Spatiotemporal Vulnerability Analysis of Large-Scale Infrastructure Systems under Cascading Failures: Case of Water Distribution Networks

Abstract

Large-scale water distribution networks (WDNs) are vulnerable to internal faults and external attacks. When exposed to various types of natural or artificial disasters, a glitch may trigger cascading failures and even paralyze the entire WDN. An improved load-based cascading failure model is proposed to analyze both structural and functional properties in a unified framework. The load–capacity relationship, load redistribution principle, and nodal breakdown probability are key factors needed to determine the evolvement of cascading failures. Different attack strategies that represent the scenarios of random failures, intermediate-level attacks, and catastrophic disasters are performed separately. The single-node attacks are repeated on each node to locate stricken components and identify weak points. Several novel metrics, such as node vulnerability, service level, and damage size, are adopted to characterize the spatiotemporal vulnerability of WDN. The simulation results illustrate that the traditional degree-based method inevitably results in the underestimation of potential losses in disaster scenarios. The improved load-based model that stresses network topology and physical significance can better reflect the reality. spatiotemporal vulnerability analysis is proven to be a powerful tool for optimizing system design and balancing the time–cost–reliability triad. Cascading failure is a ubiquitous phenomenon in network-based critical infrastructure, in which a glitch or an external disturbance can trigger large-scale disruptions and even paralyze the entire system. This study proposes a load-based model that characterize the interplay between hydraulic components in interconnected and interdependent water distribution networks. For different disaster scenarios with corresponding attack intensities, such as internal faults (e.g., leaking pipes) or intentional attacks (e.g., pump station explosions), the cascading failure model investigates the probability of the nodal breakdown and reveals the intrinsic mechanism that explains why some components may fail whereas some others may not. Specific paths for failure diffusion can be identified by probing the inherent load–capacity relationships, and eventually the damaged area can be predicted before a disaster occurs. Such simulations of potential disasters can be performed in advance to locate the weak points and risky parts, and maintain the operability and reliability. The analysis of spatiotemporal vulnerability is instructive for comprehensively improving the robustness and resilience with effective protective measures. The modeling approach can be extended into other typical critical infrastructure by substituting the definition of nodal load and relevant operation regimes.