Placement of distributed energy storage via multidimensional scaling and clustering

Energy storage has long been proposed at the distribution level, where it can provide additional benefits via ancillary services. This work studies how to place energy storage units (ESU) on a distribution feeder in the most cost-effective manner while still meeting voltage regulation requirements. The feeder also has photovoltaic (PV) generation, and the PV ability to supply reactive power is considered. The placement of the ESU is performed via a fast heuristic, in which multidimensional scaling (MDS) is used to transform the combinatorial placement problem into a continuous-valued problem by mapping buses to points in a space. In the new space, clustering algorithms can be applied to determine the ESU locations from a set of candidate locations. The method reduces computation time by an order of magnitude, allowing for various distribution feeder configurations to be quickly compared.