Guzmán et al. unveil dual-control MPC for full disturbance rejection

Standard Model Predictive Control (MPC) struggles to fully reject measurable disturbances because its cost function penalizes control effort—unlike classical feedforward compensators. This limitation has long prevented MPC from achieving ideal disturbance rejection. A team led by José Luis Guzmán at the University of Almería introduces a dual-control structure that solves this trade-off. The framework computes two separate control actions: a tracking-oriented action (penalized for effort) for set-point tracking and robustness, and a feedforward-oriented action (without control effort penalty) dedicated solely to disturbance rejection. Both actions are combined into a single control signal while explicitly enforcing process constraints.

The methodology is developed for three common MPC formulations: Dynamic Matrix Control (DMC), Generalized Predictive Control (GPC), and state-space MPC. Simulation studies, including a case study on reverse osmosis water desalination, demonstrate that the dual-control approach significantly improves rejection of measurable disturbances compared to standard MPC and classical feedforward schemes. The framework maintains constraint handling and overall control performance, making it ready for practical industrial deployment in chemical, energy, and water treatment processes.