Synthesis of plantwide control systems using a dynamic model and optimization

An optimization-based approach to synthesizing plantwide control architectures is discussed. The approach assumes that a dynamic model is available, and it extends earlier work. The plantwide controller is synthesized in three stages involving fast and slow safety variables to be controlled, followed by product variables. In each stage a mixed integer linear program is solved to generate candidate architectures. The objective function involves a tradeoff between manipulated variable moves and transient response area. The integers in the formulation determine the control architecture, and integer cuts are employed to generate second best, third best, etc., candidate solutions. Controlling component balances and adding unit operation controls completes the plantwide control system design. The Tennessee Eastman process is used to illustrate the synthesis procedure.