PMU Placement with Power Grid Partitioning for Line Outage Detection

The detection of fault locations is an important functionality in power distribution systems. In our previous work [1], we have proposed a data-driven approach to identify the locations of a minimum number of phasor measurement units (PMUs) in a power system, such that a desired fault detection accuracy is achieved. The locations of the PMUs have been determined with a feature selection method based on mutual information (MI), whereas a decision tree (DTs) has been used as a classifier. The contribution in the present work is a partitioning method which enables the application of this data-driven sensor placement approach to larger power systems. We show how a power system graph can be virtually partitioned into several child grids to create sub-zones, and optimally place a PMU in each sub-zone. The originality lies in the partitioning of a power system graph without reconfiguring the grid topology or removing any lines. Therefore, the proposed method enables the exploitation of all potentially available data sources which carry useful information for fault detection, and leads to a reduction in the number of required PMUs.