So, welcome everyone to my presentation about automated power cycles in day-long operation at SkyCells test site. My name is Manfred Quack. I'm working at SkyCells Power, and the work I'm presenting today is joint work

with Mahmoud Soliman and Rafal Noga. So, let me quickly give you an introduction or an overview of today's talk. In the introduction, I will shortly talk about the evolution of automated flight at SkyCells and provide the motivation

and goals for this talk. Then I'll go quickly through the system components and the LIDAR measurements that we have available at the test site. And as an introductionary question, I ask the question how is wind shear affecting airborne wind energy systems in general?

Then to get more specific about the SkyCells system, I will go through the equations of motion, target point control, and talk then about the encountered wind profiles during operation at the SkyCells power system test site. And then I will actually talk about how wind shear is affecting the operation

of our system specifically, and in the sense of an outlook, talk about how one could further exploit the wind shear. So, SkyCells has quite some history. We started in 2001 with using kites for propulsion,

large-scale ship propulsion, and we then also started working on the airborne wind energy systems for energy production, electric energy production, and we had our power functional

model on the scale of 20 square meters in 2011. And in January 2018, we started with the SkyPower 100 project, together with our partners, which is a research project funded by Germany.

And in 2019, I get my last presentation at the Airborne Wind Energy Conference, about extended periods of automated tetraflight at SkyCells. That was still at the smallest scale, so in the meantime, we actually scaled off the system, and we also,

as part of this project, we developed our test site in northern Germany, and at this site, we had our first flight using a 90-square-meter kite in 2019, and in 2020, we achieved day-long operation at that same site, and the data I'm going to present is

from this site, using also a 120-square-meter kite, and it's also from this period from 2020. So, looking at the timeline, there was actually one thing that one can notice, that there was an increased relevance

of the wind gradient, at least for us in SkyCells. One thing is, if you move from marine-based operation to what's land-based operation, obviously the wind-tier gradients that you encounter are actually stronger on land than at sea. So, that means you already have a larger gradient,

a stronger gradient, and because the power system is changing the tetra-length over the power cycle, you actually move through a larger range of operation altitude, and since we did an up-scaling, we also increased the maximum tetra-length,

which then further increased this operation altitude. So, for the small system, the operation altitude was around 200 meters, and for the system that we're working with currently, it's about 150 to 350 meters. So, all these things work towards increased relevance

of wind gradient, and then there is also the thing that when you fly over longer periods of time, these wind gradients will actually change with time, and that's something that you notice once you start doing day-long operation, and, well, last but not least,

you also need a lighter available to be able to present these results, and that's nice that we have at our test site, we were able to use a lighter measurement. So, then quickly, so, bringing that to questions I'm planning

to answer in this talk is how does wind-tier vary over day-long operation? I will present some lighter measurements. How does Geyser's flight control system react to duration in wind-tier? I give a short intro on the controlled variables, and then show some example of day-long flight data,

and then I go towards the outlook. Maybe I should mention that I will only focus on the power phase, just to keep this within a 10-minute talk, and also only focus on the variation of wind speed with altitude, but not on the variation of wind direction with altitude,

which is also an important topic, but it's not within the scope of this talk. So, quickly, going through the system components, just that everyone is on the same page. We have large-scale ram air kite. We have a control pod with a ram air turbine. We have a tether and a mast for launch and landing,

and a tow point that allows us to get line length and line angle measurements, and then finally we have the ground station with a winch drum, winch drivetrain, and everything is mounted on the tripod so that we can orient the main container actually

with the main wind direction using this mounting system. So, the lighter measurements that we have available outside, they have the following properties. So, we can measure these quantities, wind speed, wind direction, and vertical wing component, each at 12 stations between 40 and 290 meters.

There are some limitations when you go above 200 meters, and the update rate is around 1 hertz. We only use it for R&D and monitoring purposes. So, we don't feed the live stream to the fly control. The fly control algorithm doesn't directly use this information.

It's only used for R&D and monitoring purposes, and the key parameters of interest. So, on the right hand, you have some examples, and if you look at the black line, so that's just a 15-minute average, and then if you look at the blue line, that's the minimum, and the green line, that's the maximum over that 15-minute

and it's over that 15-minute period at the different stations over the altitude. Then there's one more point that you can compute from this data. If you just look at the difference between the maximum

and the minimum, you get some kind of a measure for the variability, and that also changes with altitude. So, we have, in this case, we have a much larger variability at the bottom than at higher levels. So, how does wind shear affect an urban wind energy system in general? I think that applies for all systems, not only our system.

