Distribution fault location using graph neural network with both node and link attributes

This invention describes a system for power companies to quickly find and classify electrical faults in their distribution lines. It works by collecting live data from sensors across the power grid, organizing this information into datasets for both the wire segments (branches) and connection points (nodes). These datasets are then fed into a specialized computer program, a trained neural network, which identifies the exact location and type of fault, triggering an alert. The claims emphasize the use of real-time sensor measurements for this process.