Research And Application On Continuous-time Nonlinear Model Predictive Control

According to the performance of industry process, the constraint and nonlinearity are so general that traditional control strategy could not offer satisfactory result. Constituting reliable optimal solution is a key issue for the constrained non-linear predictive control. In general, the nonlinear model predictive control(NMPC) also solves an on-line optimization problem at each sampling time, by using the sequential quadratic program(SQP). Sequential quadratic programming is known as one of the best algorithm in treatment of non-linear programming problem, which can get the optimal solution by transferring the initial nonlinear programming to a series of quadratic programming. The resulting nonlinear programming problems are generally nonconvex, and the on-line computational demand is high for any reasonably nontrivial systems. A general way to solve this problem is to use approximate optimization approach., in which the first control move is exactly calculated, which is actually implemented, and the rest of the control moves is approximate, which are not implemented. Input/output feedback linearization is a popular method in nonlinear control. In using MPC, for certain class of nonlinear systems, such transformation yields a linear dynamic system so that only a QP needs to be solved on-line in real time, while the original linear input constraints will change to non-linear and state-dependent constraints. Considering the state-space continuous-time system, this paper presents an iterative quadratic program(QP) routine to try to get the optimal solution. To guarantee its convergence, another iterative approach that can guarantee a feasible control solution over the complete prediction horizon is incorporated. Simulation results on both a numerical example and the CSTR demonstrate the effectiveness of the proposed method. As most industrial process are multi-input and multi-output systems, we extended this method to multivariable systems.