Study On Node Importance Ranking And Cascading Failure For Complex Networks

In the late 1990 s,the propose of the Small world network and the Scale-free network turned a new page for the development of complex network, and complex network has drawn more and more attention. As an important research field of complexity science, the research of complex network theory has evolved rapidly and has permeated every realm, such as mathematics and sciences, Life Science, engineering discipline and so on.With the rise and development of the Internet of things and the construct of smart city, many network systems have been constructed. And our social lives are relying more on these infrastructure network systems, such as power grid, communication network, Internet, airport network, logistics network and so on. These networks provide great convenience for our life. On the other hand, it tends to threaten the security of our daily lives, such as the rapid spread of the virus in computer network, large-scale blackout and so on. Most of these real network systems may be abstracted as the complex network which provides a new way to study the real network systems.From the analyze of catastrophic events in infrastructure networks, one can see that the key nodes and cascading failure have great influence on the function of infrastructure networks, hence aiming to decrease unwanted economic losses and avoid disaster events, in this paper I study node importance ranking and the cascading failure for complex networks. There are many literatures about these researches and some valuable conclusions are drawn, but there are still many problems to be solved. As a hot area of research, evaluating the importance of nodes has drawn many peoples' concern. The current methods based on global information, such as betweenness centrality, can evaluate the importance of nodes effectively, but their computational complexity is too high. Methods based on local information are of lower computational complexity, but the accuracy of the evaluation results needs to be improved. Similarly, there are also some problems to be solved about cascading failure though there have been many researches about cascading failure and some valuable conclusions are drawn. Most current cascading models didn't consider the time-varying characteristics of load, which may lead to the unreasonable redistribution of load and eventually expands the scale of cascading failure. In view of the above, considering the influence between node and line we would like to propose a new ranking method based on local information. Moreover, considering the time-varying characteristic of load, we proposed a new cascading model based on time-varying load redistribution strategy which can reduce the size of the cascading failure. In addition, we study the influence of breakdown probability of overload node on the scale of cascading failure considering the influence of human intervention on the robustness of network, and we propose the corresponding distribution strategy of protection resources. Furthermore, considering the interdependence between real network systems, we research the influence of coupled mode and dependent strength on the size of cascading failure of coupled network. The details are as follows:(1) Considering the interdependence and influence between nodes and lines, we proposed a new node importance ranking method based on the importance of line. The properties of nodes that are connected to a line are used to compute the importance of the line. Then, the influence of the line importance on the importance of node is calculated according to the contribution of nodes to the importance of lines. Finally, degree of nodes and the contribution of line importance to the importance of nodes are considered to rank the importance of nodes. As the evaluation index, the decline rate of network efficiency caused by deleting node is used to evaluate the accurate of ranking method. And the computational complexity is used to evaluate the efficiency and practicality of ranking method.(2) Considering the time-varying characteristics of load in real network systems, we proposed a time-varying load redistribution strategy based on real-time residual capacity. And based on the new load redistribution strategy we study the cascading model. A natural metric to quantify the ability of networks against cascading failure is the normalized proportion of failure nodes due to the removal of node. Then we study the influence of load redistribution strategy on the ability of network against cascading failure through exponential analysis, numerical simulation and instance simulation. Compared with the current model based on fixed load redistribution strategy, the new model can redistribute the load of failure node reasonably according to the real-time load-handling ability of its neighbors. So the new model based on time-varying load redistribution strategy has strong ability to resist disturbance and can improve the ability of networks against cascading failure when the load of node is time-varying.(3) Different from most of the current researches, this paper provides a new cascading model with an improved breakdown probability considering that the overload node may still work due to the existence of protection resources. Similarly, the normalized proportion of failure nodes due to the removal of node is chosen as the evaluation index to evaluate the ability of network against cascading failure. And we study the influence of parameters of the new model on the size of cascading failure of network through exponential analysis, numerical simulation and instance simulation. In addition, we propose the corresponding distribution strategy of protection resources, which provides a basis for establishment of the protection mechanism and the allocation of protection resources.(4) Considering the load characteristics, the influence of coupling between different networks on the ability of network against cascading failure in different conditions is researched based on two failure modes, which aims to find ways reducing the size of cascading failure of complex network. The results of numerical simulation and instance analysis show that capacity threshold is one of the key factors which influence the size of cascading failure of interdependent network. The capacity threshold directly determines the influence strength of coupling on the ability of network against cascading failure. Furthermore, compared with descending order of decoupling, ascending decoupling is more helpful in reducing the influence of coupling on the ability of network against cascading failure.