Multivariable adaptive nonlinear predictive control: Development and application to a distillation column.

摘要

Model-based predictive control (MPC) strategies have enabled a revolutionary leap in process control in the past twenty years. Even so, despite many commercial successes, conventional MPC applications can show significant performance deterioration when the controlled process exhibits significant time-varying and nonlinear characteristics. In addition, the computer resources required to rigorously address such characteristics can be prohibitive.; This research presents a novel, computationally efficient formulation of MPC that can mitigate such control performance deterioration. The new control strategy incorporates elements from the research areas of adaptive and nonlinear control, as well as MPC.; The controller is evaluated experimentally for the dual composition control of a multi-component distillation column. The controller is shown to perform well over a wide range of process conditions. The controller is easy to tune by specifying a single tuning parameter value for each controlled variable to achieve performance/robustness tradeoffs. A novel design for processes with general time-delay structures is presented. In addition, the controller is found to be applicable to non-square multivariable processes that include integrators and non-minimum phase characteristics through extensions reported by other researchers.