Maher Brahim

Tuesday, September 23, 2025, 9:00

Room 02-012, Georges-Köhler Allee 102, Freiburg 79110, Germany

Abstract: This thesis focuses on the development of an optimization-friendly mathematical model of a renewable Airborne Wind Energy System (AWES), in particular the Leading Edge Inflatable Kite from Kitepower. The aim is to accurately map the dynamics of figure-eight flight maneuvers during the reel-out phase. Therefore, the model parameters are estimated by using provided experimental flight data. Initially, the measurement data was prepared and filtered. A Kalman Filter (KF) and an Adaptive Kalman Filter (AKF) are used to derive additionally the model's control inputs from this data sets. The AKF indicates an improvement in the accuracy of the actual steering rate estimation, by more than 50% compared to the residual values of the KF. For the modeling and the optimization of the AWES, the AWEbox framework was extended by developing an one-point mass model, which takes into account the roll dynamics of the kite. The combination of the model and the measurements results in the creation of a Nonlinear Least Squares Problem (NLPS) parameter estimation. The integration of the dynamics is performed by a Radau-IIa direct collocation scheme. The findings result in an enhanced parameter identification process comprising two steps: First, the three parameters for the aerodynamic forces and a tether length offset are estimated by solving the NLPS. Therefore, the Gauss-Newton Hessian approximation (GN) is used instead of the exact Hessian, which improved the computational time by a factor of 2-3. An additional sensitivity analysis demonstrated good confidence bounds for the estimated parameters, especially for the tether length offset. Secondly, model validation is performed by using independent measurement data. Therefore, the flight trajectory with the fixed parameters and free initial conditions was divided into predefined time windows and fitted to compare the model output with the measurments. A window size of 100 measurement points achieved the lowest residuals and was able to map the system dynamics reliably.