vo þ x2 ; ℓ2
x2
0: (A.4)
ℓ2
From equations (A.3) and (A.4), one may find the time derivative of the flow velocity as
 
∂vf 1 v  v∞ ∂x1
¼ ; (A.5)
∂t ℓ1 ∂t
 
∂vf 2 vo  v ∂x2
¼ : (A.6)
∂t ℓ2 ∂t

10
This is consistent with the general actuator disc theory; for more details, refer to Sørensen (2016).

11
M. Kheiri et al. Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13

Using Fig. 14, ð∂x1 =∂tÞ and ð∂x2 =∂tÞ may be obtained as follows

∂x1 x1 ðt þ ΔtÞ  x1 ðtÞ Δs ∂s

¼ lim ¼  lim ¼  ¼ vd ; (A.7)
∂t Δt→0 Δt Δt→0 Δt ∂t

∂x2 x2 ðt þ ΔtÞ  x2 ðtÞ Δs ∂s

¼ lim ¼  lim ¼  ¼ vd : (A.8)
∂t Δt→0 Δt Δt→0 Δt ∂t
Equations (A.5) and (A.6) may then be re-written as
 
∂vf 1 v  v∞
¼ vd ; (A.9)
∂t ℓ1
 
∂vf 2 vo  v
¼ vd : (A.10)
∂t ℓ2
Substituting expressions (A.9) and (A.10) in equations (A.1) and (A.2), respectively, and performing integrations yields

1  
ρvd ðv  v∞ Þ þ pd  p∞ þ ρ v2  v2∞ ¼ 0; (A.11)
2

1  
ρvd ðvo  vÞ þ po  pdþ þ ρ v2o  v2 ¼ 0; (A.12)
2

which are simplified further to

1 1
p∞ þ ρðv∞  vd Þ2 ¼ pd þ ρv2r ; (A.13)
2 2

1 1
pdþ þ ρv2r ¼ po þ ρðvo  vd Þ2 ; (A.14)
2 2

where vr ¼ v  vd is the relative flow velocity at the actuator disc.

Appendix B. Derivation of the equation for the optimal value of the ratio of on-board turbines thrust to kite drag

From equation (28), one can obtain the drag mode induction factor as a function of C ¼ ð1=4ÞðAk =As ÞCL ðCL =CD Þ2 and κ, as below

C
a¼ : (B.1)
ð1 þ κÞ2 þ C
Thus, equation (29) may be re-written as

   2 !3
1 CL ð1 þ κÞ2 κ
PD ¼ ρAk v∞ CL
3
: (B.2)
2 CD ð1 þ κÞ2 þ C ð1 þ κÞ3

For given values of ðAk =As Þ and CL ðCL =CD Þ2 , and by letting ð∂PD =∂κÞ to zero, the local maximum or minimum points for the drag mode power are
obtained:
   2
∂ PD 1 C
  
¼ ρAk v3∞ CL L ð1 þ κ þ 3κÞ ð1 þ κÞ2 þ C  6κð1 þ κÞ2 ¼ 0; (B.3)
∂κ 2 CD

which may be simplified to

2κ3  3κ 2 þ 4Cκ þ C þ 1 ¼ 0; (B.4)

that is equation (30), given in Section 5, and it should be solved numerically (for example, by using the roots function of MATLAB).
Fig. 15 shows variation of optimal values of κ as a function of C.

References
Ahrens, U., Diehl, M., Schmehl, R. (Eds.), 2013. Airborne Wind Energy. Springer, Berlin.
Ansys Inc, 2009. ANSYS FLUENT 12.0 User's Guide. Ansys Inc, Canonsburg (PA).
Archer, C., 2013. An introduction to meteorology for airborne wind energy. In:
Ahrens, U., Diehl, M., Schmehl, R. (Eds.), Airborne Wind Energy. Springer, Berlin
Heidelberg, pp. 81–94.
Argatov, I., Rautakorpi, P., Silvennoinen, R., 2009. Estimation of the mechanical energy
output of the kite wind generator. Renew. Energy 34, 1525–1532.
Argatov, I., Rautakorpi, P., Silvennoinen, R., 2011. Apparent wind load effects on the
tether of a kite wind generator. J. Wind Eng. Ind. Aerod. 99, 1079–1088.
Avila, M., Folch, A., Houzeaux, G., Eguzkitza, B., Prieto, L., Cabez

on, D., 2013. A parallel
CFD model for wind farms. Procedia Comput. Sci. 18, 2157–2166.
Barthelmie, R.J., Rathmann, O., Frandsen, S.T., Hansen, K.S., Politis, E.,
Prospathopoulos, J., Rados, K., Cabez
on, D., Schlez, W., Phillips, J., Neubert, A.,

