The deployment method of cloud computing is commonly used for artificial intelligence algorithm in current power inspection technology. For the outstanding problems such as huge energy consumption, low efficiency and poor timeliness of cloud computing, this paper proposes a low-energy power inspection scheme based on “cloud-fog-edge” collaborative technology. The portable AI diagnostic device at the edge side and the fixed UAV library at the fog side are developed respectively, and the method of two-level traffic offloading and cooperative diagnosis is adopted to sink the AI computing power to the edge side and reduce the pressure of 5G base stations and central servers, so as to achieve the global optimization of system energy consumption and the accurate identification of power equipment defects. To complete an inspection task in cloud computing mode, the system consumes 22,000~32,000 KW·h; to complete an inspection task in cloud edge collaboration mode, the system consumes 12,000~18,000 KW·h at the highest, reducing 16.93%~63.10%; to complete an inspection task in cloud fog edge collaboration mode, the system consumes 7,400~11,000 KW·h, reducing 48.8~77.25%, while the timeliness, accuracy, data security, system stability and inspection efficiency of identification have been greatly improved.