Integrated vulnerability and dependency assessment in natural gas infrastructure and power systems: A review of traditional and emerging algorithms

Abstract

Energy pipeline networks are critical components of power systems, enabling the efficient transportation of energy resources, and are vital to societal and economic stability. However, natural gas pipelines are vulnerable to various factors, including aging infrastructure, natural disasters, accidents, and cyberattacks. In addition, these vulnerabilities can be caused by infrastructure capacity limits, which can disrupt natural gas supply and lead to electrical outages with significant societal and economic impacts. Traditional methods like Fault Tree Analysis, Event Tree Analysis, and Bayesian Networks provide valuable insights but often fail to address the complex interdependencies and dynamic risks in these systems. Recent advances in machine learning (ML), particularly Graph Neural Networks (GNNs), offer promising capabilities for modeling complex infrastructure interdependencies. This paper provides a comprehensive review comparing traditional methods, non-graph ML approaches, and graph-based techniques across vulnerability and dependency assessment applications in integrated gas-electricity systems. Challenges such as data quality, computational demands, and interpretability persist when dealing with large, complex networks, regardless of the methods employed. To provide quantitative evidence supporting these comparative findings, a controlled benchmark experiment is conducted on the IEEE 24-bus and IEEE 118-bus standard test cases, evaluating nine representative methods under a unified protocol. Future research directions include enhancing data collection, integrating hybrid models, and addressing domain-specific complexities to advance intelligent management and sustainable development of energy systems.