A Method and System for Power Grid Engineering Error Diagnosis Based on Improved Graph Convolutional Networks

The integration of power grid operation is more prominent in the era of informatization, which puts higher requirements on real-time dispatching and operation of the power grid. Objectively, it requires strengthening the global analysis and decision-making capabilities of integrated super power grids. Models and real-time data serve as the cornerstone for online analysis and decision-making. The quality of data directly affects the accuracy of online analysis and decision-making results. Massive model data supports more precise integrated analysis results, but also presents huge challenges in model data quality control. To address this issue, we propose an improved graph convolutional network-based method and system for power grid engineering error diagnosis. By adding convolutional layer feature extraction based on traditional graph convolutional networks, we can improve the accuracy of model error diagnosis and reduce training loss. At the same time, we construct a knowledge graph for power grid engineering error diagnosis with nodes and branches as the core. This solves the problem of quality inspection of model data under the comprehensive influence of multiple factors and promotes the transformation of model data from manual maintenance to intelligent maintenance.