Offshore Comparison : Makani vs HAWT

Wind :
The Offshore wind data (ref a) reveals only a small wind speed difference between 100m (typical HAWT hub height) and 300m altitude (typical AWE height), i.e. 0.5m/s -> 1.0 m/s. The power law coefficient (friction exponent) is between ~0.03 and ~0.1, as opposed to ~0.14 Onshore (ref b). Offshore is a low wind shear environment.

Other points of interest (ref a), at higher altitudes :
wind veer increases
turbulence decreases
wind speed variability increases (opposite to AWE assumption)


The Scaled Up Makani M600 (from the prototype M30 Wing 7), has a more erratic flight speed profile (ref videos).
The power vs time plot for Makani (ref b) will become more erratic, and more inefficient, further reducing the average power output over the cycle (< ~50% i.e. ~40% used here).

The M600 has a maximum power output of 600kW, with a ~40% efficiency at the generator, the ‘rated power’ output will be ~250kW.

The M600 wing span is ~26m, the aspect ratio is ~18, giving a wing area of ~38m² (ref e).
A propeller blade is ~1m long, with an area of ~0.15m².
Mean altitude is ~255m (equivalent to hub height), with a ~250m diameter sweep, and ground clearance of ~130m.
Tether length is ~440m, giving a mean elevation of ~35°.

To match a conventional wind turbine (HAWT) of 10MW ‘rated power’ output, ~40 M600 units are needed.
AWE total :
generator size => ~24MW
wing area => 1520m².
blade area => 240m² (~1600 blades).
swept area => 730000m² (doughnut/annulus only).

A 10MW HAWT (Vestas V164) has a rotor blade length of 80m, and a blade area of ~280m².
HAWT total :
generator size => ~10MW
blade(wing) area => 840m² (3 blades).
swept area => 21000m² (disc), 164m diameter.

Comparison :

Low wind shear will result in roughly the same ‘Capacity Factor’ for AWE and HAWT at the same site. Increasing the tether length will not improve power output (tether drag dominating).

The M600 extracts very little energy from the wind it intersects (sweeps). The ‘power coefficient’ (extraction efficiency) is ~3%, compared to ~30% for HAWT (the Betz Limit is ~59%). The diameter of the supersize doughnut forces a high tether elevation of 30° -> 40°, and cosine loss of 35% -> 55% (ref b). Therefore HAWT has more fuel (i.e. a higher wind power input), and can downsize the blades for the equivalent power output.

The average ‘power harvesting factor’ (zeta) for the M600 is ~7 (and for HAWT also). The M600 could improve the average zeta by reducing the cosine loss, but would need a tower.

The M600 has a relatively low ‘rated wind speed’ of 11.5m/s, compared to a typical HAWT (~13+ m/s). If this can be increased, it will allow a correspondingly smaller wing to be used.

LandSpace (SeaSpace) for HAWT (with 60 rotor diameter²) is ~6MW/km², and for M600 (with tether length radius) ~0.4MW/km².

In summary, Makani need a bigger wing, a bigger generator, a much bigger swept area (higher altitude), more SeaSpace and a system to smooth the erratic power generated, in order to produce the same ‘rated power’ output as HAWT.

Additionally, most of the Offshore cost is for the installation, cables, and grid connection (CAPEX), along with the OPEX (maintenance costs), and not for the kite (ref c). This favours big MW units, so lots of M600 units will tend to increase both, and give a higher LCOE.

Finally, HAWT Scales with bigger and heavier blades, yet produces a lower LCOE, keeping the ‘square-cube law’ in check. Blades (wings) are not solid, whereas the tether is. AWE limits are set by the tether (ref d & ref e & video), with exponential growth in tether mass (kg) when scaling, reduced aerodynamic efficiency of the kite due to tether drag, and ‘rated power’ output constrained by the maximum tether tension. A comparison with other AWE solutions such as Ampyx or SkySails, would give a similar outcome.

Conclusion :

Wind shear is greater Onshore, where ‘Capacity factor’ and ‘Annual Energy Production’ can clearly advantage AWE.

Onshore CAPEX is mainly the kite cost, and not the infrastructure, so small units make sense.

Utility Scale AWE is ‘Scale-Out’ (Solar) and not ‘Scale-Up’ (HAWT).

References :

a) Understanding of the Offshore Wind Resource up to High Altitudes <= 315m (2019)

b) Springer AWE Book, R. Schmehl et al (2014)

c) Makani keynote presentation AWEC 2017

d) Economic assessment of small-scale kite wind generators, Ivan Argatov, Valentin Shafranov (2014)

e) Makani Response to the Federal Aviation Authority

Video :

Makani Power, 2012 Testing Program

Testing Makani’s M600 energy kite in Spring 2017


It is a relevant analysis. I add it would be the same for the onshore comparison. And Makani’s is seen as one of the most promising AWES.
I think one purpose for AWES should be to avoid the cantilever effect of HAWT, but not necessarly to fly.

