Complex Graph Laplacian Regularizer for Inferencing Grid States

TL;DR

This paper introduces a graph signal processing algorithm that accurately infers complete grid states from limited measurements, improving stability and efficiency in power grid monitoring.

Contribution

The paper presents a novel GSP-based method that learns the grid graph empirically and interpolates missing data without prior grid knowledge, enabling fast and accurate state estimation.

Findings

Effective reconstruction of grid signals with fewer observations

No need for prior knowledge of grid structure

Demonstrated high accuracy on IEEE 118 bus system

Abstract

In order to maintain stable grid operations, system monitoring and control processes require the computation of grid states (e.g. voltage magnitude and angles) at high granularity. It is necessary to infer these grid states from measurements generated by a limited number of sensors like phasor measurement units (PMUs) that can be subjected to delays and losses due to channel artefacts, and/or adversarial attacks (e.g. denial of service, jamming, etc.). We propose a novel graph signal processing (GSP) based algorithm to interpolate states of the entire grid from observations of a small number of grid measurements. It is a two-stage process, where first an underlying Hermitian graph is learnt empirically from existing grid datasets. Then, the graph is used to interpolate missing grid signal samples in linear time. With our proposal, we can effectively reconstruct grid signals with significantly smaller number of observations when compared to existing traditional approaches (e.g. state estimation). In contrast to existing GSP approaches, we do not require knowledge of the underlying grid structure and parameters and are able to guarantee fast spectral optimization. We demonstrate the computational efficacy and accuracy of our proposal via practical studies conducted on the IEEE 118 bus system.