An optimal control based plantwide control design methodology and its applications.

摘要

In recent years, a plantwide perspective on designing control systems for a multi-unit chemical process has received increasing attention from people in both academia and industry. The key design tasks involve how to identify controlled and manipulated variables and how to link them together through either centralized or decentralized control structures. In this dissertation, a hierarchical plantwide control design methodology, which takes advantage of a linear dynamic process model, is presented. This methodology aims to help an inexperienced control engineer solve practical plantwide control design problems in an efficient way.; Three major characteristics of this methodology are: (1) Information about process dynamics and interaction are extracted from the linearized model by calculating an optimal static output feedback (OSOF) controller for the desired manipulated variables and measurements. Then control structure candidates are determined based on the OSOF controller, a sensitivity matrix, relative gain array (RGA), and engineering judgment. (2) A hierarchical design procedure is developed for decomposing the whole design problem into several sub-problems, and the sub-problems are solved sequentially. (3) The design procedure is implemented as a highly automated computer aided tool in MATLAB. After a user provides a linearized process model and specifies design parameters, control structure candidates are automatically generated, proportional only controllers are automatically tuned, and control performance can be evaluated based on the transients that are generated from the linearized model.; The optimal control based plantwide control design methodology has been successfully applied to two plantwide control test problems, the Tennessee Eastman Challenge (TEC) process and the Vinyl Acetate Monomer (VA) process. In both case studies, some of the identified control structures are the same as the structures published by other researchers, which shows the effectiveness of this methodology. A first-principle nonlinear dynamic model of the VA process has been developed in MATLAB in order to make the VA process based plantwide control test problem available for public access. The MATLAB code has been translated into the C language, which is compiled to provide fast simulation speed.