NONLINEAR OPTIMIZING CONTROL FOR CHEMICAL PROCESSES

The dual task of optimizing and regulatory control is examined by decomposing the problem into two separate problems. An optimizing controller can then be designed to address slow disturbances which take the process away from the most economically advantageous operating point, while a regulatory controller can be designed to attenuate fast disturbances which take a process away from specified setpoints.; The unconstrained optimization of a process at steady-state can induce dynamic system zeros on or near the origin under general circumstances. These zeros are particularly difficult to accommodate, and result in possible input multiplicities and lack of control system robustness. Therefore, limits are placed on the performance that may be expected from the implementation of linear controllers to processes that have been optimized at steady-state. A nonlinear control system is introduced which shows promise in alleviating the problems caused by linear system zeros.; Nonlinear integral control is developed in the form of a gradient system which possesses a finite region of asymptotic stability and guarantees the asymptotic convergence of the process outputs to the desired setpoints. Tuning rules are established for the single tuning parameter which determines the overall behavior of the closed-loop system. An example of a single-input, single-output system response to setpoint changes, disturbances, and modelling error illustrates the utility of the integral controller for a system that has been optimized at steady-state.; The idea of nonlinear integral control is extended to steady-state integral optimizing control, in which the optimization problem with the associated process constraints are defined in terms of a quadratic Liapunov function. The resulting gradient system shows asymptotic convergence to the optimal operating point. The formulation of the problem suggests a minimum number of integrators are used to determine the control variables. Tuning rules are developed in a qualitative way. An implementation of an integral optimizing controller to a CSTR with inequality constraints indicates that good control is achieved for cases of shifting constraints, start-up in the feasible region with and without modelling error, and the tracking of a time variable objective function.