A multi-objective dynamic RTO for plant-wide control

Abstract

This work addresses a dynamic multi-objective control problem that achieves the desired product quality while ensuring the best profitable operation. A two-layered architecture is presented: The upper layer dynamic real-time optimizer (D-RTO), designed to handle multiple objectives, determines the best Pareto optimal trajectories for the lower level MPC. The proposed optimizer computes different trade-off solutions between the objectives using lexicographic approach. Here, the quality objective is formulated as a set point tracking problem whereas the economic objective is a nonlinear cost function. The trade-off solutions, which correspond to the Pareto points in the function space, are obtained by varying the rate of convergence of the quality variable to the desired set point. The Pareto point closest to the utopian or the ideal solution of the objectives is chosen to calculate the best optimal trajectory. The applicability of the proposed control framework is demonstrated on a reactor-separator system. Performance of this dynamic controller scheme is investigated by comparing with different control methodologies that compute optimal trajectories over a fixed, predefined horizon.