Abstract: In the background of rapid development of new power systems, the deep coupling between massive multi-source heterogeneous information and diverse business brings significant challenges such as strong complexity and randomness in the power system operation. Concurrently, accelerating the construction of a flexible and intelligent new power system is a crucial strategy for energy development. There is an urgent need to establish a technology system of artificial intelligence for electric power system（AI EPS） that is intelligent, self-adaptive, and secure, in order to promote the intelligent transformation and development of the new power system. This paper reviews and summarizes the evolution and current research status of AI EPS technologies. It analyzes the technical framework, principles, and key technical methods for the new generation of AI EPS, which is based on pretrained multimodal large models. The application schemes for power large model technology in the scenarios such as perception prediction, dispatching and control decision-making, and operation planning are proposed. The technical challenges and application bottlenecks faced by electric artificial intelligence based on power large models are discussed. Finally, the application of electric artificial general intelligence technology is summarized and prospected.