Every distribution company knows the gap between the energy it buys and the energy it gets paid for. The industry calls it AT&C loss: aggregate technical and commercial loss. Reducing it is one of the fastest ways for a DISCOM to improve its finances, but only if you can tell what kind of loss you're looking at.
Three different problems under one number
- Technical losses are the physics of moving power: heat in lines and transformers. They can be reduced with network upgrades, but never to zero.
- Commercial losses are energy that reached consumers but was never correctly billed: theft by hooking or meter bypass, tampering, wrong tariff categories, faulty meters and billing errors.
- Collection losses are bills issued but not paid.
Each needs a different response. Sending vigilance teams after what is really a faulty meter wastes time and damages trust with honest consumers.
Start at the distribution transformer
The most useful unit of analysis is the distribution transformer. If you know how much energy entered it and how much was billed to every consumer it serves, the difference, minus a realistic technical loss, is the unexplained loss for that small area. Smart meters on transformers and consumer premises make this energy balance possible every billing cycle instead of once a year.
What separates theft from faults
Within a high-loss transformer, consumer-level signals point to the likely cause:
- Load at night with very low billed units can suggest a bypass.
- A sudden, lasting drop in consumption after a tamper event points to meter interference.
- A load profile that looks commercial on a domestic connection suggests category misuse.
- Gaps in readings followed by flat values usually mean a meter or communication fault, not theft.
- Vacant premises explain low consumption without any wrongdoing.
Making leads field-ready
Field teams need more than a risk score. A good lead says which consumer, why they were flagged, and what to check on site. Inspection results should flow back into the system, so the model learns which signals actually led to recoveries in your network.
Where to begin
A pilot on 12 months of historical meter and billing data for a few feeders is usually enough to show whether transformer-level analytics will pay off, before any change to live operations.
GridSentinel is our product for exactly this workflow.