Sequential recovery method and device for power system based on deep reinforcement learning

The invention discloses a sequential recovery method and device for a power system based on deep reinforcement learning. The method comprises the following steps: constructing a power system recovery model which comprises a deep reinforcement learning Q value estimation network and a Target Q network, and training the power system recovery model. According to the invention, based on the power network after a cascade failure and through a bus recovery sequence obtained after deep reinforcement learning, the recovery capability of the power network system to the cascade failure in a system recovery process is evaluated, reinforcement learning is combined with the power network, and the recovery problem of the power network is considered from the perspective of defenders; and through combination with a neural network, the implementation range of the power network is expanded; that is, an optimal recovery strategy of a large power grid can be found. 本发明公开了一种基于深度强化学习的电力系统顺序恢复方法及装置，通过构建包括深度强化学习Q值估计网络和Target Q网络的电力系