1 Like

Hi Derek,

Its been known since 2009 that Makani’s architectural down-select is a dead end technically, for reasons you cite, and many other critical flaws you don’t identify. The “value” of the program is to enhance the public image of Google’s search-engine advertising monopoly. Its a misleading baseline model for AWE.

You imply that the M600 has better potential onshore, but intractable safety and noise issues onshore drove the concept offshore, where all it ever has to do for Google PR purposes is an occasional short test, and if it crashes there will be minimal blowback. The Shell offshore partnership is synergistic PR for both giants.

A more realistic baseline unit-model on- or offshore is SkySails proposed 1000m2 power kite. A few parafoils, like the MegaFly have flown at this scale, not for AWE, but these indicate no inherent scaling barrier for exploring AWES application at around 10MW rated-unit. Some of us have envisioned topologically networked kites with many tethers and interconnections aloft. You only presumed single-tether unit-topology (kitefarm “brush” topology) so far.

You make a fine start here in critical AWES analysis. Looking forward to your further efforts,


McConney has been mostly on an all-talk format for 12 years and counting.
With their seemingly-unlimited financial backing, if they had a true economical energy solution, don’t you think you’d see one running by now, somewhere?

It is obvious. That doesn’t work. That will never work.

The Old AWES Forum archives preserve what DougS and PierreB imagined about Makani a decade ago, when the the M600 architecture was announced, and ever since.

To form their loosely changing opinions, they don’t reason from the sort of diligent engineering comparison Derek is undertaking.

The comparison os good. I agree to your analysis. I wont comment on Makanis proposal in particular, but I would like to add some argumentation pro-AWE

  • operational expenses could be lower for kites as they are more easily brought ashore for maintenance while replaced with a similar wing. Also the need for huge cranes is much less we should assume.
  • AWE is still in its infancy. It is unfair to compare a 600 kW unit with a 10 MW unit. You must compare each system at a similar scale looking at how far the technonoly couls easily scale. HAWTs are probably close to their maximum size right now (save a last few doubling of radius) while AWE should be expected to scale a lot more than 600 kW if successful. The tether length and operational height should scale with wingspan and thus give AWE a further advantage with regards to the wind profile
  • even though AWE energy extraction per km2 ground area is smaller, the difference may be offset somewhat by higher altitude for AWE at a comparable scale.

Theres more. Its not clear cut, though your analysis does strongly suggest that Makani has a lot of work cut out for itself

No, I just wrote:

Compared to DaveS’ schemes (of which the Böse-Einstein-Flapping-Oscillating-Quantum), Makani’s method is far better as it generates some electricity. But such a comparison doesn’t make sense.


I never saw Makani as “one of the most promising”. That’s why I left Alameda Islandin 2007! KiteShip actually assigned me to make a direct early determination of Makani thinking. Anyone who ever thought them promising was badly misled.

The TRL9 COTS ship-kite power-kite both conforms to advanced physics interpretations and is a far better engineering “scheme” than Makani, with their lesser apparent grasp of critical physics, like aviation Scaling Laws; and AWE IP like USP3987987.

This comparison “doesn’t make sense” to you yet, but it will if you master the applicable domain knowledge, or just wait for AWE to mature around 2030.

One more thing.

A giant windmill with the hub at 100 m and the blades at 80 meter length, has an average swept area height of 100. But the average power at the blade is less than the power at the hub because wind power is \propto w^3.

Because the gradient is so steep at lower altitudes (eg 20-50 meters above sea) this gives a further disadvantage to the HAWT windmills. This difference in wind speed might be 7.5 m/s at 30 m vs 9 m/s at 150 m. Power difference +72%. AWE will probably not fly at these low altitudes due to safety concerns, and the fact that is is not the most efficient.

AWE at the correct scale for comparison with a 10 MW windmill might be 120 m wingspan and then 1-1.5 km altitude on average.

The numbers would be more favorable for AWE if you used such numbers

Derek in part reasons from the fact that the offshore wind gradient is less pronounced than onshore, with its higher surface friction. This is usually true, but large waves, including from non- wind aligned directions are a wild card in surface effects, maybe even stalling kiteplanes at landing on “perches”.

Tallak is right to suggest that the wind gradient at sea is nevertheless still a key factor, which is true insofar as wind power increases at the cube of velocity, so even a diminished gradient still counts for a lot.

The first order offshore reality is “the sea is a harsh mistress”, an extreme environment where industrial operations are far more expensive and prone to accelerated or total loss.

1 Like

TU Delft asks the question (section 4.5, ref f) :
Why is the YoYo ‘Traction Phase’ cycle time longer than expected, and the power output less than expected (soft kite 14m² -> 25m²).

The M600 has the same behaviour (ref video).

The ‘law of conservation of energy’ describes the interchange of Potential Energy (PE) and Kinetic Energy (KE). The kite starts at the top of the loop (circle or ‘figure 8’) with gravitational PE, and on descent it is converted into KE. The kite has to depower to avoid a tether break when the maximum tether tension is reached. Reducing the lift (or increasing the reel-out speed for YoYo), slows the kite down, and KE is lost. On the climb back to the top, there is not enough KE, so the kite slows further, with only the aerodynamic lift to do the work.

