Abstract: In response to the challenges of multi-objective dynamic coupling and full process collaborative optimization in the decision-making process of power outage plans in distribution networks, research is conducted on intelligent decision-making for power outage plans based on deep reinforcement learning. Deconstructing the entire process of power outage planning, clarifying the decision-making objectives and constraints of each stage; Complete the characterization of power grid status through spatiotemporal feature extraction and deep feature encoding; Design an intelligent decision engine based on deep reinforcement learning, establish a temporal decision model, and introduce attention mechanism to optimize strategy learning; Transform the learned optimal decision strategy into executable action sequences to achieve intelligent generation of power outage planning, load transfer, and operation sequences. Comparative experiments show that the proposed method outperforms traditional methods in terms of robustness, economy, and timeliness of decision-making schemes, verifying its superiority and engineering application potential in complex power grid operation scenarios.