Parameter Identification of Power Transmission Line Based on Graph Attention Network

Transmission line parameters are the basis of power system calculation. Considering the variability and complexity within the development environment of large-scale power grids, the reliable functioning of large power system is facing considerable challenges, requiring increased requirements for the immediate analysis and flexible alerting abilities of the electric grid. Current approaches to power grid parameter identification neglect the historical trends and distinct characteristics of grid branches while also disregarding the interconnected nature of the grid's topology. As a result, they were unable to capture the correlations between the central node and its neighboring nodes. To address these shortcomings, this study introduces a graph attention network (GAT) approach for parameter identification. The approach presented in this paper shows a considerable improvement in accuracy compared to earlier methods. It also proves to be robust in dealing with noise and errors in measurements, and is capable of adapting to different situations prevalent in actual electric power grids.