Abstract: The digital simulation analysis for power grid requires a lot of manpower and time cost to comprehensively analyze simulation. The process of analysis and adjustment heavily relies on manual experience knowledge which lacks a unified form of knowledge representation method, because of the diversity of expression and particularity of application scenario. Thus, a knowledge modeling method for artificial intelligence simulation analysis of power grid is proposed. According to the characteristics of adjustment knowledge experience for digital simulation analysis, a knowledge representation method based on subject-predicate-object (SPO) is designed for adjustment process, adjustment operations and program calls, which combines qualitative and quantitative, correlation and affair knowledge. A modular knowledge-driven model is built to automate the process of power grid simulation analysis. A set of knowledge extraction and cleaning method is designed to obtain the required knowledge in the form of triples, and the knowledge graph is constructed by using the obtained knowledge triples which is developed to a visual knowledge base management system. The experiments of convergence adjustment for power flow calculation are carried out on the improved CEPRI36 node system and northeast power grid, which verify the feasibility and effectiveness