Determination of Order Reduction in Large-Scale Linear Dynamical by Using Improved Balancing Technique

Abstract

For the simplification of large-scale linear dynamical systems, a new model order reduction scheme is proposed. The proposed technique is based on Balanced Singular Perturbation Approximation (BSPA) approach. This scheme introduces a hybrid strategy that combines the Singular Perturbation Approximation (SPA) and the Balanced Truncation (BT) Method with implicit dominancy criteria incorporated, leading to a much-improved approximation approach, which the steady-state value from the problem of the BT has been circumvented. The benefit of the proposed scheme lies in matching steady-state and its applicability to the large-scale dynamical system, which has been confirmed by a test system taken from literature. And also, compared to the responses of other existing reduction methods, this method gives the lowest performance error indices and H-infinity norm to ensure its exact approximation.