An Algorithm for Model and Controller Reduction to Retain Desired States

This paper presents an algorithm for model reduction of linear large-scale systems. An objective of reduced order realization is that outputs of the reduced model will approximate the output response of the system, and states of the reduced model will look-like the desired state variables of the true model. The proposed method is based on the balanced realization concept of systems theory. Design of reduced order dynamic controller is also considered in a similar framework.