Application of aircraft’s flight testing techniques to the aerodynamic characterization of power kites

Résumé

Environmental concerns such as global warming, air pollution or urban sprawl are driving a slow transition from a fossil-fuels based economy to a new renewable, de-carbonized one. However, this transition represents a big technological challenge in which, renewable energy sources such as wind energy, are called to play a prime roll. Since the first commercial wind farms were commissioned in the Eighties, wind energy generation industry has experienced a continuous growth and technological development becoming by far the renewable generation system with a higher installed capacity. First wind turbines were located at strategically well-situated sweet spots, while, as the technology improved, newer developments allowed to exploit less than ideal locations in a continuous effort to increase wind farms capacity factors and lower production costs. However, architecture scalability is starting to show as a limitation factor, as some involved technologies are reaching an inflection point with wind turbines peak power reaching the 5MW per unit. In the last two decades, Airborne Wind Energy Systems (AWES), a new approach to wind energy generation, has been largely explored with very active research organizations mainly in Europe but also in USA and Japan. In this new approach, several technical solutions have been proposed to harvest energy from upper layers of the atmosphere, where more stable and strong winds are found. Most of the proposed systems are based on tethered aircraft (kites) flying crosswind trajectories generating an aerodynamic speed several times the real wind speed. These systems appear promising at: 1) increasing power production due to higher capacity factors and 2) reduce deployment costs due to the use of flexible structures instead of fixed, rigid ones. However, these new developments are still far from commercial deployment, with consensus TRLs in the industry ranging from 4 to 6. Long lasting autonomous operation, including autonomous take-off and landing, emergency/bad weather handling and availability factor optimization are, together with the need for technology homogenous benchmarking and airworthiness demonstration, key millstones towards AWES commercial deployment. De-risk of this development process is highly related with the capability of the industry to provide accurate simulation environments based on realistic modelling of these systems. To provide such modelling of AWES, a precise mathematical characterization of the generated aerodynamic forces and torques as a function, among others, of the kite aerodynamic speed and its attitude within the airflow (i.e. an aerodynamic model) is required. A comprehensive revision of the published techniques used for describing tethered aircraft’s flight dynamics has been carried out. Three main different approaches have been identified: 1) point-mass models, 2) rigid-body and 3) high-fidelity models, with increasingly complexity and fidelity. Point-mass models neglect all the rotational dynamics of the tethered aircraft and are often coupled with very simple aerodynamic models resulting in extremely simple simulators often used in control/optimization problems. Rigid-body approaches are coupled with more complex aerodynamic models similar to the ones used for conventional aircrafts in which the actual modeled aerodynamic forces and torques depend (among others) on the orientation of the aircraft within the airflow and are used for advanced 6DOF simulators. Finally, high-fidelity models such as multi-plate, lumped parameter or multi-body models try to increase modeling fidelity of flexible or semi-flexible aircraft (kites), which are not properly described as rigid bodies in particular problems such as the kite turning mechanism in which kite deformation plays a prime roll. Additionally, tether dynamics are differently considered in the described models, ranging from very simple massless, single rigid inelastic tethers to multiple flexible ones. Most of the revised works are lacking at the aerodynamic characterization of the simulated aircrafts, as they usually are academic approaches to the problem with ad-hoc aerodynamic models. On the other hand, modern techniques such as computer fluid dynamics (CFDs) are not always well suited for this regard due to kites' flexible nature, albeit some efforts on parametric CFD modelling have been described. Nevertheless, most of the efforts on aerodynamic characterization of tethered aircraft have been carried out based on empirical flight testing of real kites. A revision of published techniques on tethered aircrafts' flight testing has been carried out and although important efforts have been made in the last few years by using computational fluids dynamics and experimental methods, trustworthy and complete aerodynamic models of aircraft used in AWE applications are not yet available. This thesis proposed the use of the Estimation Before Modeling (EBM) technique for the aerodynamic characterization of AWES based on experimental flight data. The EBM is a two-steps method. Firstly, the state vector of the aircraft during the whole flight is reconstructed in the so called, estimation or Flight Path Reconstruction (FPR) phase. This state vector includes, aside from the generic state variables (i.e. position, velocity or attitude) the aerodynamic forces and torques. The aerodynamic coefficients are derived a-posteriori from the reconstructed forces, torques, and the other state variables. In the second step (modeling phase), a mathematical model is fitted to the reconstructed aerodynamic coefficients. The biggest advantage of the EBM technique is that both phases are independent, and the modeling phase becomes a model fitting problem isolated from the aircraft equations of motion. Consequently, different mathematical model structures and model fitting techniques