Abstract: With the development of electricity industry, power system operation and maintenance personnel generate more requirements for power data analysis, such as multi-modal data analysis and intelligent data retrieval. Traditional data analysis methods show the disadvantages of high cost, long cycle, and poor flexibility in the face of these requirements. To address this problem, this article proposes a framework of electric power data application with generative pre-trained model. This framework leverages the effective abilities of semantic understanding and content generation of generative pre-trained models, so as to allow users to implement data analysis in the form of natural language, thus improving the efficiency of data analysis. In particular, this article provides training and testing processes for the proposed application framework, which could serve as a reference for practice.