So, it should be more or less known that the power is affected cubically by the local wind speed. Well, by the wind speed, but then if you talk about wind shear, by the local wind speed on that altitude, and also by the cosine losses. Cosine losses, they have been discussed in the first airborne wind energy book

in the introductory chapter, and basically it boils down to the point that you, the power is really scaling with this term. So, cubically in the wind speed, and also cubically with this cosine term of the elevation angle, hence cosine losses.

So, that means you have a tradeoff. Flying lower is better due to the cosine terms, and flying higher is better due to the wind gradient, because typically the wind is stronger further up. And, additionally, you have, there is another reason why you maybe want to fly higher, because there is a smaller variability in the wind speed

at higher altitude, but there is also a limitation. There's some minimum safe altitude that you probably don't want to go below, at least not below ground, and then there is some maximum altitude, which is just limited by your line length that you have available. Obviously, you cannot exceed that.

So, I will go very quickly through this. The main message about this is the controlled variables. If you want to know in more detail, then you can look at the recordings from the previous talks, or also in the literature in the Avond-Winn Energy book from 2018.

So, the equations of motion that we use allows to see the two control inputs that we use. One is the deflection at the control pot, the mark delta here, which we use to change the kite orientation. And that, through the coupling,

allows us to control the flight trajectory. And that's actually how we control, yeah, we use this to control the flight trajectory. And we have another control input, which is the wind speed, and that is actually controlled, used to control the airspeed. So, we separate that basically in two control problems.

And talking a bit more detail about the flight, flight control, the trajectories are controlled using target points. So, the idea of the target point control is that you specify in the wind window two target points

by their wind window position, elevation, and azimuth angles. And then the controller, the underlying control loops make sure that you fly towards the target point. But before you can actually reach the target point, you will reach, you will enter a trigger region. And once that's triggered,

it actually changes the active target point from the first one to the second one, and you will actually fly then towards the second target point. This is handled then by the underlying control loops. The effect of this is that just by specifying the location of these target points, you actually then get the trajectory

that evolves by itself, and you can still change the shape by making basically the width of this figure larger or by changing the altitude at the elevation of these target points. And it's important to note that the target point control does actually not make any assumption on wind profile

or the specific wind strength as such. So coming to the day-long flight test. So the goal, main goal was to demonstrate that the system can be operated at a high load for a long period, and we achieved to keep the system airborne and in automatic operation

for more than 40 hours. And I picked this data because the wind conditions in that case, they actually changed quite a lot. So there was a variation from 3 to 60 meters measured at ground and 4 to 90 meters per second measured by the LIDAR. And that's what the data from the LIDAR looks like.

So on the upper graph, I picked three levels. So in right, that's the most upper level. It's 290 meters, and the other levels are lower at 140 and 40 meters above ground. And what you can see is, of course, the layering. So the red one, the red curve is always above the others.

And there is also a strong increase in wind speed around this time, 8 p.m. of the first day. And to basically show or to indicate that we were able to keep the system airborne, you see the normalized force

on the lower part over the whole flight. So we basically never exceeded a safe limit that we set. And but still the force was significantly larger than zero. So the system was really at a high load.

And I want to go a bit into more detail about what you can see from the LIDAR measurements. Because with this increase at this point, there is what you can see is that actually also the variability changes suddenly. That's an interesting transition that happens here.

What I plot in lower plot is exactly this variability at the low and high level. So in red, it's the difference between the maximum and minimum at 290 meters. And in blue, it's at the lowest level at 40 meter.

And you can see that actually before the wind changed, let me quickly, before the wind changed, at all the levels the variability was more or less the same. But after that, there was a large variability at the lower levels. But at the high levels, it was not that bad. And later, it got also bad at the high levels.

And what this looks like if you plot it as a profile is that well, first you have a smaller mean velocity. And you have the same variability all over range. So the maximum and minimum are pretty close.

And then you have this transition. And obviously, that raises the question which altitude is optimal to fly at in this case or in the both cases. And that's in the sense of an outlook. If you have this data available, you can then go

and analyze this in more detail. So starting from this equation, and assuming that we have a wind gradient known, and we have a mean line length that we operate at,

we can actually evaluate this term which corresponds to the wind speed projected into the line direction at a given height. And when you maximize that, you basically, the point that maximizes this term is your optimal altitude to your operator.

