Learning Topology of the Power Distribution Grid with and without Missing Data

TL;DR

This paper introduces a novel voltage measurement-based algorithm to accurately learn the radial topology of power distribution grids, even with incomplete data, by formulating it as a minimum-weight spanning tree problem.

Contribution

It proposes a new low-complexity method that infers grid topology using voltage variances without needing line parameters, effective even with missing data.

Findings

Algorithm accurately identifies grid topology in tests.

Method performs well with partial voltage data.

Approach reduces computational complexity compared to existing methods.

Abstract

Distribution grids refer to the part of the power grid that delivers electricity from substations to the loads. Structurally a distribution grid is operated in one of several radial/tree-like topologies that are derived from an original loopy grid graph by opening switches on some lines. Due to limited presence of real-time switch monitoring devices, the operating structure needs to be estimated indirectly. This paper presents a new learning algorithm that uses only nodal voltage measurements to determine the operational radial structure. The algorithm is based on the key result stating that the correct operating structure is the optimal solution of the minimum-weight spanning tree problem over the original loopy graph where weights on all permissible edges/lines (open or closed) is the variance of nodal voltage difference at the edge ends. Compared to existing work, this spanning tree based approach has significantly lower complexity as it does not require information on line parameters. Further, a modified learning algorithm is developed for cases when the input voltage measurements are limited to only a subset of the total grid nodes. Performance of the algorithms (with and without missing data) is demonstrated by experiments on test cases.