Frequency domain identification of passive local modules in linear dynamic networks

Abstract

We develop a novel frequency-domain approach to address the important open issue of estimating passive local modules within dynamic networks. The method applies an approach based on two stages, a non-parametric and a parametric one. The parametric stage is an extension of the vector fitting technique that incorporates energy consistency conditions as a fundamental component of the identification procedure, forming a path of the passive model in the Sanathanan–Koerner iterations. The approach includes a formulation via linear matrix inequalities to enforce energy-balance conditions resulting in a convex optimization problem. The approach is practical even under weak assumptions on noise, enabling real-world applications. Numerical simulations illustrate the potential of the developed method to effectively estimate local passive modules in dynamic networks.