Nonlinear Predictive Control Using Dynamic Integrated System Optimisation and Parameter Estimation

Abstract A nonlinear predictive controller based on state space models of the controlled plant is developed and implemented. Receding horizon long range prediction and dynamic optimization is carried out at every sampling instant by using an algorithm based on Dynamic Integrated System Optimisation and Parameter Estimation (DISOPE), a novel technique for solving optimal control problems subject to model-reality differences. States and parameters are estimated from possibly noisy output measurements by using an Extended Kalman Filter. The technique has been tested on a chemical reactor simulation example and its performance has been evaluated. The controller is flexible and is able to handle input bound and state dependent constraints.