Grid reconfiguration for congestion management of distribution grids using deep learning

Inhalt:
Due to the increasing penetration of renewable energy into the power grid, network operators are increasingly having to deal with network congestions, i.e., stations in which the capacity of the electric network is not sufficient to transport enough electricity. This is especially true for distribution system operators, who now need to take a more active role in congestion management. Typically, congestions are resolved by curtailing renewable generation., however these actions are associated with high costs. Topology reconfiguration is an alternative method to avoid network congestions by rerouting the flow of electricity through the network. In this work a topology reconfiguration algorithm based on neural networks is proposed to resolve congestions in distribution grids. The neural network suggests candidate topologies to resolve a congestion scenario. The candidate solutions are then ranked based on desirability and tested to ensure they can actually resolve the congestion. Results show that the novel method finds feasible topologies for a wide range of congestion scenarios in a representative test grid.