Title: NMPC and MHE on Power Generating Kites
Authors: Andreas Ilzhoefer and Moritz Diehl
(20-30 min)
Abstract: In the first part of this talk we show that Nonlinear Model Predictive
Control (NMPC) can be used to control a power generating kite under changing
wind conditions. We use a optimized trajectory as periodic reference for our
NMPC problem. The NMPC algorithm minimizes the deviation of the trajectory from
this reference. Nominal stability is garantied through a linearly optimal
pre-controler, which was already presented in the talk from Boris Houska. With
this feedback control technique we are able to deal with bounds for states and
controls. For example we introduce bounds that keep the kite on a "safe"
altitude. We show that the kite produces even under strong random wind
disturbances almost the same amount of energy as a kite flying on the optimized
reference with constant wind. By using the so called "real-time iteration
scheme" we achieve that this algorithm is capable of real-time use.
In the second part of this talk we use Moving Horizon Estimation (MHE) to estimate the wind velocity and the wind direction. Only the position und some other observable states of the kite are measured, while the uncertain fast changing wind is estimated. We can show again the real-time capability of this technique. The results with NMPC and MHE are surprisingly good. Under strong dryden wind disturbances we can almost harvest the same amount of energy as we could with constant know wind velocity.