Multivariable Controller Tuning for Non-square Systems with RHP Zeros by Genetic Algorithm

摘要

A simple genetic algorithm is applied by simulation for the tuning of centralized and decentralized controllers for linear non-square multivariable systems with RHP zeros. All the loops are first tuned independently using Ziegler-Nichols method based on closed loop oscillation. A detuning factor is used to obtain the range of controller parameters. This approximate range of the controller parameters is used to improve the search of genetic algorithm. This method is applied to a MIMO non-square coupled distillation column (3 inputs and 2 outputs) system. Both centralized and decentralized controllers are designed by using the genetic algorithm optimization method. The proposed centralized controller tuned by the genetic algorithm gives the best performance when compared to the decentralized controller and other analytical methods. These proposed ranges of controller parameters for genetic algorithm can make the convergence faster than the method proposed for multivariable square systems by Vlachos et al.