The high risk nature of power field operations necessitates intelligent safety management solutions.Traditional methods relying on manual experience and static mechanisms struggle to adapt to dynamic environments and complex risks.To address this issue

this paper proposes a cloud edge device collaborative safety management method integrated with intelligent perception.A lightweight edge deep learning model

combined with transfer learning

is employed to maintain model accuracy under edge resource constraints.Experimental validation confirms that the proposed cloud edge device collaborative architecture enables real time risk identification

resource optimization

and proactive early warning

thereby providing a scalable pathway to enhance safety management for power field operations.