Efficient Computation of Batch Reactor Control Profiles Under Parametric Uncertainty

Abstract

The paper outlines a method for the optimization of batch reactors when the models at hand are characterized by parametric uncertainty. A discrete (or discretized) probability distribution for the uncertain parameters is assumed. It results a differential/algebraic optimization problem (DAOP) including several model descriptions, each one corresponding to a grid point in parameter space. The DAOP is transformed to an algebraic optimization problem (AOP) using time parametrization based on the method of orthogonal collocation. This allows (i) to easily include additional algebraic path or endpoint constraints, and (ii) to use a simultaneous solution and optimization approach. Successive linear programming (SLP) is used for solving the large and sparse AOP. The proposed solution strategy is illustrated through a simulated example. The approach is computationally so effective that on-line optimization schemes become feasible.