Clearness of tuning possibilities in long-range predictive control (LRPC)

Abstract

It is shown that generalized predictive control (GPC) offers the possibility of properly controlling even unstable and nonminimum-phase plants with dead time. GPC can be adjusted so that robust behavior is achieved in order to cope with plants whose model order, dead time, and parameters are not well known or time-varying. GPC has appropriate tunings with respect to control performance and robustness. The design effort of GPC increases as a result of more difficult plants. To control unstable and nonminimum-phase processes, several control changes must be projected. The number depends on that of unstable and badly damped poles and zeros. If the plant is changing to such an extent that the output will leave behind the setpoint dynamics, GPC can be used in an adaptive scheme. The success of predictive controllers depends to a large extent on the quality of the process model. GPC offers a larger tuning variety, which is particularly useful in analyzing design methods together with concrete plants. >