Decentralzied Controller Design For Multivariable System

Most of the controlled process in the industrial field are multi-input multi-output process.The developed single loop control method cannot be directly used to control the multivariable process,due to the interactions widely existing between the variables.At present,the multivariable system control can generally be classified into two main groups:decoupling control and decentralized control.Decoupling control can provide better performance for those process that are closely coupled but have more complex control structure because of the decoupler need to be designed.Especially for the multivariable process of high dimension,it is difficult to implement in practice.decentralized control is widely employed in industrial process because it has many advantages,such as simple structure,simplified tuning and good robustness.The premise of decentralized control is variable pairing which requires the coupling characteristics of multivariable process are analyzed.In this article,the best loop pairing is obtained using the RNGA(relative normalized gain array),RLG(relative load gain)is employed to analyze the couple effect on the disturbance rejection of multivariable process.A new modified structure of decentralized control is proposed to to counteract the disadvantageous interactions by adding some controllers in the decentralized control array.A decentralized controller design method was proposed for the loop paired multivariable process based RNGA(relative normalized gain array).The SOPDT(second order plus dead time)or FOPDT(first order plus dead time)model of main loop's EOTF(effective open-loop transfer function)is obtained by Maclaurin series.The Internal model controllers were designed independently for the main loops based EOTF.A set-point filter is designed to reduce the over-shoot of closed loop response to a desired value.Simulation studies demonstrate the proposed design method can provide better performance comparing with some existing decentralized controller design method.