12
M. Kheiri et al.
Schepers, J.G., van der Pijl, S.P., 2007. Modelling and measurements of wakes in
large wind farms. J. Phys. Conf. 75 (1), 012049.
Bazilevs, Y., Hsu, M.C., Akkerman, I., Wright, S., Takizawa, K., Henicke, B., Spielman, T.,
Tezduyar, T., 2011. 3D simulation of wind turbine rotors at full scale. Part I:
geometry modeling and aerodynamics. Int. J. Numer. Meth. Fluid. 65, 207–235.
Bourgault, F., Todd, D., Beatch, J., Kheiri, M., Damron, L., Saberi Nasrabad, V., 2017.
Efficient and power smoothing drive-train concept for pumping kite generators using
hydraulics. In: Book of Abstracts of Airborne Wind Energy Conference (AWEC) 2017,
Germany, October 5-6.
Bruinzeel, L., Klop, E., Brenninkmeijer, A., Bosch, J., 2018. Ecological impact of airborne
wind energy technology: current state of knowledge and future research agenda. In:
Schmehl, R. (Ed.), Airborne Wind Energy: Advances in Technology Development and
Research. Springer, Singapore, pp. 679–701.
Burton, T., Jenkins, N., Sharpe, D., Bossanyi, E., 2011. Wind Energy Handbook. John
Wiley & Sons.
Cherubini, A., Papini, A., Vertechy, R., Fontana, M., 2015. Airborne wind energy systems:
a review of the technologies. Renew. Sustain. Energy Rev. 51, 1461–1476.
Costello, S., Costello, C., François, G., Bonvin, D., 2015. Analysis of the maximum
efficiency of kite-power systems. J. Renew. Sustain. Energy 7, 053108.
Fagiano, L., Milanese, M., 2012. Airborne wind energy: an overview. In: American
Control Conference (ACC), Montr
eal, QC, Canada. IEEE, pp. 3132–3143.
Johansen, J., Sørensen, N.N., 2004. Aerofoil characteristics from 3D CFD rotor
computations. Wind Energy 7 (4), 283–294.
Kheiri, M., Saberi Nasrabad, V., Victor, S., Bourgault, S., 2017. A wake model for
crosswind kite systems. In: Book of Abstracts of Airborne Wind Energy Conference
(AWEC) 2017, Germany, October 5-6.
Leuthold, R., Gros, S., Diehl, M., 2017. Induction in optimal control of multiple-kite
airborne wind energy systems. IFAC PapersOnLine 50–51, 153–158.
Loyd, M.L., 1980. Crosswind kite power. J. Energy 4 (3), 106–111. Article No. 80-4075.
13
Journal of Wind Engineering & Industrial Aerodynamics 181 (2018) 1–13
Lunney, E., Ban, M., Duic, N., Foley, A., 2017. A state-of-the-art review and feasibility
analysis of high altitude wind power in Northern Ireland. Renew. Sustain. Energy
Rev. 68, 899–911.
Magnusson, M., Rados, K.G., Voutsinas, S.G., 1996. A study of the flow downstream of a
wind turbine using measurements and simulations. Wind Eng. 20 (6), 389–403.
Okulov, V., van Kuik, G.A., 2009. The Betz-Joukowsky limit for the maximum power
coefficient of wind turbines. Int. Sci. J. Alternative Energy Ecol. 9, 106–111.
Rhie, C.M., Chow, W.L., 1983. Numerical study of the turbulent flow past an airfoil with
trailing edge separation. AIAA J. 21 (11), 1525–1532.
Schepers, J.G., 2003. ENDOW: Validation and Improvement of ECN's Wake Model.
Energy research Centre of the Netherlands ECN.
Schmehl, R. (Ed.), 2018. Airborne Wind Energy: Advances in Technology Development
and Research. Springer, Singapore.
Silverstein, A., 1935. Scale Effect on Clark Y Airfoil Characteristics from NACA Full-scale
Wind-tunnel Tests. NACA Report No. 502.
Sørensen, J.N., 2016. In: Peinke, J. (Ed.), General Momentum Theory for Horizontal Axis
Wind Turbines. Research Topics in Wind Energy, vol.4. Springer, Switzerland.
Thumthae, C., Chitsomboon, T., 2009. Optimal angle of attack for untwisted blade wind
turbine. Renew. Energy 34 (5), 1279–1284.
White, F.M., 2008. Fluid Mechanics, sixth ed. McGraw-Hill, New York.
Wilson, R.E., Lissaman, P., 1974. Applied Aerodynamics of Wind Power Machines. NASA
STI/Recon Technical Report N, 75.
Wolfe, W.P., Ochs, S.S., 1997. CFD calculations of S809 aerodynamic characteristics.
AIAA-97-0973. In: 35th Aerospace Sciences Meeting and Exhibit, Reno, NV, USA.
Zanon, M., Gros, S., Meyers, J., Diehl, M., 2014. Airborne wind energy: airfoil-airmass
interaction. In: The 19th International Federation of Automatic Control (IFAC) World
Congress, South Africa, August 24-29.
Near Zero, 2012. In: Inman, M. (Ed.), Energy High in the Sky: Expert Perspectives on
Airborne Wind Energy Systems (White Paper) available at: http://nearzero.org/.