A Reinforcement Learning Approach for Defending Against Multiscenario Load Redistribution Attacks

The accelerated digitalization of today's electric power infrastructure has highlighted the gravity of cyber resilience in the electricity ecosystem. Due to the expanded attack surface inherited in the digitalized operational and information technologies, the electricity sector is facing a continually escalating cyberthreat landscape and must act to prepare for more frequent and sophisticated cyberattacks as the new normal. To aid in this effort, this paper presents a novel approach to identify critical branches to strengthen and by doing so, shield the smart-grid power system from the threat of load redistribution attacks (LRAs) under a wide range of operating scenarios. Compared with conventional critical branch identification approaches, we propose a new concept, namely, chain of defense, that empowers the system operator to incorporate existing cyber protections and develop branch strengthening strategy in a more dynamic, adaptive, and flexible way. We then propose a novel reinforcement learning framework to identify the most effective chain of defense. A cross-updating search strategy is developed to specifically ensure that the identified chain of defense can safeguard the network from the two-fold damaging effects of the LRAs on operation economics and security simultaneously. The effectiveness of the proposed approach is evaluated on the IEEE 14-bus system and the European 89-bus system. Simulation results indicate that the proposed approach provides more comprehensive and effective mitigation against the damaging effects of LRAs compared with the conventional approaches.