Back to search

National Technical University of Athens (NTUA)

Ανάπτυξη μεθόδων σχεδιασμού ρυθμιστών προβλεπτικού μοντέλου για χρονικά μεταβαλλόμενα συστήματα και συστήματα κατανεμημένων παραμέτρων

Abstract

dc:description

The objective of this thesis is the development of novel Model Predictive Control (MPC) configurations for Time Varying and Distributed Parameter Systems (DPS). Regarding time varying systems, a new multiobjective MPC configuration is proposed, where the different control targets are allocated to different objective functions. Among the numerous control goals, the one of top priority is the satisfaction of the Persistent Excitation criterion, which ensures the adaptation of the process model according to the changes of the system dynamic behavior. In order to solve the non convex optimization problem that arises in the proposed MPC, a new stochastic multiobjective optimization algorithm was developed based on the principles of simulated annealing that also considers the different priority of the objective functions. Besides the implementation of the proposed algorithm in the control of a high nonlinear process, it was also successfully used in benchmark problems. Adaptive controllers were also applied for the first time in inventory management and production planning problems. The new adaptive proportional control scheme that was developed, as well as the adaptive MPC configuration, is able to confront a number of realistic cases such as changes of lead time and loss of products. The second part of this thesis refers to DPS, for which two new models based on neural networks are developed. The first model is a radial basis function neural network in which the input vector includes past values of the manipulated variables and the coordinates of the location where an estimation of the controlled variables is required. The second model combines the projection of the controlled variables onto basis functions with neural networks. The method of singular value decomposition is applied on the process snapshot matrix in order to obtain the discrete empirical eigenfunctions as well as a data base for the temporal coefficients that weigh these basis functions. These coefficient values are correlated to the manipulated variables through a neural network. The two modeling methodologies were combined with MPC methodologies for a number of problems found in literature, and their performance was proved superior in comparison to existing control methods. The same models were also successfully used by Η? controllers.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2007

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Αγγελογιαννάκη, Ελένη
  • Aggelogiannaki, Eleni

Subjects

dc:subject × 20

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/16193

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Αγγελογιαννάκη, Ελένη; Aggelogiannaki, Eleni. Ανάπτυξη μεθόδων σχεδιασμού ρυθμιστών προβλεπτικού μοντέλου για χρονικά μεταβαλλόμενα συστήματα και συστήματα κατανεμημένων παραμέτρων. National Technical University of Athens (NTUA), 2007. http://hdl.handle.net/10442/hedi/16193