Variation of the balanced POD algorithm for model reduction of linear systems

Abstract

We present a variation on an existing model reduction algorithm for linear systems based on balanced proper orthogonal decomposition (POD). In contrast to many computational approaches to balanced truncation, the algorithm variation approximates the reduced order model directly without first transforming the linear system. The algorithm is applicable to large-scale finite dimensional systems and a class of infinite dimensional systems. The algorithm variation is compared to the original balanced POD algorithm on example partial differential equation systems.