Advanced-multi-step Nonlinear Model Predictive Control

Abstract

Nonlinear Model Predictive Control (NMPC) has gained wide attention through the application of dynamic optimization. However, this approach is susceptible to computational delay, especially if the optimization problem cannot be solved within one sampling time. In this paper we propose an advanced-multi-step NMPC (amsNMPC) method based on nonlinear programming (NLP) and NLP sensitivity. This method includes two approaches: the serial approach and the parallel approach. These two approaches solve the background nonlinear programming (NLP) problem at different frequencies and update manipulated variables within each sampling time using NLP sensitivity. We present a continuous stirred tank reactor (CSTR) example to demonstrate the performance of amsNMPC and analyze the results.