关键词: 相依网络/
级联失效/
恢复模型/
择优算法
English Abstract
Enhancing resilience of interdependent networks against cascading failures under preferential recovery strategies
Wu Jia-Jian1,Gong Kai1,2,3,
Wang Cong4,
Wang Lei1
1.School of Economic Information Engineering, Southwestern University of Finance and Economics, Chengdu 611130, China;
2.Collaborative Innovation Center for the Innovation and Regulation of Internet-based Finance, Southwestern University of Finance and Economics, Chengdu 611130, China;
3.Key Laboratory for Financial Intelligence and Financial Engineering of Sichuan Province, Southwestern University of Finance and Economics, Chengdu 611130, China;
4.Visual Computing and Virtual Reality Key Laboratory of Sichuan Province, Sichuan Normal University, Chengdu 610068, China
Fund Project:Project supported by the National Natural Science Foundation of China (Grant No. 61602331), the Fundamental Research Funds for the Central Universities of China (Grant No. JBK170133), and the Scientific Research Foundation of the Education Department of Sichuan Province, China (Grant No. 17ZB0434).Received Date:24 November 2017
Accepted Date:26 January 2018
Published Online:20 April 2019
Abstract:Interdependent networks are extremely fragile because a very small node failure in one network would trigger a cascade of failures in the entire system. Therefore, the resilience of interdependent networks is always a critical issue studied by researchers in different fields. Existing studies mainly focused on protecting several influential nodes for enhancing robustness of interdependent networks before the networks suffer random failures. In reality, it is necessary to repair a failing interdependent network in time to prevent it from suffering total breakdown. Recent investigations introduce a failure-recovery model for studying the concurrent failure process and recovery process of interdependent networks based on a random recovery strategy. This stochastic strategy covers repairing a small fraction of mutual boundary nodes which are the failed neighbors of the giant connected component of each network, with a random probability of recovery . Obviously, the random recovery is simple and straightforward. Here, we analyze the recovery process of interdependent networks with two types of connectivity links, i.e., the first-type connectivity links and the second-type connectivity links, which represent the mutual boundary nodes(being also failed nodes) linked to survival nodes in current giant connected component, and linked to failed nodes out of current giant connected component in networks, respectively. We find that when mutual boundary nodes have more first-type connectivity links, the current giant connected component has higher average degree and immediately makes better interdependent network resilience, on the other hand, more second-type connectivity links generate more candidates during the recovery procedure, and indirectly make better system resilience. In short, two types of connectivity links of mutual boundary nodes both have great effects on the resilience of interdependent networks during the recovery. In this paper, we propose a new recovery strategy (preferential recovery based on connectivity link, or PRCL) to identify the mutual boundary node recovery influence in interdependent networks, based on the failure-recovery model. By defining two indexes that represent the numbers of first-type and links second-type connectivity links, respectively, we calculate the boundary influence with one parameter f by combining together with two indexes. After calculating all boundary nodes in the current process, we obtain a boundary importance index which is more accurate to indicate recovery influence of boundary node for each boundary node in interdependent networks. Our strategy is applied to interdependent networks constructed by ER random network or/and scale-free network with the same average degree. And a dynamical model of random failure based on percolation theory is used to make a comparison of performance between PRCL and other recovery strategies(including random recovery, preferential recovery based on degree, preferential recovery based on local centrality) in terms of four quantitative indices, i.e., probability of existence of the giant connected component, number of iteration steps, recovery robustness and average degree of the steady state of the giant connected component. Experiments on different interdependent networks (ER-ER/SF-SF/ER-SF/SF-ER) demonstrate that with a very small number of mutual boundary node recoveries by PRCL strategy, the resilience and robustness of entire system under the recovery process can be greatly enhanced. Finally, the only parameter f in PRCL strategy is also discussed. Our strategy is meaningful in practice as it can largely enhance interdependent network resilience and contribute to the decrease of system breakdown risk.
Keywords: interdependent networks/
cascading failures/
failures-recovery model/
preferential recovery strategy
