Abstract:In order to solve the problem that the traditional safety management method of power construction site mainly based on artificial supervision can''t meet the needs of "full coverage, full process, high efficiency" safety supervision of current power grid operation, this paper adopts the more mature artificial intelligence technology in the industry today, and constantly enhances the accuracy and correctness of feature capture and recognition through feature learning of hundreds of thousands of pictures To achieve the functions of identification of construction personnel, behavior identification of construction personnel, status identification of switches and switches, safety measures and signs, abnormal behavior identification, etc., and develop an intelligent safety management and control system for power construction. The system effectively standardizes the construction process through the process design, and carries out voice prompt and alarm for the identified abnormal information through the program design, so as to realize the automation and intellectualization of the safety management of the electric power construction site, timely discover, stop and correct the unsafe behaviors, and effectively improve the safety supervision and management effects.