Power system emergency control strategy based on severely disturbed units identification and STGCN-DDQN

With increasing uncertainties and complexity in modern power system, it is still very challenging to figure out a cost-effective emergency control strategy that can reliably react to the spatio-temporal variations. Faced with this challenge, this paper presents a power system emergency control strategy based on severely disturbed units identification and STGCN-DDQN (Spatio-Temporal Graph Convolutional Network-Double Deep Q-Network). Inspired by the novel combination of STGCN for spatio-temporal features extraction and DDQN for decision making, the STGCN model helps the DDQN agent to better capture spatio-temporal features when dealing with topology variations. Considering the problem of combination explosion, the severely disturbed units are chosen as action objects of STGCN-DDQN, which makes the agent quickly find the optimal solution. Further, the al-gorithm of emergency control strategy is designed based on the close combination of severely disturbed units identification and STGCN-DDQN. Finally, by building multiple test scenarios, the proposed method is verified based on the New England 39-bus standard system and Northeast China power system. The simulation results demonstrate that the proposed method outperforms the other schemes in terms of control performance when considering topology variations, modeling errors, power flow conditions and random noise.