Content area
Abstract
Many control and plant monitoring techniques require the states of the plant to be available, while in most practical solutions it is not possible to measure all the states of the system. There is a need for state estimators to be robust to modelling uncertainty and able to accommodate known nonlinearities. The variable structure observer, a type of Luenberger observer, has been shown to have robustness to modelling uncertainty and is able to accommodate model nonlinearities.
The main concern of this thesis is to advance the existing knowledge on variable structure observers, and to improve their performance by incorporating noise rejection and the consideration of specific types of modelling uncertainty. The research work begins with the development of a novel variable structure observer for uncertain nonlinear systems. The main feature of this technique is that it has been designed so as to achieve stability and convergence in the presence of modelling uncertainties. The development of the variable structure observer for uncertain nonlinear systems leads to the development of nonlinear variable structure observer with unknown but bounded measurement noise. The technique has been developed so as to achieve stability and convergence in the presence of unknown but bounded measurement noise.
A third novel nonlinear state estimator that has the noise rejection property of nonlinear stochastic state estimators and the robustness property of the variable structure observers is developed, by combining the unscented Kalman filter and a nonlinear variable structure observer for uncertain nonlinear systems. The main feature of the proposed technique is that it has been designed so as to achieve stability and convergence despite the presence of modelling uncertainties and measurement noise.
The state estimators developed in this thesis have been implemented in software and their performances have been evaluated with relevant simulation studies.
