| Summary: | 博士 === 國立成功大學 === 電機工程學系 === 107 === Several new methods have been proposed in this dissertation to design state observers (also called soft sensors) such that (i) the eigenvalues are specified to satisfy desired convergence performance or the state estimation errors quickly converge to zero, and (ii) simultaneously, a quadratic performance measurement of the deviation of estimates from the actual states is minimized for reducing large error during the transient period of observation. By combining the merits of both the orthogonal functions approach (OFA) and intelligent evolutionary optimization (IEO) approach, both the design issues of reduced-order observers for observable-form-based linear time-invariant dynamical systems and full-order observers for linear time-invariant state-delay systems have been studied. For observable-form-based dynamical systems, the proposed optimal reduced-order observer design method can be used to uniquely determine the observer gain matrix when the number of outputs is greater than the number of states to be estimated, whereas existing approaches fail to do so. For state-delay systems, the constraint of the linear-matrix-inequality-based condition makes the state estimation errors asymptotically converge to zero so that the steady-state errors are reduced. Furthermore, the proposed full-order observer design method with the quadratic performance measurement gives a penalty for the transient error to improve the transient error performance of state-delay systems. By solving a Sylvester equation and by fusing the OFA approach and the IEO approach, the design issue of the reduced-order observer for non-observable-form-based linear time-invariant dynamical systems has also been studied. The estimated states obtained from the designed non-observable-form-based-reduced-order observer can be directly used for state feedback control. These proposed new design methods can avoid the shortcomings of existing approaches in relevant literatures. From the demonstrative examples, it can be seen that the state estimation errors of the proposed approaches quickly converge to zero and their performance measurement values are apparently much better than those based on the existing non-optimal design methods.
|
|---|
