Nonlinear Min-Max Model Predictive Control based on Volterra models. Application to a pilot plant

This paper presents a new Nonlinear Min-Max Model Predictive Control strategy based on Volterra models. This control strategy is computationally efficient as the exact worst case cost can be computed in polynomial time. The reduced complexity of the proposed strategy allows its use in real time applications with typical prediction and control horizons. The controller has been implemented to control the temperature of a chemical reaction in the reactor of a pilot plant. A non-autoregressive second order Volterra series model has been identified from experimental data and used as a prediction model. The controller behavior is illustrated by experimental results.