42 Uwe Fechner Delft University of Technology Kluyverweg 1 2629 HS Delft Netherlands u.fechner@tudelft.nl www.kitepower.eu
Dynamic Kite-Power System Modelling
Uwe Fechner, Rolf van der Vlugt, Roland Schmehl Delft University of Technology, The Netherlands
Kite-power systems are a promising approach to reach high capacity factors even with small to medium on shore wind en-ergy systems. They need an accurate control system for the trajectory of the kite. For the development, optimization and comparison of these control systems a fast system model is needed, that is suffciently accurate to cover the main dynam-ic behaviour of all system components: The kite, the tether and the ground-station. To the best knowledge of the au-thors, no system model was published until now that covers the dynamics of all three system components. In the past in most cases a straight tether was assumed. The straight line assumption does not provide enough accuracy for high-per-formance controller design, especially for the reel-in phase. Additionally, previous models were not suffcient to devel-op controllers for automated launch and landing. We imple-mented a system model as implicit problem in Python, using different DAE solvers from the Assimulo suite. The basic mod-el structure is a particle simulation build of point-masses and spring-damper elements. To achieve realtime performance
we used the LLVM (low level virtual machine) based Numba just-in-time compiler [1]. We verfied the model in two ways: First manually by connecting two joysticks as input (for the kite and the reel-out speed of the ground-station), and ad-ditionally by connecting the same automatic control system that we already used in field tests. We compared the results to the measurement data. The simulation matched the meas-urements well. The new modelling approach enables the de-velopment of improved, optimized control systems. Realistic software-in-the-loop testing of kite-power system control-lers becomes possible. Compared to previous approaches it is easy to change model components like the ground-station or the kite.
References
[1] T. Oliphant, J. Riehl, S. K. Lam, H. Grecco, and M. Florisson. Numba: A Dynamic Python Compiler for Science. http://numba.pydata.org/, 2013. 1
