Abstract
Large scale Multi-Rotor Turbines (MRT) for electrical power generation introduce new challenges regarding the longevity of the support structure. The Vestas control challenge facilitate the development of advanced load reducing and power optimizing control methods for a MRT. This thesis aims to make a contribution of an important part of many control systems, namely the feedback. For state feedback control systems, precise estimates of the system's internal states are vital to achieve the control objective as efficiently as possible. In this thesis, a linearized and nonlinear Moving Horizon Estimator (MHE) is developed and tested for the MRT model provided by Vestas. The Uncsented Kalman Filter(UKF) and the Extended Kalman Filter(EKF) are used as benchmark estimators for MHE performance evaluation. The MHE and the Kalman variants are implemented in MATLAB using CasADi optimization toolbox.
Different measurement scenarios are tested in addition to a performance evaluation of the MHE variants with different estimation horizons. An evaluation of the real time applicability of the estimators are also performed. The wind velocities are included in the state space model of the system. All but one measurement model do not utilize wind speed measurements.
The performance of the estimators are evaluated for 3 different subgroups of system states, namely the support structure, the turbines and the wind speeds. The different estimators show different qualities, leading to no firm conclusion on what estimator is the superior for the MRT system. However, the best estimates of the individual subgroups of states are performed by the MHE versions. Out of the more realistic measurement models tested, one model is concluded to provide the highest performance.