Power Flow Analysis Using a Graph Database and Graph Neural Networks

As we know, power flow calculation is fundamental to power systems. Currently, the mainstream power flow algorithms utilize programming for power flow calculations. However, when faced with line faults or expansions and renovations, there is a need to re-iterate calculations using a large amount of data. First, with the aim of improving the speed of power flow calculations and simplifying the process of power flow calculations after adjustments to the power system structure, the emerging technology of graph databases is highly compatible with power grid topologies. Under the condition that data for various parts of the power grid and their relationships are known, graph databases and related theories can be used to build graph models, and graph computing theories can be applied to optimize the power flow calculation process, thereby accelerating the calculation. Secondly, graph database-based power flow calculations belong to parallel computing, whereas traditional power flow calculations are serial computations. In theory, parallel computing is faster than serial computing. By comparing traditional power flow calculations with graph database-based power flow calculations, the speed advantages of using graph databases in power flow calculations can be established. Finally, due to the rise of artificial intelligence, this paper also explores the use of graph neural networks for power flow calculations in power grids, concluding that using graph neural networks for training can yield accurate power flow calculation results.