Distribution Network Topology Identification via Variance Analysis of Smart Meter Data

Accurate topology knowledge is vital for modern distribution networks, yet many data-driven identification methods are limited by unrealistic assumptions, such as requiring high-resolution data, known line parameters, or uncorrelated power injections. This paper proposes a novel, purely data-driven algorithm that identifies the operational radial topology using only smart meter voltage and power data. Our two-step method first proposes edges based on voltage statistics and then validates them using a windowed decorrelation technique. A key advantage is its effectiveness under realistic, positively correlated load conditions where many algorithms fail, without needing line parameters or active grid intervention. Simulations on IEEE test feeders demonstrate superior accuracy over existing methods that ignore load correlation.