GENERALIZED SYSTEM IDENTIFICATION FOR NONLINEAR MODEL PREDICTIVE CONTROL

Abstract

This work articulates an overarching and integrated System Identification procedure for systematically deriving black-box nonlinear continuous-time and discrete-time multiple-input multiple-output system models for Nonlinear Model Predictive Control. This framework successfully identified both continuous-time and discrete-time black-box internal models for a highly nonlinear Continuous Stirred Tank Reactor system that enabled Nonlinear Model Predictive Controllers to achieve effective control in both servo and regulator problems across wider operating ranges, even with ~1% output measurement error. These controllers also had reasonable per-iteration times of ~0.1 seconds and ~1 second with the continuous- and discrete- time models, respectively. By demonstrating how such system models could be identified for Nonlinear Model Predictive Control without prior knowledge of system dynamics, this opens further possibilities for direct data-driven methodologies for model-based control which, in the face of process uncertainties or modeling limitations, allow rapid and stable control over a wider operating range.