Integrated multi-objective predictive control for multi-unit system

An integrated architecture comprising of a state estimator, dynamic optimizer and model predictive control (MPC) is designed in this work for an output feedback multi-objective control of a process system involving multiple units. The output feedback control uses a multi-rate extended Kalman Filter (EKF) for state estimation. Measurement delays in the arrival of the measurements of the infrequently sampled primary process variable are fused using a computationally efficient sampled-state augmentation approach. Certainty equivalence is assumed, and the state estimates are used by a dynamic multi-objective optimizer (D-MOO) followed by the coordinator MPC to implement feasible inputs to the plant. The trade-off between multiple objectives are handled by the D-MOO using the augmented -constraint method (AUGMECON) to generate the Pareto optimal solutions. This method computes efficient solution by incorporating slack variable in the optimization. The best solution among the Pareto optimal points is chosen close to the Utopian point. The significance of this algorithm in comparison with the conventional weight-based multi-objective control is discussed. The proposed algorithm is implemented on case study of a multi-unit system involving a series of two reactors followed by a separator.