Sequential topology recovery of complex power systems based on reinforcement learning

Cascading failure is among the most critical threats to the security and resilience of modern power systems and has attracted a wealth of research interest in the past decade. Most of the existing studies have investigated the issue of cascading failure on complex power systems mainly from the attacker's perspective. From the perspective of a system defender or operator, fast restoration of the power system to normal operation is also important. In this paper, we consider cascading failure in conjunction with the restoration process involving repairing of the failed nodes in a sequential fashion. Based on a realistic power flow model depicting cascading failures, we apply reinforcement learning to develop a practical and effective strategy to identify an optimal sequential restoration process for large-scale power systems. Simulation results on three benchmark power systems demonstrate the learning ability and the effectiveness of the proposed strategy.