Multi-step Ahead Prediction of the Multivariable Linear Discrete Stochastic System Output in Succesive One-step Manner

This paper gives a new multi-step ahead prediction method for the multivariable linear discrete stochastic system output. The multi-step predictions are obtained by succesively employing the one-step ahead prediction relationship included in the system model itself. This new method is equivalent to the existing multivariable prediction method via recursion of diophantine equations. It possesses the advatages both in greatly reducing computational effort and simplifying programming, and has been successfully used in the generalized predictive control.