Assessment and Enhancement of Power Grid Resilience to Shocks and Stresses

Abstract

The rising impact of large-scale power outages on economy and society in the last decades has drawn attention to an emerging need for power system resilience. Power outages, for example due to cascading failures and triggered by extreme weather, have become more frequent and severe, and extreme weather events are likely to occur more often in the future because of climate change. Power networks are also put under additional stress due to many strategies that aim to reduce carbon emissions, for instance electrification of heat and transport, whilst increasing dependency on electricity supply. The aims of this thesis are to assess and enhance power network resilience to shocks and stresses. As such, the thesis contributes to a number of areas in power systems research, particularly addressing the understanding and prevention of cascading failures, which are caused by extreme weather, through probabilistic and risk-based decision strategies. Simulation-based and data-driven methods are combined to present a comprehensive and holistic approach to power network resilience. Two novel cascading failure models specifically designed for resilience analyses are developed, validated, and applied to investigate cascade propagation for a number of scenarios. Whilst cascade acceleration has been observed in extreme cascades in the past, detailed analyses of cascading failures using historical utility outage data show a cascade acceleration in less extreme and more common cascades. Building on the cascading failure models, preventive actions, and in particular intentional islanding, are shown to be operational methods that can reduce the risk of cascading failures, reduce lost load, and improve power network resilience. The decision to apply preventive actions must be based on a risk assessment and be made dependent on the specific circumstances of the event. To address this, probabilistic approaches, that consider uncertainty inherent in the impact of extreme events on power networks, are assessed and optimised. Subsequently, two decision-making frameworks for identifying preventive actions, that are likely to perform best considering uncertainty about upcoming extreme events, are developed and evaluated. As computational complexity of decision-making remains an issue, a novel real-time decision-making tool is developed using machine learning. The tool is able to accurately and efficiently identify the most suitable preventive action for a given extreme event, based only on event parameters that are readily available ahead of the event. Finally, a data-driven methodology for evaluating power network resilience based on utility outage data is proposed and applied. A novel and comprehensive data set of power outages is created, allowing for in-depth analysis of causes and implications of power outages in both spatial and temporal dimension, and supplementing simulation-driven approaches to power network resilience.

Date of Award	1 Aug 2022
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Ian Cotton (Supervisor) & Robin Preece (Supervisor)