On model reduction of large-scale systems

The model reduction problem for time-invariant linear systems is considered. The solution of the problem involves the approximation of a matrix. An approximation method proposed by Aoki (1968) requires for an evaluation of the aggregation error a knowledge of the solution of the original system. A description is presented of a different approach for approximating the matrix. The result obtained is a simple expression which can be used for aggregation error assessment.