the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Nov 2024
System design and scaling trends for airborne wind energy
Abstract. So far, the size of horizontal axis wind turbines (HAWTs) has steadily increased, but recent studies and market decisions suggest that this trend may come to an end. Airborne wind energy (AWE) is an innovative technology that differs from the operating principles of HAWTs. It uses tethered flying devices, denoted as kites, to harvest higher-altitude wind resources. Kites eliminate the need for a tower but introduce a penalty in power generation since the kite has to spend part of its aerodynamic force to counter its weight. The differences between the two technologies lead to different scaling behaviours, and understanding these and the design drivers of AWE systems is essential for developing this technology further. To this end, we developed a multi-disciplinary design, analysis and optimisation (MDAO) framework which employs models evaluating the wind resource, power curve, energy production, overall component and operation costs, and various economic metrics. This framework was used to design fixed-wing ground-generation (GG) AWE systems based on the objective of minimising the levelised cost of energy (LCoE). The variables used to define the system were the wing area, aspect ratio, tether diameter and rated power of the generator. The framework was employed to find optimal system designs for rated power ranging from 100 kW to 2000 kW. The results show that kite mass, energy storage, and tether replacements are the key LCoE-driving factors. Moreover, in contradistinction to HAWTs, the total lifetime operational costs are equal to or higher than the initial investment costs. This distribution of costs over the project’s lifetime, rather than as a large upfront investment, could make it easier to secure project financing. The scaling results show that the LCoE-driven optimum lies within the 100 kW to 1000 kW system size. The reason for this is that the kite mass penalty increases the cut-in and rated wind speeds, reducing the capacity factor of the larger systems. Sensitivity analyses with respect to extreme scenarios considering technological advancements, financial uncertainties and environmental conditions show that this optimum is robust within our modelling assumptions.
- Preprint
(1132 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on wes-2024-161', Anonymous Referee #1, 16 Dec 2024
The authors present a comprehensive analysis on design and scaling of airborne wind energy with interesting results. Specifically, the authors are assessing how it scales with size under certain assumptions, identifying the key cost drivers for airborne wind as well as providing estimated levelized costs of energy.
Items that should be considered are:
- There seems to be no variance between the mass dependence of horizontal take-off and landing (HTOL) kites, ground-generation kites (ground gen) and air-generation kites (air-gen).
a) HTOL kites would expect higher structural mass to handle landing loads, but probably better than VTOL configuration
b) VTOL lift-based kites would expect added mass due to VTOL weight as mentioned in the article
c) VTOL kites with air-gen would expect added mass of VTOL weight, as well as higher drag of tether due to having a conductive tether (thicker diameter and more mass).
i) Perhaps add a comment on system mass depending on configuration and how this is (not?) accounted for in the analysis
- The article uses a lot of colors. When printing the article for reading to do a proper review, it was difficult to discern differences between the meanings, and I had to use digital version for assessing the figures. Consider to add dashed lines for it to be readable in black and white.
- Page 2 line 39: Recommend to use lift instead of thrust for power production. It is typically the lift of the kite that produces pulling force, not the thrust.
- Looking quickly through the code base at github it is not clear how the tether drag is accounted for. I see that you mention it on page 10, where it is considered as a lumped at the kite, but it remains a question if this is correct, or representative enough. At the core, there are two main factors that limit the cut-in speeds: kite mass and tether drag where the latter tends to dominate. This means that the lumped model might not capture the tether drag effect to a sufficient degree.
- In terms of figure 22, and scaling of airborne wind, I would say that it is too early to say anything about scaling beyond 1MW, but that this is indicative based on our current understanding (and of course the square cube law in terms of mass). The output of this figure is also closely tied to the assumptions in the article.All in all, this is an excellent paper on scaling trends for airborne wind energy and provides valuable insight to the reader.
