Iterative nonlinear model predictive control. Stability, robustness and applications

Abstract: This paper describes the control of a batch pH reactor by a nonlinear predictive controller that improves performance by using data of past batches. The control strategy combines the feedback features of a nonlinear predictive controller with the learning capabilities of run-to-run control. The inclusion of real-time data collected during the on-going batch run in addition to those from the past runs make the control strategy capable not only of eliminating repeated errors but also of responding to new disturbances that occur during the run. The paper uses these ideas to devise an integrated controller that increases the capabilities of Nonlinear Model Predictive Control (nmpc) with batch-wise learning. This controller tries to improve existing strategies by the use of a nonlinear controller devised along the last-run trajectory as well as by the inclusion of filters. A comparison with a similar controller based upon a linear model is performed. Simulation results are presented in order to illustrate performance improvements that can be achieved by the new method over the conventional iterative controllers. Although the controller is designed for discrete-time systems, it can be applied to stable continuous plants after discretization. Copyright c©2005 IFAC