Incremental nonlinear dynamic data reconciliation

Der, Ralf
Lummer, Olaf
List, Thomas
Measurement noise reduction and parameter estimation is a topic of central importance in plant control. The complexity of real world plants and the working conditions in practice require robust real{time algorithms which are easy to implement, simple to use and economic in computer ressources. The state of the art is given by the novel approach of Liebman et al. called the NDDR (nonlinear dynamic data reconciliation) which is based on nonlinear dynamic programming. We present in the present paper a new algorithm based more traditionally on gradient descent methods supplemented with a self control of the parameters of the algorithm. It uses an iterative method for the rectification and correction of state variables and system parameters, what makes it a true on{line algorithm. Despite its simplicity, the perfomance of the new algorithm proved superior to that of the NDDR in the applications considered so far.
Appeared / Erschienen in: 
Report Nr.3/1998 des Instituts fuer Informatik der Universitaet Leipzig
Pubdate / Erscheinungsdatum: 
Pages / Seitenanzahl: 
1998-15.pdf395.08 KB