Robust Predictive PI Controller Tuning

Abstract

Listen

Translate article

Two decades ago, a predictive PI controller was introduced for regulation of dead time dominated linear time-invariant first-order processes. The introduction involved a simple tuning rule with consideration on easy implementation and good performance. Later, the predictive PI controller has been improved by having an additional low-order filter in the controller structure against high-order modeling uncertainties. Also, several publications have been released with a detailed analysis on performance and robustness in particular with comparison of predictive and conventional PI(D) controllers. However, there have not been tuning rules with a design parameter for fine-tuning performance and dealing with robustness. In this paper, stability and robustness of a predictive PI without an additional filter for any time-invariant system is considered resulting in tuning rules for an optimal performance with a targeted robustness against model mismatch. Consequently, simple controller tuning rules are given for a first-order plus dead time and integrating plus dead time systems.