Separating technical from non-technical loss without full conductor data.

Why the split matters

Technical loss is physics: energy dissipated as heat in conductors, transformers and connections. Non-technical loss is everything else: unmetered connections, meter tampering, billing gaps, faulty CTs. The two demand opposite responses. Technical loss is an engineering investment decision, reconductoring or rebalancing. Non-technical loss is an operational and social one, inspection, enforcement and customer engagement. An operator who cannot separate them is choosing between those responses blind, and the money spent on the wrong response is simply gone.

What the physics actually requires

Computing technical loss from first principles needs the network as built: conductor types and lengths per segment, transformer ratings and their load-dependent loss curves, and a load profile per segment to integrate against, since resistive loss scales with the square of current and reacts sharply to how load is distributed in time and across phases. On many distribution networks, and on most rural ones, part of this record simply does not exist. The as-built drawings disagree with the field, the conductor on a segment was replaced during a repair and never recorded, and the transformer's nameplate is the only test data it has ever had.

The tempting move is to substitute typical values: a standard conductor assumption, a textbook transformer curve, a generic load shape. The computation then completes, produces a technical loss figure with decimal places, and the balance of the energy gap gets labelled non-technical. Every substitution made along the way has quietly migrated into an accusation, because the residual is what inspection teams get sent to chase.

What can be said honestly

Our approach is to compute what the evidence supports and report the rest as unresolved. Where the physical record exists for a segment, technical loss is computed and carries its inputs. Where it does not, the system reports the gap, its reason, and what would close it: a field survey of that feeder, a conductor audit, a transformer test. The energy gap for the affected zone is then presented as a whole, labelled as unseparated, rather than being forced through an assumed split.

Meter-level evidence can still narrow things down without conductor data. Feeder-level balancing isolates which zones the losses live in. Temporal signatures help: technical loss tracks load squared, while a constant offset at night on a lightly loaded feeder points elsewhere. Consumption discontinuities at individual meters, sudden sustained drops with no tariff or season to explain them, are meter-level facts that stand on their own. Each of these is evidence with a basis, and each is presented as what it is, rather than being laundered into a single confident percentage.

The number that survives scrutiny

The test we design against is not the monthly report. It is the meeting where the loss figure is challenged: by a regulator, a lender, or the community whose feeder is being singled out. A figure assembled from recorded physics and labelled evidence survives that meeting. A figure resting on silent substitutions does not, and it takes the credibility of every other number in the system down with it. This is the same rule that runs through all of our work: where the input is missing, the system says so, because the alternative is a number that works until someone asks how it was made.