Research On Cascading Failure Models On Multiple Network Structure

With the rapid development of human society network,the security and reliability of the network have various serious problems.Critical infrastructure networks in reality,such as power networks,communications networks,Internet networks and transportation networks often suffer from natural disasters or deliberately attacks,the operating rules of these network systems will change dramatically.Therefore,the robustness of the network will be significantly reduced.When the network structure changes,the overall load of the network will change,and the loads of some nodes or edges need to be redistributed.When the nodes and edges load in a network exceeds their maximum allowable capacity,these nodes or edges become congested or even collapse.If this process continues,large areas of failure will occur at nodes or edges in the network.From the perspective of complex networks,this process is called"cascading failure".The large-scale cascading failure once occurs,it can often have highly destructive power and influence.Therefore,the robustness and security of these key infrastructure network systems have attracted great attention from various countries and governments,and attracted extensive attention from scholars at home and abroad.In recent years,the cascading failure of complex networks has been extensively studied and has become an important branch of the study of complex network dynamics.Based on the research results at home and abroad,this thesis studies the cascading failure models from the perspectives of single-layer network,network with community structure and double-layer network structure,for how to improve the robustness of the network,the optimization models are established.The main research results of this thesis are as follows:(1)For the cascading failure models of the Internet,a local congestion-aware routing strategy is proposed.The dynamic load on the node is redefined,and the congestion function to represent the congested extent of node is proposed.We construct the "cost" weight function considering shortest path routing strategy and node congestion function.When routing the packets in the network,the next hop node is selected according to the "cost" weight.The cascading failure models caused by deliberate attacks on two typical network models are studied.It is found that there is an optimal value for the tunable parameter of adjusting the "cost" weight function,this optimal value enables the network to achieve the strongest robustness.At the same time,the influences of the tolerance parameters of nodes,the generation rate of packets,the attacking rate of nodes and the time delay of information transmission on the network robustness are also analyzed.(2)For the network model with overlapping community structure,the community reconnection algorithm is proposed,and the cascading failure model caused by deliberate attack is studied.The reconnection algorithm can change the network community structure from overlapping to non-overlapping,while the degree distribution characteristics of the network remain unchanged.Simulations were made on the synthetic network generated by the algorithm proposed by Lancichinetti and Fortunato and on the actual data set of the electric power network in the western United States,the simulations found that the higher the community overlap,the stronger the network robustness;moreover,the average degree of network nodes,the mixed parameters of network,the number of overlapping nodes in the initial community network structure and the number of communities in which the overlapping nodes belong all have some influence on the robustness of the networks.(3)For the two-layer network model,an attack method based on grey information is proposed,and the cascading failure model validations are carried out on the interdependent network and the interconnected network.Through numerical simulations,it is found that for these two kinds of networks,the network robustness can be improved by reducing the accuracy of attacking information under the three coupling pattern of assortative,disassortative and random coupling.There exsits a critical phenomenon for attacking information accuracy,when the value is greater than the critical value,the robustness of the network is almost unchanged.For the interconnected network,the network robustness increases with tolerance parameter for any coupling preference.The robustness of the assortative coupling and random coupling of the network model increases with the coupling probability,however,for disassortative coupling,there exists a critical phenomenon for coupling probability,the higher the coupling probability,the stronger the robustness of the network isn't.Finally,two kinds of two-layer network models are validated using the actual interconnected autonomous system level Internet in Korea and Japan.(4)For the robust performance of complex networks,the intelligent optimization models are established.For single layer networks,the multi-objective optimization models that comprehensively consider the operational cost of the edges in the networks and the robustness of the networks are established.The NNIA(Non-dominated Neighbor Immune Algorithm)is applied to solve the optimization models.In the solutions,we find the edges that can facilitate the propagation of cascading failure and the edges that can suppress the propagation of cascading failure.This provides a certain reference for how to take remedial measures to weaken the damage caused by cascading failures.Network decision makers can also make more realistic choices based ondifferent operational costs.For two-layer interconnected networks,a single-objective optimization model considering the coupling preference between networks is established.Typical ant colony algorithm,memetic algorithm and simulated annealing algorithm are used to solve the model.The simulations show that the network robustness of the three algorithms is significantly better than that of the three typical coupling patterns.Finally,it is verified on the actual autonomous system level Internet in Korea and Japan.