A Novel Topology Recovery Method for Low Voltage Distribution Networks

TL;DR

Low voltage distribution networks suffer from poor visibility due to sparse measurements and undocumented topology changes, hindering fault detection and optimization. The paper proposes a topology-recovery method that uses incremental voltages and currents ($\Delta V$, $\Delta I$) from smart meters and an AC power-flow model to infer the hidden, radial LVDN structure, without relying on DC approximations. It constructs sample correlation $\Sigma$ and precision $\Theta = \Sigma^{-1}$ from leaf data and derives node distances $D_{ij} = 1/|\Theta_{ij}|$ with a grouping threshold $\tau = 0.2$ to recover hidden layers iteratively. Simulations across 11–25 node systems show 100% topology-recovery accuracy with sufficient samples, validating robustness and practical applicability for real-world distribution networks. The approach promises improved outage management and grid-aware operation and points to future extensions to unknown line impedances and meshed-grid configurations.

Abstract

Low voltage distribution networks (LVDNs) suffer from limited visibility due to sparse or nonexistent measurement systems, leaving distribution network service providers with incomplete data. Maintenance activities, such as transformer upgrades and power line replacements, sometimes go undocumented, leading to unmonitored topology changes. This lack of oversight hinders network optimization, fault detection, and outage management, as utilities cannot fully monitor or control the system. With the rise of electric vehicles, having an accurate understanding of LVDN topology is crucial to avoid infrastructure damage from potential overloads. This paper introduces a method to reconstruct LVDN topology using incremental voltage and current changes from smart meters at customer endpoints. The approach identifies and maps network topologies with high accuracy, overcoming limitations of prior methods by discarding unrealistic assumptions. Specifically, it addresses grids with fewer than three pole connections and employs an AC power flow model over simplified DC approximations. Simulations across diverse configurations validate the method's effectiveness in accurately reconstructing LVDN topologies, enhancing real-world applicability.

Figures (6)

• Figure 1: A typical LVDN with 100% smart meter coverage of residential loads
• Figure 2: For the 11-node test system, histograms for Node 2 display the distribution of voltage incremental changes ($\Delta V$) and current changes ($\Delta I$) across the system.
• Figure 3: Schematic of a 6-node grid used in the proposed method derivation. Green nodes represent observable nodes, while light blue nodes indicate hidden nodes.
• Figure 4: Graph representations of test systems for simulation results. Light blue nodes indicate internal hidden poles, while green nodes represent observable consumer loads with smart meters. (a) 11-node system, (b) 15-node system, (c) 20-node system, (d) 25-node system.
• Figure 5: Step-by-step grid reconstruction process for a 15-node test system. The sequence illustrates the progressive stages of reconstructing the grid topology.