Reference system-based model predictive control of nonlinear processes

Reference System Synthesis (RSS) is a model based control technique that has as its major advantage a transparent tuning methodology that is equally well applicable to linear as well as nonlinear system. A drawback of this technique is its inability to handle nonminimum phase processes and process constraints. A new strategy known as Reference System Model Predictive Control (RS-MPC) was developed that addresses this issue while still retaining the transparency of tuning of the original RSS approach. RS-MPC aims to fill a void in predictive control technology that is caused by the lack of intuitiveness of tuning of currently popular MPC schemes. Most available schemes tune the controller via selection of objective function weights. These weights penalize deviations of the outputs from setpoint and excessive movement in the inputs. No clear guidelines for the selection of the objective function weights exist in the literature. Further, a feel for the closed loop response is rarely possible at the design stage in techniques that use this weight based tuning. Because its tuning is based on a specification of the desired closed loop response, RS-MPC offers significant advantages, particularly in the context of nonlinear systems. It is observed that weight based tuning techniques fail in delivering a uniform response when the underlying process's dynamics are different from those used to tune the controller. RS-MPC on the other hand consistently delivers the asked performance without needing a retuning of the controller.; Presented herein is the theoretical development of the RS-MPC algorithm as well as a simulation based study of the above mentioned advantages. The RS-MPC algorithm is outlined in three different forms so as to enhance the understanding of it as well as to allow a comparison with existing predictive control strategies. The handling of nonminimum phase plants and the tuning issues are then explored in considerable detail. The remaining part of the thesis uses simulation examples to demonstrate the tuning advantages of RS-MPC in the context of SISO and MIMO, minimum phase and nonminimum phase, as well as linear and nonlinear processes. It is concluded that RS-MPC is an excellent alternative to conventional weight based tuning control algorithms.