On September 10, 2020, Makani released a significant amount of material detailing their development in a blog post by Paula Echeverri, former Chief Engineer and Flight Testing Lead.
The release contains numerous videos from flights, 1000+ pages of technical documentation, software source codes of firmware and flight simulators, raw flight data, and a documentary as well as a non-assertion pledge for all their patents. This is a treasure chest for us at Kitekraft , a fellow airborne wind energy developer that is following a similar technical concept. To release all this data after shutting down the company to help advance the airborne wind energy technology is incredibly generous and a testament to the mission-driven people at Makani. We already had a chance to dive into Makani’s materials, providing us with incredible insights. It helped us confirm many assumptions we had and it will likely contribute to save us years of development time. We at Kitekraft are very grateful to everyone at Makani for pushing the whole technology to the point where it is now and showing what is possible! As Paula wrote — “innovation is a relay race” — we agree and take this as a challenge to boldly push the next generation of airborne wind energy technology forward.
In this blog post, we outline some of the challenges faced by Makani and explain if those can be — or are already — solved by Kitekraft's approach.
Roll Excursion During Hovering
Makani’s kite was susceptible to “roll excursion” in hovering. This was likely the most severe problem of the concept and led to several crashes or almost-crashes of small and large kites. Here are links to several Makani-videos:
Makani used a Y-bridle tether connection at the kite which in principle should give a restoring moment for any roll excursion, both in crosswind flight as well as in hover. However, this requires that the tether is almost horizontal at the kite meaning that the kite’s hover altitude is limited, and that sufficient tether tension is maintained. The roll angle was not well damped and the concept apparently worked only under some conditions. This “existential problem” was not resolved until the end as concluded by Makani in the technical reports p. 33 and following.
In contrast, Kitekraft 's kite has a large roll control authority, not only in crosswind flight due to a single tether connection point (further discussion below), but also in hover mode, due to the different airframe design and rotor placements: The rotors are in front of the biplane/boxplane wings close to the wing tips. During hover the flaperons of the wings are in the propeller downwash. Deflecting the flaperons is therefore similar to thrust vectoring. Since they are far from the center of mass, it offers to actuate large roll moments so that we can do both:
Stable flight in crosswind mode: Kitekraft kite flying figure-eights, after a hover launch.
Stable flight in hover mode. Kitekraft kite in an early test in the machine hall performing hovering on a single tether connection while being heavily disturbed, in particular in roll (hover-yaw). We also flew the kite in hover outside for more than an hour in a row and sometimes with strong side winds without any problems.
Our flight controller can easily stabilize our kite’s roll angle in hover using the flaperons. This has been working well for more than a year now with our prototype and we expect an equally well behavior when scaling to larger kites.
Note that we cannot use the moment generated by using differential propeller rotation directions as in small multicopter-drones, as this offers not nearly enough control authority, which was similarly concluded in Makani’s reports.
Note also that strong adverse effects of our configuration are not expected. Maybe even on the contrary: Makani’s rotors are placed in front of the empennage (tail with rudder and elevator). As the rotors extract energy, the empennage’s control surfaces see a lower airspeed reducing their effectiveness. This requires extra large control surfaces (further details are in Makani’s reports). In contrast, our kite slows down airspeed at the wingtips not affecting the empennage. It does affect the main wings, but an elliptical lift distribution is desired to minimize induced drag and the rotors may rotate so to counteract the vortices generated at the wing tips. By that, our kite can achieve an efficient lift distribution even with a rectangular planform wing without taper or washout simplifying airframe design and manufacturing.
Makani’s kite was susceptible to aeroelastic oscillations. Here are two videos showing resonance effects:
Had this design been scaled up to multi-MW kites, most likely it would have gotten worse. Susceptibility to flutter was also in part a reason for the low efficiency of Makani’s M600 prototype according to the technical reports (further discussion below).
In contrast, our biplane/boxplane is a truss-like structure and therefore exceptionally stiff. Even at larger-scale, we do not expect such challenges. As another advantage, a biplane/boxplane kite of the same wingspan of a Makani monoplane kite has about 2x the power as the wing area is roughly doubled, making our system much more compact for reaching a certain power level and thus further decreases the susceptibility to aeroelastic issues. Again, our CFD simulations and wind tunnel tests show no evidence of significant adverse effects of our biplane/boxplane configuration.
Makani aimed to build a MW-scale system for offshore grid injection right away. While incredibly bold, building a new technology in large-scale right away is hard to accomplish: Also see the story of Growian.
