Research Article
Cloud-Edge Collaboration Based Power IoT Scene Perception Mechanism
@INPROCEEDINGS{10.1007/978-3-031-23141-4_8, author={Sujie Shao and Congzhang Shao and Cheng Zhong and Shaoyong Guo and Pengcheng Lu}, title={Cloud-Edge Collaboration Based Power IoT Scene Perception Mechanism}, proceedings={Game Theory for Networks. 11th International EAI Conference, GameNets 2022, Virtual Event, July 7--8, 2022, Proceedings}, proceedings_a={GAMENETS}, year={2023}, month={1}, keywords={Cloud-edge collaboration Power IoT Scene perception Transfer learning Edge intelligence}, doi={10.1007/978-3-031-23141-4_8} }- Sujie Shao
Congzhang Shao
Cheng Zhong
Shaoyong Guo
Pengcheng Lu
Year: 2023
Cloud-Edge Collaboration Based Power IoT Scene Perception Mechanism
GAMENETS
Springer
DOI: 10.1007/978-3-031-23141-4_8
Abstract
Fast and high-quality scene perception is an important guarantee for the efficient, stable and reliable operation of the power Internet of things, which can assist the decision-making of upper-level applications. The transmission delay of scene perception based on cloud computing is high, so it is difficult to meet the needs of real-time decision-making the mode based on edge computing is not competent for all real-time perception tasks due to the limited computing resources. For this reason, this paper proposes a scene awareness mechanism of the power Internet of things based on cloud-edge collaboration. A scene information awareness architecture based on cloud-edge collaboration is constructed, and a scene information processing flow that distinguishes dynamic instances, static instances and general instances is designed to support local scene information edge awareness and global scene cloud synthesis. Focusing on the construction of high-precision neural network recognition model of high-frequency dynamic examples, using the idea of transfer learning, a neural network model training framework based on cloud-edge collaboration is designed. Simulation results show that the scene perception mechanism proposed in this paper can effectively reduce the perception processing delay and model training time on the basis of accurately perceiving the scene, and improve the adaptability of the perception model to high dynamic scenes.
