Approximate aggregation of linear dynamic systems (Q1821064)

A two stage method for approximate aggregation of linear dynamic systems is proposed that involves selecting an aggregation matrix on the basis of the controllability and observability properties of the system, and then reducing the order using the chosen aggregation matrix. The system and aggregated model are given by \(\dot x=Ax+Bu\), \(z=Cx\) and \(\dot y=Fy+Gu\), \(y(0)=y_ 0\), \(\hat z=C L_1y\) respectively, where \(L_1\) is the chosen aggregation matrix and \(F\), \(G\), \(y_0\) are chosen to minimize \[ J=\int^T_0 [z(t)-\hat z(t)]^T [z(t)-\hat z(t)] \,dt. \] For given \(L_1\), explicit formulas are obtained for \(F\), \(G\) and \(y_0\).