Cascading failures in complex infrastructure systems
Cascading failures in complex infrastructure systems
复制标题
DOI：
10.1016/j.strusafe.2008.06.007
复制
发表时间：
2009-01-01
影响因子：
5.8
通讯作者：
Vemuru, Srivishnu Mohan
中科院分区：
文献类型：
--
作者：
Duenas-Osorio, Leonardo;Vemuru, Srivishnu Mohan
本文研究了连锁故障在复杂基础设施系统风险和可靠性评估中的作用。这些系统的传统可靠性评估仅限于寻找预定义组件之间的路径，并且不包括增加的流量需求或流量容量的影响。网络流与基于中断的中断相关联，这会使基于路径的性能预测恶化。在这项研究中，由于连锁故障的过载建模与公差参数，测量网络元件的流量容量相对于实际的电力传输系统中的流量需求。自然灾害和恶意的有针对性的破坏构成触发事件，由于流量重新分配，这些事件演变为广泛的故障。据观察，仅在网络组件的公差的改善并不能确保系统的鲁棒性或保护不成比例的级联故障。拓扑结构的变化，需要增加级联的鲁棒性在现实的公差水平。有趣的是，一小部分网络组件的有针对性的拓扑中断可能会比触发类似部分元件故障的地震或闪电事件更严重地影响系统级性能。此外，无论危害的性质如何，一旦发生中断被调查网络的触发事件，由于级联故障造成的额外性能损失可能比初始性能损失大几个数量级。这些结果强化了这样一个概念，即管理网络不可用的风险需要冗余拓扑结构，增加流量承载能力和其他非传统的后果减少策略的组合，如布局均匀化和故意包括网络孤岛的薄弱环节。此外，公认的想法，罕见的性能损失事件发生的频率呈指数下降的性能降低加剧与更频繁的网络漏洞，导致从最初的危险引起的故障和随后的故障引起的故障影响的对比。这些复合灾害级联有害影响可能对基础设施故障预防策略产生深远影响。(c)2008爱思唯尔有限公司保留所有权利。
This paper studies the effect of cascading failures in the risk and reliability assessment of complex infrastructure systems. Conventional reliability assessment for these systems is limited to finding paths between predefined components and does not include the effect of increased flow demand or flow capacity. Network flows are associated with congestion-based disruptions which can worsen path-based predictions of performance. In this research, overloads due to cascading failures are modeled with a tolerance parameter a that measures network element flow capacity relative to flow demands in practical power transmission systems. Natural hazards and malevolent targeted disruptions constitute the triggering events that evolve into widespread failures due to flow redistribution. It is observed that improvements in network component tolerance alone do not ensure system robustness or protection against disproportionate cascading failures. Topological changes are needed to increase cascading robustness at realistic tolerance levels. Interestingly, targeted topological disruptions of a small fraction of network components can affect system-level performance more severely than earthquake or lightning events that trigger similar fractions of element failure. Also, regardless of the nature of the hazards, once the triggering events that disrupt the networks under investigation occur, the additional loss of performance due to cascading failures can be orders of magnitude larger than the initial loss of performance. These results reinforce the notion that managing the risk of network unavailability requires a combination of redundant topology, increased flow carrying capacity, and other non-conventional consequence reduction strategies, Such as layout homogenization and the deliberate inclusion of weak links for network islanding. Furthermore, accepted ideas that rare loss of performance events occur exponentially less frequent as the performance reduction intensifies contrast with more frequent network vulnerabilities that result from initial hazard-induced failures and subsequent cascading-induced failure effects. These compound hazard-cascading detrimental effects can have profound implications on infrastructure failure prevention strategies. (c) 2008 Elsevier Ltd. All rights reserved.