Model Reduction and Dynamic Aggregation of Grid-Forming Inverter Networks

Olaoluwapo Ajala, Nathan Baeckeland, Brian Johnson, Sairaj Dhople, Alejandro Dominguez-Garcia

Research output: Contribution to journalArticlepeer-review

4 Scopus citations


This paper presents a model-order reduction and dynamic aggregation strategy for grid-forming inverter-based power networks. The reduced-order models preserve the network current dynamics as well as the action of the inverter current-reference limiter. Inverters based on droop, virtual synchronous machine, and dispatchable virtual oscillator control are considered, a generic model for all three control strategies is presented, and a smooth function approximation is utilized to represent the action of the current-reference limiter. The network is assumed to be composed of lines with homogeneous l/r ratios. Given such a system, our approach involves three steps. First, time-domain Kron reduction is used to reduce the dimensions of the electrical network model. Next, dynamic aggregate models are developed for parallel-connected inverters. Finally, singular perturbation analysis is used to systematically eliminate fast-varying dynamics in both the network model and the grid-forming inverter single/aggregate models. Numerical simulation results benchmark the response of the reduced-order aggregate models against the full-order models from which they are derived, and we demonstrate significant savings in computation cost with limited loss of accuracy.

Original languageEnglish (US)
Pages (from-to)5475-5490
Number of pages16
JournalIEEE Transactions on Power Systems
Issue number6
StatePublished - Nov 1 2023

Bibliographical note

Publisher Copyright:


  • Current limitation
  • dispatchable virtual oscillator control
  • droop control
  • dynamic aggregation
  • grid-forming control
  • reduced-order modeling
  • singular perturbation analysis
  • virtual synchronous machine


Dive into the research topics of 'Model Reduction and Dynamic Aggregation of Grid-Forming Inverter Networks'. Together they form a unique fingerprint.

Cite this