A gap is not an option
Metering values have to be delivered. A value that is missing, whether because the meter did not answer or because validationThe first step in VEE - automated checks of metering data against technical and logical limits to detect obviously erroneous values. rejected it, cannot travel through the chain as an empty slot. SettlementTurning metering data into money. Consumption and production per metering point become what each party owes or is owed. needs a number.
So it is estimated. That is the second step of VEEValidation, Estimation, Editing - systematic process to ensure quality of raw metering data from electricity meters., and it is not a workaround, it is part of the standard.
The calculation runs as a six-step cascade based on historical patterns. The cascade tries the best basis first and falls downwards: the same metering pointThe fundamental unit in Elhub - the unique point in the electricity network where energy is measured and settled, identified by an 18-digit GSRN number. earlier, corresponding periods, patterns from comparable points. The further down the cascade a value was calculated, the weaker the basis.
The labelling matters more than the calculation
This is the real point, and it is easy to miss because the calculation is the interesting part.
An estimated value that is not labelled as estimated is worse than no value at all. It looks like a measurement. It sums like a measurement. It gets compared against other measurements. And the day someone has to explain a deviation, there is no way to tell what was measured from what was calculated.
That is why traceability is the requirement, not precision. The cascade must be able to say which step a value came from, so that an estimated value can always be told apart from a measured one. That property is what makes the number fit to be used.
A good estimationThe second step in VEE - calculation of missing or rejected metering values based on historical patterns, neighbouring meters or statistical models. engine is not the one that guesses best. It is the one that is most honest about how well it guessed.
When a thousand meters go down at once
Individual gaps are everyday work. Mass outages are something else, and they need a rule of their own.
When many meters go down at the same time, typically in a power cut or a communication failure across a whole area, the values are estimated automatically and labelled as estimated. Then five days run for the real values to arrive from the meters. If they come, they replace the estimates. If they do not, the estimate stands.
The deadline exists because both extremes are bad. Without one, an estimate would never become final, and nobody could trust that settlementTurning metering data into money. Consumption and production per metering point become what each party owes or is owed. was finished. With too short a deadline, real values arriving slightly late would never get in.
A missing number and a known outage are not the same thing
This detail is the best part of the whole mechanism, and it is worth understanding.
If the collection system reports "power outage" rather than merely "missing", the value is set to finally estimated immediately, and the five-day deadline lapses.
The reason is simple once you see it: in a power cut there is no real value to wait for. The meter does not hold a value it failed to deliver. The installation consumed nothing, because there was no power. Keeping a deadline open for data that by definition will not arrive is pretending the matter is unresolved.
The system therefore behaves differently depending on how much it knows about the cause, not just about the outcome. That is a higher ambition than most data models hold, and it pays off in fewer open cases.
What this requires of collection
The point above has an upstream consequence worth saying out loud.
The distinction between "missing" and "power outage" only exists if the collection system actually sends it. A head-end systemHead-End System - central system for collecting and managing data from smart meters in AMI networks. that reports all absence identically forces every outage case to be treated as unresolved for five days, even where the cause was known that same night.
It is a good example of how quality in a chain is not decided in the last link. The engine can only be as precise as the messages it receives.
Where we have built it
The six-step cascade, the five-day deadline and the distinction between "missing" and "power outage" are how SEFTATvimenning’s VEE engine: validates, estimates and edits metering series with versioned time series and a full audit trail, before data goes to Elhub. does it. The engine labels every value with the step it came from, so an estimated value can never be mistaken for a measured one further down the chain.
It is verified against simulated market interfaces and not in production at a customer yet. That is worth knowing when reading a description like this one: the mechanism is built and tested, it has not stood through three winters.