can be used without the need to reformulate the system process model. To demonstrate the application of the EBM technique to tethered aircraft, an ad-hoc experimental setup has been developed together with a flight-testing methodology for kites. Several flight tests have been carried out and the EBM technique has been applied to the obtained measurements. A conservative approach has been applied during this thesis work. Firstly, a simpler experimental setup was developed with low-cost, commercial of the shelf (COTS) hardware, and a four-line, leading edge inflatable (LEI) surf-kite. The obtained data was used to validate a first version of a FPR algorithm and reconstruct the full state vector of the kite. After the first flights were concluded and the FPR algorithm was validated, conclusions on the required experimental setup improvements were obtained and implemented. The FPR algorithm was then updated and reformulated to be used also with two-line rigid frame delta (RFD) kites and the reconstructed aerodynamic forces and torques were used to provide some characteristics of the aerodynamic behavior of both LEI and RFD kites. The main elements of the first version of the experimental setup used to validate the proposed FPR method were: 1) Kite: Two LEI, 10m2 and 13m2 surf-kites were selected for the first flights. The size of the particular tested kite was chosen depending on wind speed in order to keep the aerodynamic forces within the design range. The kite was instrumented with an onboard Inertial Measurement Unit (IMU), a GPS, a magnetometer and a pitot tube with static and dynamic pressure sensors. Those instruments are secured to the kite by means of ad-hoc designed, 3D printed rigs, and the provided signals are recorded onboard the kite. 2) Control Bar: The kite is manually controlled by a human pilot on the ground. A COTS kitesurfing control bar was firstly used. The control bar is instrumented with load-cells in each tether and for the first flights, distance sensors were also included to provide measurements of the control bar state. The signals provided by those sensors are recorded by a laptop on the ground. After analyzing the obtained measurements, it was determined that the selected sensors were working within their expected measurement ranges and the provided data could be used to reconstruct the space state trajectory of the kites. However, the static and dynamic pressure signals from the onboard pitot tube were found to be heavily noisy and biased, and the need for a new sensor was concluded. Additionally, some of the reconstructed state variables, such as the kite attitude and the kite trajectory, were compared with the outputs of an independent estimator provided by the software Px4 and showed in good agreement. Two basic control maneuvers were studied to validate the coherence of the reconstructed state vector of the kite, a pull-up maneuver and a steering maneuver. In both cases the kinematics of the kites were in good agreement with the command inputs from the control-bar, showing coherent correlation between the kite attitude, angular rates and kite ground position and velocity. During this maneuver, the reconstructed behavior of the aerodynamic forces and torques showed coherent with the evolution of the kite attitude and aerodynamic speed. This validation allowed us to conclude that the FPR algorithm produced coherent results. However, the poor performance of the onboard dynamic and static pressure sensors, together with the lack of direct measurements during the flight of the wind direction and speed, and the kite attitude within the airflow, resulted in the uncertainness of the a-posteriori reconstructed air-data (angle of attack (AoA), angle of side slip (AoS) and aerodynamic speed (TAS)). In particular, the estimated AoA resulted abnormally high during the studied maneuvers in comparison with conventional aircraft and other published studies with bigger LEI kites. In order to verify these findings and to increase the overall precision of the FPR algorithm, it was learned that the experimental setup should provide direct measurements of the kite AoA and AoS, and the wind state vector. From the experience obtained during these flight campaigns, it was also concluded that i) the handling qualities of the kites equipped with the required onboard sensors were heavily affected due to the added payload and the modification of the kite CoG location and ii) the COTS control bar instrumented with load-cells and distance sensors showed particularly intrusive as the weight of the load-cells inserted in the control lines heavily affected the controllability of the kite, resulting even in the kite stalling and crashing when flying in wind lulls. The final version of the experimental setup was designed to address the observed limitations during the proof-of-concept flight tests and included: 1) A new hand-made control bar designed to minimize the interference of its sensors with the kite flying qualities. 2)A new two-line RFD kite, which allowed to place the on-board sensors without modifying the kite CoG location, was expected to fly faster and with a smaller AoA, and provided and easier and better modelling of its inertial characteristics. The LEI kites were also conserved in the new experimental setup for comparison, albeit changing the position of the onboard instruments to improve their flying qualities resulted not possible due to their semi-rigid nature. 3) A wind station providing direct measurements over the wind vector. This wind station was mounted on the top of a portable tripod and is designed as a rotatory platform pointing to the wind by means of a wind vane. The wind station is instrumented with an IMU, a magnetometer and a pitot tube with static and dynamic pressure sensors. The signals provided by these sensors are recorded onboard the wind station. 