The cycle time is stretched out, restoring the PE, while the power output becomes more erratic (negative for the M600), decreasing the average power output over the cycle.

‘Scaling Up’ just makes things worse, more depower on descent, followed by a harder longer climb (with an exponentially increasing tether mass (kg)), see M30 -> M600 videos.

References :

f) Quasi-Steady Model of a Pumping Kite Power System, Rolf van der Vlugt et al (2018)


And also the irregular wind power because the swept area is not quite perpendicular to the wind direction, as mentioned just before the Detailed Analysis on

and leading to variable cosine and power according to the position of the kite within the wind window. Power variations also add losses to losses.

And also the power/space use ratio for crosswind AWES is very low in spite of a very high power/kite area ratio (at least before a re-evaluation).

As a result some AWES seen as less efficient, being more or less stationary, could better maximize the space used.

You cant use conservation of energy because the wind will add energy as you go, and the amount of power depends on many factors…

Let me add an explanation: when the kite slows down dt gravity, it gets «more powerful» as a result… there is energy going everywhere.

As the kite moves some kinetic energy occurs, whether going up or down. Some similarity can work with Solar balloon jumping when the balloon rises before lifting a person by its kinetic energy.

The potential energy has to fully do with kinetic energy only when the gravity fully occurs.

Aerodynamic forces are higher compared to gravity. And there is a significant difference between the 1 ton M600 and 20 or 30 kg TU Delft wings, leading to a huge difference of kinetic energy, taking also account of M600 speed being the double (so the energy is squared) _ by aerodynamic force, not by gravity _ of that of TD wing.

In the paper in reference there are some measures with gravity incl. and excl. and they are right.

Let me add another thing - mass also evens out speed through the loop. A glider can glide pretty far without stopping. Though mass makes gravity slowdown more prominent, it does make cosine losses less prominent.

I’m going to ask the dafties questions…
Why does more mass reduce the significance cosine losses? Some sort of momentum punch through?
Is it only relevant for electrical drag mode? Or can mechanical drag mode use it too?

How does it play through the full cycle?
And sorry if so but…
Where should I have read the answers before?

Doh… Think I just sussed it out… A less steep circle to fly with a higher line elevation keeps the speed more constant. Is that right? But surely cosine losses are still there…

More mass does not reduce cosine losses, but rather the effect change of cosine losses has on speed.

Since the speed of a kite is a steady equilibrium, any external force acting as drag or thrust is counteracted by an opposite force once the speed is different from the equilibrium speed. This is a reason why simple energy conservation calculations dont work well for AWE. Another reason is simply that for most architectures, (I assume) energy losses to drag are significant.

If we still to an energy calculation if a kite, say a 10 meter wingspan kite looping at 100 m diameter weighing 100 kg, and with a speed at the bottom turn of 30 m/s, the energy at the bittom turn is E=\frac{1}{2} m v^2, 45 kJ. The energy lost going from the botton turn to the top turn is E=m g h, 98 kJ. So for this example the kinetic energy would be completely lost in the loop.

Even so, there would be some kinetic energy left in the kite during the critical «upwards point». Given just a steady state solution, the kite might have stalled, but the dynamic solution could show that the kite pulls through this point fine.

Changing the elevation angle does help a bit. At 30 degrees, the example above would have resulted on the potential energy of 84 kJ, and the speed at the bottom turn would be even larger compared to looping at zero elevation (incidentally there is less cosine loss with a little elevation, at the bottom turn).

The power generation of course will experience huge variations in ouput power as the kite needs to use more power to maintain its own speed, while at the same time not being able to maintain the optimal energy extraction speed.

Now on to variations due to cosine losses: If we are at an elevation angle of 30 degree looping from zero degree elevation to 60 degree, the change in speed will be 1:2. Like before, we look at the kinetic energy at bottom and top turn. The speed at bottom turn is still 30 m/s, the speed at top turn is 15 m/s, looking at the steady state solution. The dynamic solution would show the two speed being closer in value due to momentum punch through. This is what I meant by mass reducing cosine loss variations.

Anyways, energy considerations are in there, but its not enough to draw any conclusions. From my point of view, simulations seem the best tool to analyze the dynamic speed if the kite theough a loop. I’d love if anyone came up with some good rule if thumb for these things.

A bit of trivia: when the kite is travelling at the upwards point, you cannot easily generate a lift force pointing to counteract gravity. If you tried this, you would also be changing the angle of attack of the kite. So aerodynamic forces and momentum punch through is the only way to keep a (non networked) kite moving past this point…

Btw: sorry if all if this doesn’t make sense, It’s not easy to express all of this in a few paragraphs

1 Like

Thanks @tallakt
Yes whilst flying around 1 plane of rotation the changes in line angle, AoA etc are great.
Yes it is usually easier to interpret a set of graphs and animations for data this interdependent… Being able to move around the graphs and animations in 3d with control of parameter sliders really helps too.
Would be great to develop a more intuitively understandable representation of the dynamic.
Thanks again