RLS Estimation of Uncertainties in the Model Parameters and Decentralized Adaptive PI controller Using PSO for Chemical Multivariable Coupled System

Abstract

This paper proposes a decentralized adaptive PSO-PI (particle swarm optimization-Proportional Integral) control strategy for an uncertain coupled multivariable system using input/output data. The example used to be examined is that of a distillation column, in which the model is considered unknown as well as these uncertainties. First, we have developed a recursive least squares (RLS) with exponential forgetting factor to estimate the parameters of the nominal system and these uncertainties. Then, the design of the decentralized adaptive PSO-PI controller is developed by the inverted decoupling network combined with the PSO-PI controllers, and all updated through the estimated parameters using RLS to get the adaptive controllers. To automatically adjust the parameters of the PSO-PI controllers with robustness within the closed-loop, a tuning formula is developed based on the fitness function of the PSO technique. The obtained results in comparative form with IMC-PI controller demonstrate that the proposed approach can find better performance for estimating and controlling multivariable process with a nominal model and variations of gains for different setpoint changes and disturbance rejection.