the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Jan 2025
Kite as a Sensor: Wind and State Estimation in Tethered Flying Systems
Abstract. Airborne wind energy systems (AWESs) leverage the generally less variable and higher wind speeds at increased altitudes by utilizing kites, with significantly reduced material costs compared to conventional wind turbines. Energy is commonly harnessed by flying crosswind trajectories, which allow the kite to achieve speeds significantly higher than the ambient wind speed. However, the airborne nature of these systems demands active control and makes them highly sensitive to changes in wind conditions, making accurate wind measurements essential for steering the kite along its optimal trajectory. This paper presents an advanced sensor fusion technique based on an iterated extended Kalman filter (EKF) for state and wind estimation for AWESs. By integrating position, velocity, tether force, and reeling speed, this method provides accurate estimations of system dynamics, including kite orientation and tether shape. The estimates of the wind speed and direction are compared to lidar measurements, showing a strong agreement across various atmospheric conditions. The results demonstrate that this approach can effectively capture the transient dynamics of atmospheric wind using sensors typically already present in AWESs, making it suitable for supervisory control strategies and ultimately enhancing energy efficiency and system reliability across diverse atmospheric conditions.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Wind Energy Science. R.S. is a co-founder of and advisor for the start-up company Kitepower B.V., which is commercially developing a 100 kW kite power system and provided their test facilities and staff for performing the in situ measurements described in this article. Both authors were financially supported by the European Union’s Meridional project, which also provided funding for Kitepower B.V.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.- Preprint
(16509 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on wes-2024-182', Anonymous Referee #1, 13 Feb 2025
Further development and optimization of Airborne Wind Energy Systems requires of advance characterization and monitorization of tethered aircrafts flight dynamics. However, this can result challenging specially for flexible soft kites, as they are highly susceptible to changes in wind speed and direction. The paper presents a sensor fusion technique which is an evolution of previous works based on EKFs, effectively estimating wind velocity at flying altitude using the kite as a sensor. Albeit there is no significant novelty in the theory of the methods used, the application of the EKF and the collected experimental data, in particular, the implementation of a model that accounts for both tether sag and kite control unit inertia and aerodynamics, should be interesting for the community.
Overall, the article is well written, figures are high quality and the mathematical formulations and the technical information are presented accordingly. However, I have one major concern for the publication of this paper. For me is very difficult to follow the reasoning about the EKF implementation/design, specially the sensors used and the observation model of both the V3 and V9 kites flight tests. For example:175-“Overall, the sensors that are least susceptible to the intrinsic deformations of the soft kite and the high accelerations of the system, and thus more reliable, are the GPS, the load cell (for tether force), and the mechanism measuring the tether length and tether angles. These sensors can maintain their accuracy despite the flexible nature of the kite. Consequently, the proposed sensor fusion model primarily relies on these measurements, resulting in a minimal sensor setup consisting of a GPS (for position and velocity), a load cell, and a tether length measurement”.
311-“The required minimum measurements are the position and velocity of the kite wing”
393- “In this section, we explore various sensor setups and model configurations. The different EKF models are detailed in Table 2. The additional measurements listed in the table are used alongside the minimum required sensors for a system with a KCU, which include the position, velocity, acceleration of the kite wing, tether force, and reel-out speed.”
In my opinion, the paper should be optimized for increased clarity and conciseness. The authors should make an effort to facilitate the reader the matching between caps 4 and 5.
I also have some minor comments:- Line 115-120 - Further discussion about direct measurement of in-situ aerodynamic angles of attack and sideslip is welcomed (Oehler and Schmehl, 2019). Using booms for isolating aerodynamic sensors from aircraft’s perturbations is a well-known practice in the aerospace industry during development phases.
- Line 143 - Fig2. could be improved by showing a detailed view/scheme of the implementation of the load cell sensing the tension of the tether without interfering with the reel in-out system.
- Further detail about how airborne data is logged/transmitted to the ground and synchronized with in the ground measured data is also welcomed.
- Line 396, 405-410 - In my opinion, the output of the Px4 position and velocity estimations should not be used as measurements for the EKF as errors are not guaranteed to be zero mean and gaussian. Instead, raw GPS position and velocity from PixHawk Gps should be used, plus a measurement error model to guarantee that the measurement noise described in eq.24a and 24b is Gaussian white noise. (R.Borobia et al. 2018). This change will eliminate the dynamics of the PixHawk onboard estimator increasing the stability of the filter.
- Line 411 - In Fig9. The measured Euler angles are the estimated ones by Px4?
- Line 419 - Calculation of Yaw angle assuming alignment of the kite body axis with aerodynamic velocity vector assumes no side-slip during the flight. However, direct measurement of side-slip angle showed non zero values for a inflatable kite (R.Borobia et al. 2021)
- 520 -530 - The underpredicted side-force could be related to assuming zero-side slip angle?
-
RC2: 'Comment on wes-2024-182', Anonymous Referee #2, 25 Feb 2025
Notes on WES-Cayon & Schmehl 2025
Version Preprint
This paper addresses some very interesting questions on kite dynamics using fusion techniques for the analysis of experiments. It draws on the substantial experience of the Delft team, which is undoubtedly the world's leading centre for kite studies. The paper is composed of 41 pages including 4 pages only for the introduction. Section 2, 3 and 4 present the material, i.e. system, sensors and fusion technics used for analysis. Section 5 presents many results in 4 subsections (kite kinematics, system dynamics (aerodynamic identification and turning dynamics), wind estimation and turbulence measurements). It ends with a broad conclusion.
The article could probably have been split into 2 separate articles. That would have made the point clearer. The language is highly technical, which often makes it difficult to read. It is often difficult to find the definition of one of the many variables. In such cases, an appropriate nomenclature is essential. Given the complexity of the problem being addressed and the large number of variables involved, the use of full variable names in the text should be preferred most of the time for easier reading.
One of the main gap in the document is a clear definition of the reference frames used in this study. With all the information available, the reader probably has all the information needed to find the definition of each variable. But this definition is uncertain, and the reader may make a mistake. An appendix at the end of the document gives the values of the experimental parameters and the model used during the tests. This should ensure reproducibility of the results. The codes are also provided.
Minor comments
Line 80, The authors claim that their model improves on the existing one by considering the sag of the lines and the dynamics of the KCU. We expect them to present comparisons of measurement results with and without taking these quantities into account. This would make it possible to visualize concretely the impact of taking these factors into account.
Line 112-114: the KCU appears to be a reference for the orientation of the kite, which does not
Line 113-114 The depower angle deserves a more rigorous definition.
Line 119 Fig.3.a does not exist. Replace by Fig.3
Line 160 Figure 3 shows 4 boxes on 4 battens. We need to specify which one is the IMU and which other sensors are in the others.
Other comments are available with notes directly in the article
Data sets
Kite power flight data acquired on 8 October 2019 M. Schelbergen et al. https://doi.org/10.4121/19376174.V1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 424 | 54 | 6 | 484 | 7 | 5 |
- HTML: 424
- PDF: 54
- XML: 6
- Total: 484
- BibTeX: 7
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1