A VARIABLE HORIZON PREDICTOR FOR CONTROLLING DEAD TIME DOMINANT PROCESSES, MULTIVARIABLE INTERACTIVE PROCESSES, AND PROCESSES WITH TIME VARIANT DYNAMICS

摘要

A Variable Horizon Predictor (VHP) compensates for dead time, measured disturbances, and time variant dynamics in process control loops. The VHP is comprised of a process model, an uncorrected prediction vector unit, a prediction correction unit, a corrected prediction vector unit, and a horizon selection unit (54). The VHP applies a controller output signal to the process model to form a prediction vector that is stored in the uncorrected prediction vector unit and predicts the process output signal over a prediction horizon. The prediction is corrected by the prediction correction unit, which compares the prediction vector to the process output signal. The corrected prediction vector is then stored in the corrected prediction vector unit. The horizon selection unit isolates a portion of the prediction vector to form a subvector and feeds the subvector to an error vector calculator to form an error signal that is applied to the controller. A multivariable interactive control system is implemented by coupling a plurality of VHPs and a plurality of loop interaction compensators to a process having a plurality of control inputs and a plurality of control outputs.