Distribution Transformer Overload and Hot-Spot Monitor for Power Utilities

The problem: transformers fail in the heat, and nobody saw it coming

Every summer, neighbourhoods in Pakistani cities lose power for hours or days because a distribution transformer has burnt out. Air conditioner load climbs in the evening heat, one phase carries far more than the others, the oil runs hot, insulation ages fast, and eventually the transformer fails. Replacing it takes time and money, and the street is in the dark until then.

Most distribution transformers have no monitoring at all. Utilities learn about overloading from complaints after the failure. Yet the physics is well understood: winding temperature, driven by load and ambient temperature, determines how fast insulation ages. Measure the load and oil temperature, apply a standard model, and you can see which transformers are being cooked.

What already exists and the gap

• Substation monitoring systems watch large power transformers closely, using detailed thermal models, but cost far too much for thousands of small distribution units.
• Commercial distribution transformer monitors exist from several vendors, usually as part of large smart grid contracts.
• Smart meters at homes show household use, not the loading of the transformer they share.
• Utility load surveys measure transformer loads occasionally by hand, not continuously.

The gap is a low-cost, clip-on unit that a lineman can install without opening anything, and that reports risk in a form a sub-division office can act on before summer peak.

Inspiration

IEEE Standard C57.91, the guide for loading mineral-oil-immersed transformers, gives the thermal model: the hottest spot temperature equals ambient temperature, plus the rise of top oil over ambient, plus the rise of the winding hot spot over top oil, each depending on load. It also gives the ageing acceleration factor that turns hot-spot temperature into loss of insulation life. A published IEEE paper, "Oil-immersed transformer temperature monitoring by ANSI/IEEE C57.91 model", implemented this model on a microcontroller. Your project brings the same idea to a modern, connected, low-cost design.

What you build

• The sensor unit clips current transformers around the low-voltage cables and straps temperature probes to the tank and to free air in shade.
• The firmware samples current waveforms, computes RMS values, reads temperatures and sends a compact report every few minutes, storing data during network outages.
• The server runs the thermal model per transformer, estimates hot-spot temperature and loss of life, and flags phase imbalance.
• The utility engineer sees a ranked list of transformers at risk and which phases to rebalance.

The technically interesting part

• Measuring current safely. Split-core current transformers clamp around cables without breaking any circuit. Design the burden resistor and analogue front end, calibrate against a clamp meter and check accuracy at low and high load.
• The thermal model. Implement the C57.91 equations with exponential time constants, then calibrate the unknown parameters (rated rises, time constants) for each transformer by fitting predicted to measured top-oil temperature.
• Phase imbalance. Single-phase house connections pile onto one phase. Quantify imbalance and estimate the reduction in hot-spot temperature if loads were moved.
• Reliable telemetry. Outages are exactly when the device loses power. Use a small battery or supercapacitor, timestamp events, and send buffered data when the network returns.
• Field robustness. Heat, dust, rain and electrical noise around transformers are harsh. Choose enclosures, cable glands and filtering carefully, and record failures honestly.

How to evaluate it

• Current accuracy: error against a calibrated clamp meter across the load range.
• Model accuracy: predicted against measured top-oil temperature over several days, on the bench or in the field.
• Uptime: share of expected reports received during the trial, including across outages.
• Usefulness: whether the risk ranking matches the engineers' own knowledge of problem transformers.

Suggested plan

• Weeks 1 to 4: current sensing front end, calibration, temperature probes.
• Weeks 5 to 9: firmware, buffering, GSM or LoRa uplink, MQTT ingestion.
• Weeks 10 to 14: thermal model, calibration method, dashboard.
• Second semester: bench trial with a small transformer and controlled loads, enclosure, supervised field trial with a utility if possible, report.

Path to a product

Distribution companies across Pakistan and South Asia lose money and goodwill on every burnt transformer. A low-cost monitor, sold or leased to utilities and to housing societies and industrial estates that own their own transformers, is a credible business. The project also builds the embedded, power and data skills that utilities and energy companies hire for.

Scope warning

Never open a transformer, touch live terminals or climb a pole yourself; all field work must be done by authorised utility staff, with your sensors installed by them. Do most development on the bench with a small transformer and safe loads. Do not attempt dissolved gas analysis or any oil sampling.

Learn what you need on Sefism

• Embedded systems and IoT roadmap: sensors, firmware and connectivity.
• Embedded systems and IoT cheat sheet: ADC sampling, power and communication basics.
• ESP-IDF Programming Guide: building the firmware on the ESP32.
• MQTT Essentials: reliable telemetry over unreliable networks.
• Grafana Documentation: the utility dashboard.
• How do I become an embedded systems engineer?: where this project fits in your career.

Get help with this idea

Safety and calibration are where this project succeeds or fails. Tauseef can help you plan the bench setup, approach a utility for a supervised trial and scope two semesters realistically. If you are new here, create a free account first.

• Discuss this idea with a mentor
• Send a query to Tauseef
• Book a 1:1 session