Dept. of Electrical Engineering, Shahid Bahonar University of Kerman, Kerman, I. R. of Iran
In this paper, a new method is investigated for model order reduction of high order systems based on moment matching technique. In this method, at first, full order model is expanded by Legendre wavelet function which is included in orthogonal functions. A suitable fixed structure model is considered as reduced order model whose parameters are unknown. These unknown parameters are determined using Harmony Search (HS) algorithm by minimizing the errors between the l first coefficients of Legendre wavelet expansion of full and reduced systems. The Routh criterion is applied for specifying the stability conditions. Therefore, the stability condition is considered as constraints in optimization problem. To present the ability of the proposed method, four test systems are introduced. The obtained results are compared with other conventional techniques such as Balanced Truncation (BT) method and Hankel Singular Value (HSV) method. The obtained results show the proposed approach performs very well with respect to other reduction methods.