Nontrivial Kron Reduction for Power Grid Dynamics Modeling

TL;DR

This paper investigates the limitations of Kron reduction in power grid modeling, revealing that reduced nodes can significantly influence the entire system's dynamics and noise characteristics, and proposes a method to improve the reduction accuracy.

Contribution

It demonstrates the impact of reduced nodes on system dynamics and noise, and introduces a Mori-Zwanzig formalism-based approach to improve Kron reduction accuracy.

Findings

Kron reduction can significantly affect non-reduced node dynamics.

Noise in reduced nodes can influence the entire system unexpectedly.

A Mori-Zwanzig formalism improves the modeling accuracy of reduced power grids.

Abstract

The Kron reduction is used in power grid modeling when the analysis can -- supposedly -- be restricted to a subset of nodes. Typically, when one is interested in the phases' dynamics, it is common to reduce the load buses and focus on the generators' behavior. The rationale behind this reduction is that voltage phases at load buses adapt quickly to their neighbors' phases and, at the timescale of generators, they have virtually no dynamics. We show that the dynamics of the Kron-reduced part of a network can have a significant impact on the dynamics of the non-reduced buses. Therefore, Kron reduction should be used with care and, depending on the context, reduced nodes cannot be simply ignored. We demonstrate that the noise in the reduced part can unexpectedly affect the non-reduced part, even under the assumption that nodal disturbances are independent. Therefore, the common assumption that the noise in the non-reduced part is uncorrelated may lead to inaccurate assessments of the grid's behavior. To cope with such shortcomings of the Kron reduction, we show how to properly incorporate the contribution of the reduced buses into the reduced model using the Mori-Zwanzig formalism.