Premium
Comparative analysis of model reduction strategies for circuit based periodic control problems
Author(s) -
Hossain MohammadSahadet,
Tahsin Aniqa,
Omar Sufi Galib,
Hossain Khan Ekram
Publication year - 2021
Publication title -
asian journal of control
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.769
H-Index - 53
eISSN - 1934-6093
pISSN - 1561-8625
DOI - 10.1002/asjc.2312
Subject(s) - mathematics , linear system , lti system theory , model order reduction , krylov subspace , lyapunov function , observability , cholesky decomposition , discretization , control theory (sociology) , computer science , algorithm , mathematical analysis , projection (relational algebra) , nonlinear system , eigenvalues and eigenvectors , physics , quantum mechanics , control (management) , artificial intelligence
This paper is a comparative analysis of two prominent iterative algorithms for model order reduction of linear time‐varying (LTV) periodic systems where the system's matrices are singular. Our proposed method is based on a reformulation of the LTV model to an equivalent linear time‐invariant (LTI) model using a suitable discretization procedure. The resulting LTI model is reduced in two ways, once by applying a balanced truncation method and once by applying a Krylov‐based method known as iterative rational Krylov algorithm (IRKA). During the application of balanced truncation, the low‐rank Cholesky factorized alternating directions implicit (LRCF‐ADI) method is used to estimate the solutions of the corresponding LTI form of Lyapunov equations. Since the system's matrices are singular, the concept of pseudo‐inverse is adopted to compute the shift parameters needed in the LRCF‐ADI iterations. For the Krylov‐based IRKA, our work is twofold. We solve the time‐invariant Lyapunov equation for the observability Gramian and apply a moment‐matching Krylov technique. The accuracy and effectiveness of the two proposed techniques are demonstrated with the help of frequency response graphs, bode plots, and eigenstructure of the main and reduced models.