Dynamic Transfer Method of Substation Diagram Model Based on Intelligent Middle Platform

With the increasing complexity of power grid operations and the growing demand for efficient data integration and intelligent management, dynamic and accurate transfer of substation whole-station diagram models has become a critical challenge. This paper proposes a novel dynamic transfer method based on an intelligent middle platform that integrates multisource data-including SCADA voltage and load measurements, historical alarm texts, and meteorological temperature and humidity information-to construct a comprehensive power grid graph model reflecting substation structure and operational status. The method combines physical modeling, BERT-based text encoding, and environmental factor transformation to achieve effective feature fusion into a unified input vector. A safety assessment mechanism is incorporated, leveraging system safety factors calculated from voltage, load, and operational constraints to ensure secure operation. Furthermore, a graph neural network with Graph Attention Network (GAT) architecture dynamically aggregates node information to generate adaptive scheduling suggestions, which are modulated by safety factors to produce accurate dispatch commands. Validation through a representative three-node substation case study demonstrates that the approach enhances voltage profile optimization, load utilization, and reactive power compensation while maintaining safe operational boundaries, confirming its practical value for intelligent substation management.