GEOMAGNETIC DISTURBANCE VULNERABILITY STUDIES AND ARTIFICIAL INTELLIGENCE-BASED MONITORING FOR A REAL U.S. POWER GRID

The adverse impacts of geomagnetic-induced current (GIC) flow in the power grid during geomagnetic disturbance (GMD) events have been extensively documented. Demonstrated by notable occurrences such as the 1989 Hydro-Quebec blackout and the 2003 Swedish grid collapse, these impacts highlight the potential for severe disruptions. Consequently, electric utility companies, government regulatory agencies, and power system researchers have maintained a keen interest in this subject, particularly due to the escalating significance of the electric grid that is fueled by ongoing electrification initiatives across various industries. Moreover, heightened system stress and the integration of emerging technologies have intensified uncertainties surrounding the potential ramifications of GMDs.

In the context of extreme 1-in-100 year scenarios, the first portion of this dissertation conducted a thorough GMD vulnerability study of the Dominion Energy Virginia (DEV) grid that covers parts of Virginia and North Carolina. The study utilized load flow, harmonic, and thermal analyses to comprehensively assess grid security in these GMD scenarios. The findings indicate that the probability of immediate voltage collapse or equipment damage due to GMD is less than once in 100 years unless the GMD event coincides with multiple contingencies in the DEV grid.

The second portion of the dissertation involved a no-load GIC field test performed at a DEV substation to validate existing knowledge on transformer response to GIC. This test is the first in 20 years worldwide and the first-ever long-duration GIC field test in the U.S. This dissertation details the pre-test impact studies performed which deployed electromagnetic transient (EMT) simulations. The field test measurements enabled the development and validation of thermal and EMT models for the tested transformers, which facilitated the study of the transformer’s GIC response under loaded conditions. Furthermore, the validated models aided the investigation of grid-following and grid-forming inverter-based resources operations under GMD conditions; highlighting similarities and differences in their responses to GIC-induced harmonics and higher reactive power demand in the grid.

The last portion of the dissertation involved the advancement of effects-based GIC monitoring through the deployment of a convolutional neural network, surpassing previous attempts in literature with enhanced accuracy, sensitivity, and resolution.