Three Solutions to Phase Unbalance in LV Distribution Networks

The Problem: What Phase Unbalance Actually Is

In a low-voltage distribution network, three-phase power reaches a neighbourhood transformer. From there, individual homes are connected one at a time — each on a single phase. In theory, connections should be distributed evenly: roughly one-third of homes on L1, one-third on L2, one-third on L3. The sum of the three currents would cancel out in the neutral, thermal stress would be shared equally across the transformer's three windings, and the system would work as designed.

In practice, this does not happen. When a new subscriber is connected, the electrician connects to whichever phase is most accessible on the pole — usually the lowest wire, the closest one, the one already wired to a nearby house. There is no systematic check of which phase is least loaded. Over time, one phase ends up carrying two or three times the current of the others.

This is the phase unbalance problem. It is not a theoretical concern: the REGIDESO field teams in Bujumbura deal with its consequences daily.

What Actually Happens to the Transformer

A concrete example from the field: a 160 kVA, 30/0.4 kV distribution transformer with L1 drawing 180 A, L2 drawing 120 A, and L3 drawing only 40 A. The transformer is running at its thermal limit on one winding while the other two are underloaded. The equipment effectively loses 20 to 30% of its usable capacity, even though it is nominally rated for the total load.

Three consequences cascade from this situation:

• High neutral current. The vector sum of the three unequal phase currents does not cancel to zero — it returns through the neutral conductor as a real current, causing Joule losses and heating that are not billed to any customer.
• Voltage unbalance. The overloaded phase sees its voltage drop (dim lights, motors that struggle to start). The underloaded phases see their voltage rise, which can damage appliances connected to them.
• Premature transformer failure. The single overloaded winding runs hot. The dielectric oil degrades faster in that zone, the insulation softens, and eventually the winding fails. A transformer that should last 20+ years can be destroyed in a few years in a heavily unbalanced network.

Currently, this is managed reactively. When a neighbourhood starts seeing repeated blown fuses or visible voltage drops, a team is sent. They measure current on each phase with a clamp meter — usually in the evening during peak consumption between 18:00 and 21:00 — note the readings on paper, and manually reassign some connections. This is slow, imprecise, and the problem returns as soon as new subscribers are added.

Solution 1 — The Pragmatic Approach: An Automated Excel Tool

The first solution requires no new hardware. It replaces the empirical, experience-based rebalancing with a mathematical decision support tool built in a spreadsheet.

The workflow is simple: technicians take their existing clamp meter readings and enter the current values for each phase into the Excel file. An algorithm then computes the optimal reconnection plan — which specific loads to move from which phase to which — to minimise the standard deviation of phase currents. The output is a direct instruction: “Move approximately 30 A from L1 to L3.”

Cost: Essentially zero. The company already has laptops and clamp meters. The only investment is the time to build the tool (a few days) and the time to train technicians to use it (one to two hours per person).

Limitation: Clamp meter readings are point-in-time. A transformer that looks balanced at 14:00 may be severely unbalanced at 19:00 when consumption peaks. This tool improves the quality of each intervention but does not address the underlying dynamic behaviour of the network.

Conclusion: This is the fastest solution to deploy. It turns an intuitive, error-prone process into a reliable one. For the most critical transformers, it can provide immediate relief.

Solution 2 — The Professional Approach: Fluke Power Analysers and Process Redesign

The second solution addresses the limitation of point-in-time measurements. It combines professional diagnostic equipment with a structural change to how new connections are managed.

The Equipment

Portable power quality analysers — such as the Fluke 1770 or Fluke 435 series — are clamped onto the transformer's BT panel and left in place for 48 to 72 hours. Over that period, the device records the full load profile: three-phase currents and voltages, neutral current, harmonic content, power factor. The result is a complete picture of how the transformer actually behaves over a typical consumption cycle, not just a snapshot from one afternoon.

This data answers the question that the clamp meter cannot: when is the unbalance worst, how severe is it, and which phase needs how much load shifted?

