Clustering nodes in a directed acyclic graph by identifying corridors of coherent flow

This paper proposes a novel method for clustering nodes based on prevailing power flow conditions within a power grid. To this end, first, the network's active power flow state is modelled as a directed acyclic graph. This digraph explicitly represents where power is flowing and this can help in monitoring and analysing system vulnerabilities. The directed acyclic graph representation also allows easy identification of those buses that solely provide or absorb active power: these are pure source and sink nodes, respectively. An iterative path-finding procedure is applied to every node in the system, to enumerate the sources that is fed by, and the downstream sinks towards which it forwards power. The novel clustering algorithm is then applied, to group together those nodes which share the same set of reachable sources and sinks. This novel clustering methodology is proposed in the first instance as a tool to boost the situational awareness of control room operators by better summarising aggregate power flow dispositions in large grids. The proposed methodology is applied to two sample grids, and an analogy to river systems is articulated, applying such notions as tributaries, distributaries and the central mainstream to electrical networks.