Unlocking feedforward capabilities in Model Predictive Control algorithms to deal with measurable disturbances

Disturbance rejection is a central objective in process control, especially when measurable disturbances can be exploited through feedforward action. Although Model Predictive Control (MPC) can include disturbance models, standard formulations do not generally achieve complete rejection because the cost function penalises control effort, preventing MPC from reproducing classical feedforward behaviour. This work proposes a framework that embeds true feedforward capabilities in MPC without removing the nominal control-effort penalty. A dual-control structure computes two actions simultaneously: a tracking-oriented action for set-point tracking and robustness, and a feedforward-oriented action for measurable disturbance rejection. Both are combined into a single constrained control signal, while the feedforward component is formulated without effort penalisation to enable full compensation. The methodology is developed for Dynamic Matrix Control (DMC), Generalised Predictive Control (GPC), and state-space MPC. Simulation studies, including a reverse osmosis case study, show improved disturbance rejection while preserving constraint handling and tracking performance.