The Process Change

Equipment alone is not enough. The root cause of unbalance is the absence of a phase assignment protocol at the moment of connection. The process fix has two parts:

• Visual coding on the network. Coloured rings on poles identify which wire corresponds to L1, L2, and L3. Every technician and electrician in the field can immediately see the phase assignment at any point on any feeder.
• Mandatory phase routing at connection time. When a new subscriber is added, the engineering office — based on the current load profile of that transformer — specifies which phase the connection must go to. The electrician follows a written instruction sheet. The meter is only validated after the correct phase connection is confirmed.

Cost: High. A single Fluke 1770 or 435 unit costs between 5,000 and 8,000 euros. For a utility the size of REGIDESO, acquiring several units represents a significant capital expenditure, plus the cost of technical training and the time to implement the visual coding across the network.

Conclusion: This is the most robust medium-term solution. It treats both the symptom (through accurate diagnostics) and the cause (through process discipline at connection time). It corresponds to international industry standards for distribution network management.

Solution 3 — The Dynamic Approach: A Low-Cost IoT Monitoring Device

The third solution is forward-looking. It does not wait for a problem to become visible — it detects the drift toward unbalance in real time and sends an alert before the transformer is damaged.

The Concept

A small monitoring enclosure is installed permanently at the transformer's BT panel. Inside: a microcontroller (ESP32 or Arduino), three split-core current transformers that clamp around each phase conductor without any electrical interruption, and a GSM module.

The device continuously measures the current on each phase and computes the unbalance ratio. If the unbalance exceeds a defined threshold — say, 15% — for more than two consecutive hours, it automatically sends an SMS alert to the maintenance team. The on-call engineer can then decide whether to dispatch a crew immediately or schedule a rebalancing intervention before the situation deteriorates further.

Why Split-Core Transformers Matter Here

Split-core (or open-core) current transformers are non-intrusive: they clip around an existing conductor without disconnecting it, without interrupting service, and without requiring any modification to the installation. This is essential for a deployment on live distribution infrastructure where planned outages are costly and difficult to schedule.

Cost and Development Path

A single enclosure, built from off-the-shelf components and programmed in-house (or through a local tech partner), would cost between 400 and 600 euros per unit. The main investment is not hardware — it is the software development time and the ongoing cost of SIM card subscriptions for the GSM alerts.

This solution is a proof of concept. A logical next step would be to aggregate the alert data on a central server, building over time a picture of which transformers in the network are consistently stressed — feeding directly into a predictive maintenance workflow rather than waiting for failure reports.

Conclusion: This moves the operation from corrective to predictive maintenance. The hardware cost per unit is low, the deployment is non-intrusive, and the long-term value — catching transformer stress early, before a 160 kVA unit burns out — is substantial. It does require internal R&D capacity or a development partner.

Comparing the Three Solutions

The three approaches are not mutually exclusive. They sit at different levels of a maturity ladder and address different parts of the same problem:

A realistic deployment path: start with Solution 1 today — no procurement, no waiting. Use it to identify the five or ten most critical transformers in the network and rebalance them. Use the freed-up engineering capacity to develop the IoT device (Solution 3) as a PoC on those same transformers. Use the resulting long-term data to build the case for acquiring one or two Fluke units (Solution 2) for periodic network-wide audits.

The Bigger Picture: From Corrective to Predictive

A transformer that is properly balanced can double its service life — from a few years in a saturated, unbalanced network to over 20 years under normal conditions. A 160 kVA distribution transformer costs several thousand euros to replace, plus the labour and logistics of a workshop repair cycle and a field installation. Multiply that across a network with dozens of transformers in stressed zones, and the financial case for active load management is clear.

But the more important shift is organisational. Corrective maintenance — fixing things after they break — is expensive and operationally consuming. Every transformer failure is an emergency: a neighbourhood without power, a workshop job that pushes other work back, a team scrambling to find a replacement unit. Predictive maintenance turns that emergency into a scheduled intervention. The failure is anticipated, the crew is prepared, the replacement is sourced in advance.

The technical tools to make this shift exist. The Excel algorithm can be built this week. The IoT device can be prototyped in a few months. The Fluke analysers are commercial off-the-shelf equipment. None of these solutions require waiting for the large-scale 2027 modernisation programme. They can start creating value now, with the teams and infrastructure already in place.

That is, in the end, what this project is about. Not the technology itself, but the transition from reacting to anticipating — and the argument that this transition is within reach, today, at a cost that is justified by the damage it prevents.