耦合网络相继故障及保护有效性研究

发布时间：2018-05-24 02:36

本文选题：相互依存网络 + 相互连接网络　；参考：《北京邮电大学》2015年硕士论文

【摘要】：相继故障是复杂网络研究领域中的热点问题之一,长期以来受到专家学者们的广泛关注。它研究的是网络少部分元素发生故障,对整体性能所造成的灾难性损害。对于这个问题的讨论有助于人们深入了解现实系统发生大规模崩溃时的内在机制,提高现代基础设施的安全可靠性等等。近些年来,随着科学与技术的迅猛发展,各类基础系统设施之间的依赖变得越来越强,耦合网络模型的提出加深了人们对现代复杂系统内在机制的认识,讨论耦合网络上的相继故障具有很现实的意义。考虑到耦合关系的多样性,相互依存网络(Interdependent Network)和相互连接网络(Interconnected Network)是耦合网络中常见的两类模型,本文分别针对这两类网络的相继故障特性和保护策略展开讨论,希望我们的研究结论能够为实际中网络系统的设计和灾害防护,提供一些可行性的建议。首先,我们注意到在大多数的前人研究中,网络的拓扑结构在相继故障过程中都发生了变化。对于相互连接网络,本文基于网络中节点负载(Load)动态分布的假设,提出了一种更贴合实际的相继故障模型,认为网络的结构不应该随着级联失效的传播而发生改变。通过与单层无标度网络(Scale-Free Network)的抗毁性能进行对比,我们发现相互连接的无标度网络具有较高的脆弱性。而且,当网络遭受蓄意攻击时,提高网络异质性会降低相互连接网络的鲁棒性。接下来在考察网络间的耦合偏好(Coupling Preference)与级联特性的关系时,仿真结果表明前者对后者没有产生明显的影响。最后,将模型应用到了北京市交通系统(公交网和地铁网相互耦合形成的相互连接网络),研究其抗毁性特征。得出系统受到蓄意攻击后,地铁网将会遭受更大的损害。而如果初始状态,公交网而非地铁网受到了蓄意攻击,这会使得系统发生更加严重的相继故障。上述结论可以为相互连接网络系统的设计以及最优化提供一些实际的参考建议。其次,对于相互依存网络,我们早前的工作曾讨论过有负载的相继故障模型的特性,得到有负载的互依存网络的脆弱性大大增加,而且随着网络异质性参数的变化呈现出一种“谷型”分布特征。由于有负载的相互依存网络呈现出较强的脆弱性,针对其提出有效的保护措施,具有非常重要的意义。首先,我们将在单网中得到广泛应用的无代价保护策略,直接引入到相互依存网络,发现大多数情况下这会导致网络发生更严重的相继故障。接下来,考虑到相互依存网络的依存关系特性,我们提出了一种改进型的无代价保护策略,并呈现出较好的保护效果,在某些网络参数下,可以使得网络鲁棒性提升3倍有余。对于不同异质性的网络,改进后的保护策略存在一定的容忍度有效保护范围。通过上述细致的分析讨论表明,在有效保护范围内,调整到合理的保护强度,可以使得网络具有最佳抵抗级联失效的能力。综上所述,本文提出了基于负载动态再分布的相互连接网络相继故障模型,并针对有负载的相互依存网络相继故障模型设计了一种无代价保护策略,有助于揭示如：通讯、电力、交通、Internet等现代基础设施系统级联失效的内在规律和特征,为保证系统的安全可靠运行提供了重要指导。同时,在日益复杂、多变、融合的外部环境下,针对异常行为导致的网络故障问题,本研究成果提供了新的理论与方法,并提出了具有较强实际意义的参考建议。
[Abstract]:Successive failures are one of the hot issues in the field of complex network research. It has been widely concerned by experts and scholars for a long time. It studies the disastrous damage to the overall performance of a few elements of the network. The discussion of this problem helps people to understand the large-scale collapse of the real system. Internal mechanism, improving the security and reliability of modern infrastructure, and so on. In recent years, with the rapid development of science and technology, the dependence between various basic systems and facilities becomes more and more strong. The proposed coupling network model has deepened people's understanding of the internal mechanism of modern complex systems, and discussed successive failures on the coupled network. Considering the diversity of the coupling relationship, the interdependent network (Interdependent Network) and the interconnected network (Interconnected Network) are two common types of models in the coupling network. This paper discusses the successive fault characteristics and protection strategies of the two types of networks. In the actual design of network system and disaster prevention, some feasible suggestions are provided.
First, we note that in most previous studies, the topology of the network has changed in the process of successive failures. For interconnected networks, based on the assumption of the dynamic distribution of node load (Load) in the network, this paper proposes a more practical sequential fault model, and considers that the structure of the network should not be cascaded with the cascade. The propagation of failure has changed. By comparing with the survivability of Scale-Free Network, we find that the interconnected scale-free networks have high vulnerability. Moreover, when the network suffers a deliberate attack, increasing network heterogeneity will reduce the robustness of the interconnected network. The simulation results show that the former has no obvious influence on the latter when the coupling preference (Coupling Preference) is connected with the cascade characteristics. Finally, the model is applied to the Beijing traffic system (the interconnected network coupled with the bus network and the subway network) to study its destruction characteristics. The subway network will suffer more damage. If the initial state, the bus network rather than the subway network is deliberately attacked, this will make the system more serious failure. The conclusion can provide some practical suggestions for the design and optimization of the interconnected network system.
Secondly, for the interdependent network, our earlier work has discussed the characteristics of the load sequential fault model, and the vulnerability of the load dependent interdependent networks is greatly increased, and with the changes of network heterogeneity parameters, a "valley type" distribution is presented. Frailty, it is of great significance to put forward effective protection measures. First, we will get a cost free protection strategy widely used in the single network and directly into the interdependent network. It is found that in most cases it will lead to more serious failure of the network. Next, we consider the dependency network dependency. We propose an improved non cost protection strategy and show a better protection effect. Under some network parameters, we can make the network robustness up to 3 times more. For different heterogeneous networks, the improved protection strategy has a certain tolerance range. It is shown that in the effective protection range, the reasonable protection strength can make the network have the best ability to resist cascading failures.
To sum up, this paper proposes a sequential fault model of interconnected networks based on load dynamic redistribution, and designs a non cost protection strategy for the sequential failure model of a load dependent interdependent network, which helps to reveal the inherent laws and characteristics of cascading failures of modern infrastructure systems, such as communication, electricity, traffic, and Internet. It provides an important guide for ensuring the safe and reliable operation of the system. At the same time, in the increasingly complex, changeable and integrated external environment, this research provides a new theory and method for the network fault caused by abnormal behavior, and puts forward some useful reference suggestions.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O157.5

【共引文献】