Real-time PI controller tuning via unfalsified control

Abstract

In this paper, we present a new unfalsified adaptive control algorithm. This algorithm leads to a real-time controller tuning method. The algorithm consists of two main elements: 1) Switching of controllers in a controller set by the e -hysteresis switching algorithm and 2) Optimization of the controller set via an evolutionary algorithm (EA). The real-time controller tuning is demonstrated for a nonminimum-phase continuous stirred tank reactor (CSTR) model.