(214a) Passivity-Based Plantwide Control Design by Flowsheet Decomposition

Design of control structures for chemical plants typically occurs at the unit operations level.   Controllers can be designed for desired closed loop stability and performance properties for each unit.  The units, when placed in series retain these properties, but the presence of recycle streams changes the dynamics of the entire system.  Control of recycle streams has been studied in detail throughout the control literature. However, these studies usually do not include global stability proofs for integrated systems which may have many nested recycles.  Furthermore, current methods are useful for analysis but can't be easily used to construct plantwide control systems without analyzing the dynamics of the entire system.

In this paper, we utilize passive systems theory to develop a constructive procedure to design control laws for complex networks of chemical process systems. We use a natural storage function for passivity design that is derived from the second law of thermodynamics.  The method has many similarities with the tangent plane method developed by Gibbs to show stability of multiphase and reactive equilibrium systems.  We show that the storage function can be generalized to non-equilibrium systems and that chemical process systems are made passive by controlling one extensive variable for each independent phase and extent of reaction.  The control laws are implemented locally and individual units become passive.  If all units in a chemical processing plant are designed to be locally passive, the entire network will be passive and retain desirable properties like bounded input-output stability and Lyapunov stability, with the proper choice of inter-connection variables.  

We present a method to construct process control systems for large scale and complex networks of chemical processes with a minimum amount of process information.  We utilize extensive variable inventory control and identify the minimum number of controlled inventories to assure passivity of the entire network.  The remaining degrees of freedom are available for flow control and process optimization. The paper has a theoretical and a practical component.  New results on passivity of complex networks and graph-decomposition of process flowsheets are presented. The chemical looping combustion process is used as an illustrative example for the procedure.