Frontiers in Energy Research (May 2022)
A R-GCN-Based Correlation Characteristics Extraction Method for Power Grid Infrastructure Planning and Analysis
Abstract
For a large number of grid infrastructure projects, various interrelationships may have an impact on portfolio optimization to a certain extent. At present, there are few qualitative analyses considering linkages among massive power grid infrastructure projects. In order to overcome the limitations of the existing studies, this paper proposes a method for extracting the correlation characteristics of massive power grid infrastructure projects based on relational graph convolutional neural network (R-GCN). The correlation characteristics of power grid infrastructure projects with different voltage levels, engineering attributes and project properties are comprehensively considered. R-GCN generalizes the traditional graph convolutional neural network and can process multi-relational data, building an encoder and identifying multiple relations between entities in the project library by accessing different layers to solve corresponding modeling problems, so as to accurately identify the linkages among a large number of power grid infrastructure projects, and further improve the rationality of portfolio optimization.
Keywords
