In this paper multivariable generalizations of the SISO unified process model (UPCM) and unified prediction model (UPM) used in the SISO unified predictive controller are presented. the necessary assumptions regarding the multivariable UPCM (MUPCM) are established. the properties of the multivariable UPM (MUPM) are examined and the conditions for a zero steady state prediction error are established.
The MUPCM has a general noise model and unites all multivariable, discrete, linear process models in transfer function form. the MUPM uses an output filter which unites all relevant output filters used in predictive controllers. Furthermore, the elements of the output can be predicted over different horizons. the MUPCM and MUPM and the algorithm used to calculate the MUPM are chosen such that the predictions of the outputs can be calculated as fast as possible, not only for the general model, but also for special cases such as ARX, ARIMAX, Box-Jenkins and FIR models.
[1]
David W. Clarke,et al.
Generalized predictive control - Part I. The basic algorithm
,
1987,
Autom..
[2]
P. Falb,et al.
Invariants and Canonical Forms under Dynamic Compensation
,
1976
.
[3]
Lennart Ljung,et al.
System Identification: Theory for the User
,
1987
.
[4]
R. Guidorzi.
Canonical structures in the identification of multivariable systems
,
1975,
Autom..
[5]
Graham C. Goodwin,et al.
The role of the interactor matrix in multivariable stochastic adaptive control
,
1984,
Autom..
[6]
Michel Kinnaert,et al.
Adaptive generalized predictive controller for MIMO systems
,
1989
.