4) A new multi-hole pitot tube coupled with a COTS air data computer capable of providing calibrated, high quality measurements of the kite AoA and AoS plus static and dynamic pressures. Additionally, the EKF algorithm was consequently updated to incorporate the new measurements and adapted to be used also with the RFD kites. Two flight campaigns were performed with the new experimental setup, one with the 10m2 LEI kite and one with the RFD kite. These flight campaigns provided quantitative and qualitative results about the aerodynamic characteristics of both type of kites and revealed significance differences among them. Flight testing with the final version of the experimental setup and FPR algorithm resulted in the following conclusions: After the last flight campaign with the 10m2 LEI kite, we can conclude that the flying qualities of the kite are still far from ideal due to the forwarded location of its CoG. In an effort to minimize this effect and to increase the flight safety, this campaign was carried out in the coast of Huelva, looking for steadier sea winds in comparison with the gusty ones found in Madrid. However, the LEI kite crashed only 65 seconds after launch, and one of the multi-hole pitot tubes was destroyed. Unfortunately, the amount and quality of the recorded data was limited. From the direct measurements of the calibrated pitot tube and air-data computer, we can confirm that the LEI kite flew most of the time with an AoA greater than 50o This measurement corroborates the reconstructed AoA of the first proof of concept flights and further validates the integrity of the FPR algorithm. The observed AoA constitutes a big difference with other works carried out with much bigger LEI kites during power producing flights. A 2-second-long maneuver could be identified within the flight in which the kite flew a crosswind trajectory with higher TAS and lower AoA. During this maneuver, the AoA sensor was not saturated, and the recorded data was used to feed the FPR algorithm. A comparison of the reconstructed lift coefficient (CL) with the kite AoA () showed an inverse relation among them. It has been concluded that this effect is coherent with a post-stall aerodynamic behavior of the wing and differs from the observed behavior of large power-producing LEI kites. The flight campaign with the RFD kite showed excellent handling qualities of the kite, and close to the behavior of the clean kite, rewarding the efforts done to keep the original CoG location. This allowed the pilot to perform a 200-second-long flight without any incident and provided abundant and quality data for the FPR algorithm. The adaptation of the FPR algorithm to the two-line kites was also validated. As with the LEI kites, the kinematics of the kite showed coherent during the flight. Four figure-eight maneuvers were selected to study the aerodynamic characteristics of the kite. The CD, CY, CL and Cm coefficients were reconstructed, and a statistical analysis of the reconstructed coefficients was carried out. From this analysis, we can conclude that the kite had a diminished aerodynamic performance in comparison with theoretical models of an equivalent wing with the same aspect ratio. For such a wing, and a low-speed attached-flow, a lift-curve slope C_(L_α )=4 was expected, while an average C_(L_α )=2.54 was obtained. On the other hand, the induced drag predicted by the lifting-line theory is only around C_L^2/πARe≈0.057 for AR=7 and a plan form efficiency factor of e=0.8, versus the obtained C_D≈0.4 Albeit smaller AoA were registered during the flight of the RFD kite in comparison with the LEI kite, the magnitude of the estimated AoA was still significantly high. However, opposite to the LEI kite, the RFD kite lift coefficient did not show a post stall behavior (the CL versus  diagrams show that CL grows monotonically even after reaching an AoA higher than 30o). This may be due to nonlinear aerodynamic effects, in which the convex curvature of the flow near the leading edge produces a suction that increases the CL The presented experimental setup and flight testing methodology have played a distinctive role in UC3M Tethered Aerospace System and Airborne Wind Energy Group. Related academic activities of the group have resulted in multiple publications in JCR indexed journals, conference presentations and nine Bachelor and Master Theses. Future work in the application of EBM techniques to tethered aircraft like the ones used in AWEs, may tackle three main aspects: 1) Validate the described aerodynamic behavior with bigger LEI kites flying bigger and wider crosswinds trajectories. The application of the described instrumentation to an AWE system, such as Kitepower B.V system, could provide valuable information about the aerodynamic characteristics of power-producing kites. This information could be used to improve our knowledge of some aerodynamic coefficients which are key when modeling the system performance or to improve the kite control system. Additionally, the multivariate orthogonal functions method described by Morelli can be tested to provide a full aerodynamic model. 2) Improving the experimental setup towards a mechanical (fly-by-wire) system which eliminates de manual inputs to the system. This new setup should be able to stand higher loads on the tethers, allowing for faster and wider crosswind trajectories, while at the same time eliminating the uncertainness introduced in the measurements by the manual control bar. 3) Improving the FPR algorithm with a tether model, which will allow to introduce the sag of the tether in the filter process model. This will provide a better projection of the tether tensions over the aircraft body-axes and will further improve the precision of the EKF, especially if wider trajectories are studied, in which the straight line approach for the tether cannot be hold. Some of the described tasks, particularly the improvement of the experimental setup towards fly-by-wire operation, are currently ongoing in the UC3M GreenKite-2 project.