From linear to nonlinear model predictive control: Comparison of different algorithms

Performances of predictive control strategies which make use of different forms of linearized models are compared. The linear models range from a simple fixed model, linearized around a reference steady state, to a locally linearized model, which is updated at every sampling time by making use of a full first-principle nonlinear model. An intermediate choice is represented by the use of a weighted sum of different local models. Not only the process nonlinearity but also its directionality and the process-model mismatch affect the differences in performance of the controllers. All these aspects therefore need to be taken into account in order to determine which is the control algorithm that is more advantageous to implement. A terpolymerization reactor is used as a test process.