Structure Identification of NDS with Descriptor Subsystems under Asynchronous, Non-Uniform, and Slow-Rate Sampling

TL;DR

A novel two-stage structure identification algorithm that leverages system zero-order moments, a concept traditionally used in model order reduction, to bridge system identification and model reduction and demonstrates its effectiveness in recovering subsystem interconnections from irregular sampling data.

Abstract

Networked dynamic systems (NDS) exhibit collective behavior shaped by subsystem dynamics and complex interconnections, yet identifying these interconnections remains challenging due to irregularities in sampled data, including asynchronous, non-uniform, and low-rate sampling. This paper proposes a novel two-stage structure identification algorithm that leverages system zero-order moments, a concept traditionally used in model order reduction, to bridge system identification and model reduction. First, zero-order moments are estimated from steady-state time-domain outputs; second, subsystem interconnections are explicitly reconstructed from these moments. The method generalizes existing approaches by handling asynchronous, non-uniform, and slow sampling simultaneously, eliminating constraints on input signal periodicity and extending applicability to multi-input multi-output NDS with arbitrary interconnections. Unlike black-box identification techniques, our approach explicitly recovers subsystem interconnection structures. Validation on the IEEE 14-bus system demonstrates the algorithm's effectiveness in recovering subsystem interconnections from irregular sampling data.

Figures (3)

• Figure 1: IEEE 14-bus power system
• Figure 2: Norm of Relative Estimation Error of $\bar{\eta}$
• Figure 3: Norm of Relative Estimation Error of $\theta$