线性离散随机系统输出的多步预报法

Making Multi-step ahead Predictions for the Discrete-time Linear Stochastic System Output in a Successive One-step Manner

给出一种以逐次单步预报方式实现线性离散随机系统输出的多步预报的新方法,其基本思想是直接根据系统模型自身包含的一步递推预报关系,重复地加以利用,依次得出诸个多步预报值。文中证明了这种方法所得计算结果与Clarke提出的递推求解丢番方程的方法相同,但计算量可减少许多,且更便于在微型计算机中用汇编语言实现。

This paper gives a new method for making multi-step ahead predicti-ons of the discrete-time linear stochastic system output in a successive one-step manner The basic idea here is that these multi-step ahead predictions may simply be obtained one-by-one,repeatly using the systemmodel itself which just implies the recursive relationship in each one-stepahead predicting The paper proves tnat the new method leads to the same result as that of Clarke'S method (1984) employing a recursive solution to the Diophantine equations However,the new method is much easier for implementing via assembler language in microcomputer,and requires much less computational effort