Doctoral Dissertations

Date of Award

5-1990

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Nuclear Engineering

Major Professor

Rafael B. Perez

Committee Members

L. F. Miller, Belle R. Upadhyaya, J. M. Bailey, Pedro J. Otaduy

Abstract

In this dissertation a methodology for large-scale system decomposition is developed. The decomposition process involves dividing the large-scale system into a number of smaller subsystems for control purposes and creates the problem of how to treat the interactions among subsystems. The technique presented here follows a development similar to optimal control theory and results in expressions for the control action and the interactions among subsystems. First, the control system designer is responsible for determining how the system is to be divided. The state equations for the subsystems are written, and in each equation all terms involving interactions of a subsystems with other subsystems are eliminated and replaced by a single interaction term. Then, following optimal control theory, a cost function is defined which minimizes the difference between the desired state of the subsystem and the state of the internal controller model. A second cost function is defined which is used to ascertain the interactions between subsystems. This cost function is designed to minimize the difference between the controller model of the subsystem by adjusting the interaction terms in the state equations. The cost functions chosen for determining the control and interaction variables also allow the resulting two-point boundary value problem to be transformed into an initial value problem. This is accomplished by choosing the cost functions so that the initial values of the adjoint equations can be determined. The decomposition and control technique is demonstrated by developing seven subsystem controllers for the Advanced Liquid Metal Reactor, one of the advanced nuclear reactors designs being sponsored by the U.S. Department of Energy. This reactor has a total of nine nuclear reactors and three turbine generators, arranged in groups of three reactors per turbine. A testing environment for the controllers is developed which allows the controllers to be tested in a distributed manner as they would be implemented in a real plant. The testing environment consists of computer programs that allow reactor simulation on a parallel computer to communicate with each of the controllers, which are executing on a network of Sun workstations.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS