Design of decentralized multivariable PI controllers using the particle swarm optimization algorithm

This study introduces a technique for the design of multivariable decentralized proportional-integral controllers (Pl). This problem is addressed by first analyzing the pairing possibilities of the manipulated and controlled variables, which result in minor interaction between the loops; controller tuning is then performed by using the particle swarm optimization algorithm. The control objective is to design the multivariable controller for the several viable pairing possibilities, minimizing the error between the controller outputs and the set-point variables. Simulation results presented allow the conclusion that the proposed technique can tune the controller with all the loops closed, reducing the interaction effect. It is also shown that for this problem the particle swarm algorithm is faster in converging than a genetic algorithm.