Power Network Component Vulnerability Analysis: A Machine Learning Approach

Abstract

The U.S. has defined a number of critical infrastructures, the disruption of which “would have a debilitating impact on security, national economic security, national public health or safety, or any combination of those matters.” Among these critical infrastructures is the electric power network that has a crucial role in enabling the operation of societies and industries. In the past decades, the functionality of the power network has been vulnerable to numerous disruptive events, including natural hazards, human-made events, or common failures. This work leverages several publicly available big data sets to lay the foundation for a comprehensive characterization and analysis of the U.S. power network in order to propose a network component vulnerability measure adopting machine learning techniques. The non-linear machine learning model is implemented to create smarter component cataloging for vulnerability analysis based on its geographic location and criticality. The findings could be useful by the grid stakeholder and policymakers to (i) evaluate network stability, (ii) understand the risk of cascading failure, and (iii) improve the resilience of the overall network and moving toward resilient smart grids.