Task Offloading Based-on Deep Reinforcement Learning for Microgrid

Authors:

Abstract
Nowadays, microgrid plays an important role in building local energy system to generate and consume power. However, there are numerous computing tasks and high delay in the data processing from users. Mobile edge computing (MEC) technology emerges as a new method, especially through the way of the edge-terminal collaborative. Furthermore, we propose a new MEC architecture in the background of microgrid, and then use deep Q network (DQN) to minimize the total processing latency to optimize offloading decision and resource allocation. Experiments results show that the proposed DQN algorithm can effectively reduce latency and improve convergence compared to MEC-only algorithms and local-only algorithms.
Funding Information
  • National Natural Science Foundation of China (U2006222)

This publication has 14 references indexed in Scilit: