Voltage-based clustering to identify connectivity relationships in distribution networks

We present a novel technique to infer connectivity relationships corresponding a set of measurement or monitoring locations in a distribution grid. Our approach uses a time series of voltage measurements from each location and partitions these into clusters based on whether the locations belong to same or different sub-circuits in the grid. For instance, customer meters connected to the same phase may be clustered into one group. The methods allow distributors to verify the connectivity model of their distribution network and also improve its accuracy. Furthermore, the methods support incremental meter deployment and can be applied to any subset of measurement locations. We present initial experimental results with the help of real voltage time series measurements collected from a microgrid.