Optimization for sequential communication line attack in interdependent power-communication network

In this paper, we construct a novel model to study cascading failures that occur in the interdependent power-communication network. Based on this model, we employ the Q-learning algorithm to search for the optimal attack sequence against the lines in the communication network with the aim to bring the most destructive damage to the power grid. The effectiveness of the deduced optimal attack sequence is validated through numerical simulations, in which the IEEE-39 bus test system is used as the power grid and the Barabasi-Albert (BA) scale-free network is modeled as the communication network. It is found that in the interdependent power-communication network, large-scale failure in the power grid can be caused by attacking only a small fraction of lines in the communication network. In addition, under the same number of attack actions, the attack sequence resulted from Q-learning algorithm is more destructive than random attack sequence and the attack sequence based on traditional complex network characteristics. Furthermore, according to the analysis of simulation results, we identify the most vulnerable lines in the communication network and propose the corresponding protection strategy. The proposed protection strategy proves to be able to effectively reduce the number of independent optimal attack sequences that the attacker can launch and improve the robustness of the whole network. (C) 2021 Elsevier B.V. All rights reserved.