Research on Intelligent Management Architecture for Power Grid Data Catalog Based on Artificial Intelligence

With the digital transformation of the power grid, the scale and complexity of power grid data have significantly increased, encompassing various types such as SCADA systems, GIS data, and dispatch operation logs. This necessitates efficient and intelligent management methods. This paper proposes an intelligent classification method for power grid data catalogs based on the Transformer model. Specifically, the T5 model is employed to convert the classification task into a text generation task, while table embedding techniques are integrated to enhance data representation. To address the issue of class imbalance, Focal Loss is introduced to optimize classification performance, and soft prompt learning is incorporated to improve the training effectiveness for small sample data. Experimental results demonstrate that the proposed method outperforms traditional approaches in power grid data classification, achieving higher classification accuracy and better generalization capability. Furthermore, intelligent retrieval and knowledge graph construction enable efficient data management and intelligent querying. This research provides a novel technological approach to power grid data management, contributing to the enhancement of intelligent power grid operations.