Model-order reduction using truncated modal balanced realization

In this paper, a modal reduction approach based on the truncated balanced realization technique is proposed to compute reduced-order system models (ROMs). The method enables efficient computation of Gramian covariance matrices in terms of the system frequency response and can be used for control design. Methods for approximating the high-dimensional system dynamics with a low-dimensional model are examined and compared, and the prediction error of the ROMs is evaluated. The proposed technique is tested on a 46-machine dynamic equivalent of the Mexican interconnected system. The issues of accuracy, applicability, and computational effort are discussed.