Research

The connectivity structure of an LV network — which customers share conductors, how branches join, where the network ends — is rarely well-documented. Utility GIS records are often incomplete, out-of-date, or simply absent for older network sections.

This research develops methods to infer the topology from smart meter data alone. Smart meters already collect voltage and current time series at customer premises; the challenge is reading the network structure out of those signals without any additional sensors or site visits.

Getting topology right matters because almost every downstream analysis — load flow, loss estimation, fault location, phase identification — depends on knowing how customers are connected.

Distinct from topology, impedance estimation focuses on the electrical properties of conductors — specifically the resistance and reactance of the cables connecting customers and substations. These values determine how voltage drops across the network, how losses accumulate, and how power flows under varying load conditions.

As with topology, utility records for cable impedances are often unreliable or missing. This research explores how to estimate impedances from smart meter observations, either jointly with topology or as a follow-on step. Accurate impedance estimates also help validate and refine topology inferences.

Distribution transformers are the gateway between the medium-voltage grid and LV customers. Knowing how a transformer is loaded, what voltage it is presenting, and whether it is under stress is valuable for network management — but instrumenting transformers directly is expensive and often impractical at scale.

This research develops virtual sensing methods: estimating transformer-level quantities from smart meter readings collected downstream at customer premises. No hardware changes are required at the transformer. The goal is to make transformer monitoring accessible for networks where deployment of dedicated sensors is not economically viable.

LV networks are three-phase, but individual customers connect to just one. Which phase a customer is on affects load balance, voltage quality, and the interpretation of their meter readings. In practice, phase records are frequently incomplete or incorrect — particularly for older networks where connections have changed over time.

This research infers phase assignment from the correlation structure of voltage time series across the smart meter fleet. Customers on the same phase exhibit distinct patterns compared to those on different phases. Phase grouping extends this to cluster all customers simultaneously by inferred phase membership, without requiring any direct measurement or prior knowledge of the network topology.

Understanding the current state of a network is useful; understanding how it responds to change is essential for planning. This research builds tools for scenario exploration — stress-testing networks against hypothetical futures to understand their sensitivity and identify potential problems before they occur.

Scenarios of interest include:

• What happens to voltage profiles and transformer loading as rooftop solar uptake increases in a feeder area?
• How does the network behave as EV adoption grows and charging patterns shift customer load curves?
• If a large industrial customer's consumption changes, how does that propagate through the network?
• What is the impact of adding or rerouting customers between substations?

The aim is to give operators and planners an interactive tool for probing hypothetical changes — not just reporting current state, but exploring what-if futures alongside the analytics methods developed in the other research areas.