And I showed this for these two cases. So the one is at 8 p.m. and the other at 10 p.m. So just before and after this wind increase because the profile has changed just as an example. And what one can see in the first case for this profile, if you operate at an average line length

of 600 meters it actually doesn't make that much difference if you operate at 150 meters or at 250 meters. But if you have a shorter line length, then definitely it makes sense to fly lower. And at the case for the second profile, the differences are even more significant, right?

So if you had only 400 meters of operation length available in average, then you would lose quite a significant amount. And remember that this scales cubically. So if you take this difference, it's probably between 15

and 13 meters per second, but that difference goes into the power then cubically. And the nice thing about target point control, it's actually quite straightforward to just adapt the elevation angle to match this correct optimum altitude.

So with this, come to the conclusions. So yeah, we were able to operate over exceeding 14 hours in automatic operation at conditions that have been shown by LIDAR data to be challenging in the sense

that the variation in the wind speed itself and also in the wind gradients over time is quite significant. And one aspect that helps us is that we can separate a wind and flight trajectory control using target point control. And target point control doesn't make any specific assumption on wind speed or wind gradient.

And also I talked about the possibility of further adapting the target point elevation to increase the overall performance. The wind variation as it changes with altitude, that can be really a significant disturbance signal

and it should also be considered when choosing the optimal altitude to operate. There are some open questions. So one question is how accurately can you actually estimate the wind profile if you don't have a LIDAR? Are there any rare or extreme wind shear situations which would still require specific treatment?

And well, in this talk I only talked about the variation in wind speed with the altitude, but talking about the wind direction would definitely also be an interesting topic. So with this, I would like to close and ready.

Yeah, thank you for the very nice presentation and impressive results and congratulations obviously on the day-long operation. So I was curious, on the reel-out phase, are you maintaining a constant elevation angle

over the entire phase and how difficult would it be to prescribe a varying elevation angle which is optimal for each of the tether lengths along the reel-out? Yeah, so that's actually an interesting question. So we can do both. So we have both operation schemes. But if you're at short line length, then usually it's better

to first reel-out at a constant elevation angle. And then from the data that you can see, it makes sense then at the longer line length maybe to operate at a constant altitude. Yeah, and in the field,

have you tried varying the elevation angle during the reel-out and are there any operational or dynamic challenges that have come up in dynamically varying it? So yeah, so we do these kind of tests to see and of course if you have a day-long operation, then you have the

chance to do such variations of parameters. Yeah, I'm thinking during a single reel-out phase though, are there any dynamic issues that come up if you're starting at an elevation angle of 40 degrees and coming down to 20 degrees or something like that? Well, what you definitely noticed,

I mean there's the underlining system dynamics, right, if you go too high with the elevation, basically there is not enough wind to continue reeling out. So that's one thing. You're out of the wind window. Yeah. That's one aspect. And the other thing is this term with the variability, right?

So sometimes you might think it's better to really fly very, very low just because, well, of the cosine losses. But if you have this wind shear situations where we have strong variability at the lower levels, that's why I put this data up, is maybe it's not a nice point to fly in, right?

So and I think that should be addressed a bit more because in the models that are being discussed, I haven't seen a lot about this in the literature so far. Thank you. I won't monopolize the discussion anymore.

Thanks for the nice presentation, Moffat. I was wondering what the typical height range is that the system sweeps during a real-life phase. So that was on the one slide. So for the large system, it's about between 150 and 350. But yeah, we're in the middle typically.

So but that's the range that we operate. Yeah. Thanks a lot. Also, I have a question about the real in phase. So how do you depower? What's the ratio between, let's say, peak to average power in a typical cycle? And during these 40 hours, what was the average electrical power that you generate?

I'm sorry if you're able to say these things. I would be very curious. So maybe about the last point. So that's going to be the keynote at 2 p.m. I think. Maybe it's shifted a bit now. And there will be a bit more about this.

So we basically, so I can rather say that this system is designed to achieve an annual yield of 800 megawatt hours in, yeah, an annual yield 100 megawatt hours. And in the development, in the path there, we're about halfway there too.

So building a system that has a power curve that can achieve such an annual yield at the site like in northern Germany as a reference site, that is more or less what I can say about this. The peak to average, no, I don't have that numbers now.