Analysis and Correction of Ill-Conditioned Model in Multivariable Model Predictive Control

摘要

The ill-conditioned model is a common problem in model predictive control. The model ill-conditioned can lead to control performance declining obviously from steady-state model of process in this paper. The direction of output movement is relevant to whether the model is ill-conditioned by simulation and analysis. Model mismatch also leads to model ill-conditioned becoming more serious. The geometry tools and SVD in linear algebra are used to analyze the essential reason of ill-conditioned model, and an offline strategy is proposed which can solve the ill-conditioned model problem together with existing online strategies. Finally, the simulations are used to prove the conclusions which presented in this paper are correct.