Constrained nonlinear predictive control using Volterra models

This paper presents a constrained model predictive control strategy that can be used to control nonlinear processes described by Volterra models. The controller exploits the special structure of the model to achieve an online feasible solution to the general optimization problem. Efficient solutions can be found, especially for second order Volterra models, that can solve the nonlinear problem by iteration of the linear solution, based on the particular structure of the model. This iterative procedure makes use of an analytical solution in the unconstrained case or a QP solution if constraints exist allowing an easy solution to the nonlinear problem. The complexity is therefore reduced compared to the computations needed to solve the general constrained non-linear optimization by a non-linear programming (NLP) technique. The procedure is tested on a simulated non-linear continuous stirred reactor and compared to other nonlinear MPC strategies both in complexity and computation time, showing that the proposed structure is a good candidate for a great class of industrial processes