As outlined in an earlier blog post, Kitekraft ’s approach is to rapidly learn at small-scale and low costs, commercialize fast with the smallest economically viable system, e.g. in microgrids, and then scale up with a proven design, which is, as we believe, the fastest way to get a commercial MW-scale system. As dozens of systems are then deployed at real customer sites early on, we will gather customer validation and years or even decades of accumulated operational flight data quickly to prove the reliability and commercial viability. This is essential to give customers enough confidence to make larger investments in flying wind turbine systems.
Power Generation Efficiency
In the technical report p. 26, 59 as well as 253 and following, further challenges of Makani’s kites are explained. It provides a justification for the low power generation efficiency. While much of it can be solved with design iterations like a better optimized airfoil (“Oktoberkite”), some issues are deeper. Here again, the Y-bridle tether connection is mentioned as a primary cause, as it limits the control authority for flying tighter circles. This led to numerous further effects, e.g. the requirement to fly slower, reducing efficiency. It should be noted however, that Makani’s simulation models predicted the measured worse performance with the as-built kite, such that design changes can be expected to achieve the much better specified power curve in reality.
Kitekraft’s kite does not have a Y-bridle tether connection. In the crosswind figure-eight flight shown in the video above, the kite flew stable with rather high tether roll angles of up to ±45°. As our kite is a biplane/boxplane which can handle high loads with its truss-like structure, a Y-bridle tether connection is not required and a single point connection can be used. So there are fewer and weaker cross-couplings in the design and a similar performance problem as occurred to Makani may therefore be avoided easier at Kitekraft .
It should be further noted that a single tether connection allows flying figure-eights in contrast to a fixed Y-bridle. It would also allow flying circles. The former has, however, advantages in high winds when power generation is saturated, which was another challenge detailed by Makani here p. 185 and following. In our research paper cited below, we motivate to fly figure-eights with downward-stroke in the middle to solve this task:
Florian Bauer, Daniel Petzold, Ralph M. Kennel, Filippo Campagnolo, and Roland Schmehl. “Control of a Drag Power Kite over the Entire Wind Speed Range”. AIAA Journal of Guidance, Control, and Dynamics. July 3, 2019. DOI: 10.2514/1.G004207. URL: https://arc.aiaa.org/doi/10.2514/1.G004207. Download: Preprint-PDF
As Makani’s kites flew circles, the tether must be untwisted by a motor all the time. As power and communication signals must be transmitted, a slip ring is required. The motor and the slip ring are components that can fail and did so: https://youtu.be/YmSYaQu7Xk4
In addition, it is actually easier to fly figure-eights for the control algorithm and has some further advantages as outlined above, although this was not available for Makani with the choice of a fixed Y-bridle tether connection. The challenges stated above and the required solutions for the Makani kite add further complexity in the system as well as in the engineering- and design phase.
Obviously, a Kitekraft system does not need components to untwist the tether as we fly figure-eights. We generally strive for simplicity and try to minimize the amount of sensors and actuators required (while maintaining redundancies).
Simulation and Sizing Models
Makani used a more traditional approach to sizing the kite using, e.g., linearized models and conventional software packages like ASWING, as detailed in the technical documents, e.g. here, p. 7 and following. This required several manual iterative loops by engineers and in the end did not cover all effects unique to energy kites.
At Kitekraft we implemented a systems engineering framework in C++ which covers all major effects including, e.g., tether moments on the kite or effects of airspeed increases or decreases on airfoils and control surfaces caused by rotors. The aerodynamics model and solving algorithm is so efficient that it can be executed (and was so during hoverings and our figure-eight flight) in real-time thousands of times per second on the flight controller to optimize the control deflections without any linearizations around some trim point (which does not exist for kites, as also concluded by Makani). The systems engineering framework is not only used to validate flight control algorithms, but also for sizing the kite with a first principles-based approach. This enables us to design optimal kites very fast. Our figure-eight flight validated most of our models, control algorithms, and general approaches.
Challenges are expected to emerge also in our further development. Confirming many of our assumptions and solutions with the information provided by Makani though, gives us validation and confidence in our approach. This adds to our recent success of flying our rigid kite prototype autonomously in figure-eights with 800V tether power transmission after just over a year since company foundation, essentially accomplished with just the four engineers of the Kitekraft founding team and help by students, interns, and friends.
Our developments have been fueled by the prior research and work of the broader airborne wind energy community, including in particular Makani who boldly pushed the boundaries by even flying a 737-wingspan kite offshore. Makani’s release is especially for us at Kitekraft a massive resource in many more aspects than outlined in this blog post — we believe to have good solutions on even further challenges outlined by the team of Makani or evident in the videos, and for sure we can learn and utilize solutions discovered by Makani on other aspects.
This is an exciting time for developing such a new technology, which, we believe, is an essential solution to help fighting climate change!