Q-learning-based sequential recovery of interdependent power-communication network after cascading failures

Wei Huang, Yuxin Gao, Tianyi Zhang, Hua Gao

Abstract: Enhancing robustness of cyber-physical system under cascading failures still remains to be an important problem. In this paper, we propose a novel coupled model with the process of cascading failure, the DC power flow model and sequential bus/branch recovery in the interdependent power-communication network. Then, we adopt the Q-learning algorithm to search out the optimal recovery sequence with the minimal number of recovery times. By comparing the recovery costs of sequential bus recovery and sequential branch recovery, it is found that the strategy of branch recovery requires less recovery cost. Besides, we have also compared the performance of Q-learning-based branch recovery strategy with those of several other topology-related branch recovery strategies. Experimental results show that the Q-learning-based algorithm can effectively find the optimal sequence of branch recovery.