Citation: https://doi.org/10.5194/wes-2024-161-RC1 -
RC2: 'Comment on wes-2024-161', Anonymous Referee #2, 31 Dec 2024
General comment:
This work is very much desired and appreciated, as the system design of airborne winter energy systems with respect to scaling and maximum viable or techno economically optimal system size is a critical consideration in the design and in the techno economic viability. This work also validates the assumption that AWE systems require and offer substantially financing models With lower upfront capital cost and increased operation and maintenance and system replacement cost over the operational time in comparison to conventional wind energy systems.
Edits
Please define consistently and clearly throughout the paper which are independent and dependent variables.
A variable dependency network diagram could be useful prior to the system design framework under 2.2
Lines 129 to 131 appear convoluted and not fully consistent. Please define what remains constant.
In Line 140 and later one could change from “Developers” to “Developments” or “Technology Developments” to move from entity to process
144 Please correct “area of operation A_oper is the ground area density”. Either area or areas density. If area density, the please define the quantity the area relates to. “Area per what”
622 and several times before: Please, always specify which cost you refer to OpEx or CapEx or even more detailed.
626 to 629. The repetition of the working principle in the conclusions is not required.
Citation: https://doi.org/10.5194/wes-2024-161-RC2 -
RC3: 'Comment on wes-2024-161', Anonymous Referee #3, 31 Dec 2024
This paper presents a tradespace characterization and sizing optimization for a particular airborne wind energy (AWE) architecture. The technical work is of high quality and the paper's overall methodology is well suited to the problem statement. The scientific topic and literature contribution are valuable given the early stage of AWE technology. However, there are still notable issues that need to be addressed prior to publication. My comments, both major and minor, are:
- I suggest deleting the first sentence of the abstract. This paper doesn't want to venture into the can of worms of the connection between power rating and LCOE and whether HAWTs will continue to grow in rotor diameter and power rating or not. This is an AWE paper and the opening sentence should focus on AWE not HAWTs.
- After the first sentence of the introduction, it would be appropriate to explain why wind turbine sizes have grown. The increased power ratings and diameters have been associated with sharp decreases in the cost of energy for a given project size, such that wind is now one of the cheapest modes of generation. This context is otherwise missing.
- Page 2, line 36, suggest new paragraph before, "The rotor nacelle assembly..."
- Page 3, line 50- I do not understand why onboard storage is required for the AWE system. Power electronics coupled to the generator suitably "smooth out" the grid signal for wind turbines. Why does that approach not work here? The addition of storage (ultracapacitors in this work) certainly adds a steep cost penalty to the design.
- This paper leans heavily on the prior work in "Power curve modeling and scaling of fixed-wind ground-generation airborne wind energy systems" by 2/3 of the same authors. I imagine there was some discussion at one point as to whether the two papers should be combined into one or kept separate. The first paper also delves into tradespace exploration for some of the same design variables and their link to power performance. I understand that this work focuses on LCOE and overall system sizing, but the similarities between the two papers should be addressed more explicitly than is currently done. Otherwise, it is unclear to the reader what the novel contribution to the literature is in this paper. I would suggest a clearly worded paragraph near the end of the Introduction that says something like, "Prior work by the authors presented the model and the power performance tradeoffs [cite]. This paper builds on the prior work and makes the novel contribution of..."
- I find the definition of design variables and constraints in sections 2.1.2 and 2.1.3 confusing. The design variables include maximum wing loading and maximum tether stress. These material/structural limits are more commonly applied as constraints in other MDO papers. I would have expected more tangible parameters such as tether diameter or tether/wing material as the DVs. The discussion of max wing loading and tether stress around Line 127 also uses phrasing that makes them sound like constraints. Furthering my confusion is the list of optimized variables on Line 185 (reel-out stroke length, wind lift coefficient, kite speed, pattern radius, elevation angle, cone angle) that are not included in Table 2, yet are optimized as though they are DVs. My guess from the XDSM diagram is that there are different nestings of sizing optimizations occurring. For the purposes of discussion, better to mention all DVs in Section 2.1.2 and then also tag which modules/nesting level they are associated with.
- Section 2.6.2 mentions a hollow-core fiber tether. Is there a copper wire/cable for communication, control, SCADA-type parameters coming to-and-from the kite?
