Model Predictive Control with Direct Feedthrough with Application on a Mist Reactor

Abstract

Theoretical derivation of a controller formulation for systems with direct feedthrough from the input to the output is considered. The resulting model predictive control (MPC) scheme requires the use of a predictor for the output. Two forms of predictor are compared. This work opens up the possibility of applying MPCs to practical systems with direct feedthrough. The controller is applied to a multivariable mist reactor for cell culture. A linear parameter-varying (LPV) model is applied with the controller. This enables the controller to have low computational complexity and high accuracy in controlling the temperature and humidity in the reactor. With the successful implementation of the LPV model to the controller, the proposed MPC scheme with direct feedthrough is shown to be feasible for application in practical systems.