Power Edge Node Resource Scheduling Mechanism Empowered by Deep Reinforcement Learning

This paper proposes a dynamic scheduling mechanism based on Deep Reinforcement Learning (DRL) to optimize resource allocation in Power Internet of Things (PIoT) edge nodes. A three-layer cloud-edge-terminal architecture is developed, integrating Deep Q-Network (DQN) and Temporal Convolutional Network (TCN) to enhance task offloading and resource allocation. The mechanism adapts strategies dynamically to real-time network conditions and task demands, minimizing total task delay and resource occupancy while ensuring delay balance. Experimental results demonstrate that the proposed method outperforms traditional heuristic algorithms across various load conditions, enhancing resource utilization and computational performance.