Order Reduction of Large-Scale Dynamical Systems for Benchmark Problems

This paper aims to introduce a new model order reduction (MOR) method for simplifying the complexity of large-scale stable linear dynamical (LSSL) systems. The Balanced Truncation (BT) and Padé approximation methods are used to obtain the denominator and numerator polynomial coefficients of the reduced-order model (ROM). This method is used to address the shortcomings of the Padé approximation and BT methods. In this procedure, the stability and steady-state value of the LSSL system are guaranteed to be well preserved in the reduced-order model. The proposed technique has been applied successfully to the SISO system and has been extended to multi-dimensional systems. The proposed technique is confirmed by applying it to benchmark examples of 1006th and 120th orders. The results are compared to other well-known methods as well as recent work on performance indices and time domain specifications.