- I liked the inclusion of fatigue estimation for the tether! That is often tough to do in these conceptual design levels of fidelity.
- I believe there is some confusion with regards to BOS costs and the categorization of CapEx and OpEx. I am also suspicious that the cost comparison between AWE and HAWTs is not apples-to-apples. I therefore am not sold on the summary conclusion that AWE spreads out its costs over the lifetime better than HAWTs. I encourage the authors to review the annual Cost of Wind report from NREL (latest edition: https://www.nrel.gov/docs/fy25osti/91775.pdf) as that might clear some points up. The reason for my confusion and suspicion is:
+ Line 407 says that BOS consists of "site preparation, ..., operation maintenance and decomissioning". This has me confused because BOS is considered an upfront CapEx expenditure, yet operation and maintenance is included? If so, then what constitutes OpEx?
+ Figure 16 shows CapEx to be purely the kite system cost and BOS to be included in the OpEx category. This is a different convention than I am used to.
+ The authors need to cite a reference for wind energy costs to support their claim that AWE spreads out the costs over the lifetime more evenly. Somewhere between Lines 525-530 would be appropriate. The units in Figure 16 are not the same as the Cost of Wind reports, in addition to the difference in categorization of the costs, so I am not convinced by the claim but unable to do the mental math to prove the authors case.
+ In Line 605, the authors state that only a single kite was considered and no farm-level sizing, costs, or performance effects are accounted for. This is a bit surprising because this is a key driver for turbine upscaling. Fewer turbine positions for a given plant rating drives down the BOS costs and leads to lower LCOE. To compare AWE LCOE trends in the results here vs HAWCT LCOE trends in the marketplace without ensuring that the BOS cost trends are consistent makes me suspicious. I would suggest the authors provide a back-of-the-envelope calculation for HAWTs in a table here to ensure consistency in the numbers and assumptions. Finally, unless I missed it earlier, this assumption of looking at single turbine costs only should also be stated earlier.- Discussion in the paragraph at Line 540 might be easier to understand by mentioning that the square-cubed law is likely at play.
- Table 9 is labeled as "optimized", but this is *not* the output of an optimizer, correct? This is the "best performer" of the parametric study in Table 8? If so, I would remove the word "optimized" from the caption and the discussion in paragraph at Line 540. Furthermore, in that paragraph, only the kite area is discussed for the trends observed. What about discussion for why the other parameters change or stay constant?
- Why wasn't full system optimization used to generate Table 9 instead of just a parametric study with DV steps that some might argue are too coarse? I understand that it can be arduous to show consistent trends in a family of optimized designs as it requires lots of restarts at different initial conditions, etc, to ensure a robust and consistent output. However, without that I feel like the authors are not taking full advantage of the model capability they have presented and the insight into the design drivers and physics isn't as clear as it could be.
+ Given the parametric approach taken in Table 9, I am also suspicious about Table 11, given the nice round numbers. Was that a true optimization or just the best performer is a parametric sweep?- The discussion in 3.2.1 around the capstone plot in Figure 22 could be stronger. Why is 500kW a robust optimal size for this AWE architecture across all of the scenarios? Even when mass is 50% less and wing area increases, the optimal rating stays the same- why? I believe the explanation that the kite has to use part of its aero lift to keep aloft, instead of using 100% for power generation, but does that explanation hold if a mass reduction by 50% doesn't have an impact on the design? What if the mass reduction were greater, or even in the 90-99% range? Is it because the steps in rated power are too coarse? What other hypothetical scenario would cause this optimal size to shift?
- In Table 11, using the Scenario number to label the columns is confusing and requires lots of page flipping to understand what is going on. Some text-based shorthand to describe each column would help.
- If Table 11 is from a parametric search, please do not use the word Optimized in the caption (same comment as above).
Citation: https://doi.org/10.5194/wes-2024-161-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 153 | 67 | 8 | 228 | 1 | 2 |
- HTML: 153
- PDF: 67
- XML: 8
- Total: 228
- BibTeX: 1
- EndNote: 2
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1