Transient Stability Emergency Control Strategy of Power System Based on Deep Reinforcement Learning

In order to form the on-line emergency control according to the real-time wide area measurement data, a transient stability emergency control strategy based on deep reinforcement learning is proposed in this paper. In this paper, the emergency control problem of power system is regarded as a sequential decision problem, and an emergency control model based on deep reinforcement learning is established to learn the mapping relationship between power system state and emergency control actions; secondly, according to the transfer learning technology, the model training process is divided into two stages: pre-training and full training, which can improve the adaptability of the model to multiple scenarios by model and parameter transfer; finally, the rainbow algorithm is used to solve the optimal emergency strategy of generator tripping and load shedding, and the strategy is verified in New England 10-machine and 39-bus system. The results show that the proposed method has high training efficiency and can restore the stability of the unstable system.