Reinforcement Learning based Optimization of Line Switching off during Cascading failures in Power Grids

Nowadays, large-scale power outages caused by cascading failures occur frequently around the world. It is of great significance to study the control strategies for cascading failures to ensure the safety and reliability of power grids. During cascading failures, the shift of the flow from the broken elements to the others will further cause the overload of the other elements. After the initial failure occurs, if we immediately switch off a group of key lines to control the transfer of the flow, the cascading failure could be effectively mitigated. The line switching-off optimization method based on reinforcement learning is studied and the cascading failure's control methods for known failure scenes and random failure scenes are proposed respectively. Simulations results on IEEE 118-bus system validate our proposed method.