Export data: microgeneration, SEG and storage

Export is not negative import. A home with solar panels does not have a meter that counts backwards at noon. It has a meter with more than one register, and exported electricity is counted on a separate one. The import register measures energy drawn from the network and only ever advances. The export register measures energy pushed back to the network and also only ever advances. Two counters, both moving forwards, measuring two different physical flows through the same connection.

This matters because the identifier follows the register. In module 3 you met the MPAN as the reference for an electricity meter point. A site that both consumes and exports has two of them: the ordinary import MPAN and a separate export MPAN. They are distinct meter points in the registration systems, they can carry different market arrangements, and they are settled separately. If you are reconciling a solar customer and you have one MPAN in your hand, you are holding half the story.

The physics behind the two registers is straightforward once you picture the moment of measurement. At any instant the site is either drawing net power from the network or pushing net power out to it. On-site consumption is served by the panels first, so what reaches the meter is the difference. Between 07:00 and 07:30 in January the panels contribute almost nothing and the whole household load appears on the import register. At 13:00 in June the panels may cover the whole load and the surplus flows outward, so the import register stands still and the export register advances. Across a single day both registers can advance, each in its own half hours.

Two practical consequences follow. First, self-consumption is invisible to both registers. The kilowatt-hours the household generates and uses itself never cross the meter in either direction, so no settlement system and no export tariff ever sees them. Any analysis that tries to infer a home's solar generation from its metered data is inferring only the exported surplus. Second, an export register that reads zero is not proof that there is no generation. It may mean the export meter point was never set up, or that the register exists but the export MPAN was never registered to a supplier, which is a data problem rather than a physics one.

Gas has no equivalent. Gas meter points measure a one-way flow of volume and there is no domestic export arrangement to mirror, which is one of the several places where the electricity and gas data models stop being symmetrical.

The takeaway to carry forward: two registers, two MPANs, one connection, and never a negative number. Every commercial arrangement in the rest of this module is built on top of that separation.

The Smart Export Guarantee (SEG) is what turned the export register from a technical curiosity into a billing input. Under the scheme, electricity suppliers above a size threshold, set at 150,000 domestic customers, are obliged to offer a tariff that pays small generators for the electricity they export to the grid. Suppliers below that threshold may offer a SEG tariff voluntarily. Ofgem sets the obligation and the rate is left to the market, so tariffs differ between suppliers and a household can hold its export contract with a different supplier from its import contract.

The point that matters for a data professional is the prerequisite chain. A SEG payment cannot be made from goodwill or from an annual estimate. It requires a metering arrangement capable of recording export at half-hourly granularity, an export MPAN registered so that the exported volumes have somewhere to land, and a contract between the generator and a SEG licensee. Miss any one of the three and the payment does not happen, however many kilowatt-hours the panels produced.

This is why the SEG changed the incentives around export data quality. Before it, an export register was mostly an engineering detail: nobody lost money if it was set up late or read badly. Once export volumes carry a price, every weakness in the chain becomes a commercial dispute. A missing export MPAN is a customer who is not being paid. A meter fitted without an export register is an installation that has to be revisited. A read that fails validation is a payment that slips a month. Export data became commercially load-bearing, and load-bearing data gets governed.

It also reshaped the relationship between the household and its supplier. The import side of the account and the export side of the account are separate positions with separate identifiers, separate volumes and separate prices. Reconciling them means joining two meter points that share a postal address but nothing else in the data model. If you have ever wondered why a solar customer's statements can be so hard to follow, this is the reason.

The takeaway to carry forward: SEG is a data obligation dressed as a tariff. Half hourly export metering, a registered export MPAN and a SEG contract are the three things a payment stands on.

Module 9 traced the import chain through Market-wide Half-Hourly Settlement (MHHS): retrieval by the Smart Data Service, shaping by the Load Shaping Service where a read is missing, aggregation by the Market-wide Data Service, all wired through the Data Integration Platform (DIP). Export meter points travel the same road. The MHHS migration that began on 22 October 2025 and runs to May 2027 moves export MPANs into the new arrangements alongside import MPANs, and the same Settlement Period granularity applies to both.

The practical effect is that export stops being a rounding error in settlement and becomes a settled volume with the same standing as consumption. Before market-wide half-hourly arrangements, domestic-scale export was thinly measured and coarsely handled. Under MHHS an exporting site contributes an actual half-hourly volume to its GSP Group in the direction that reduces the net demand of that group. The settlement calendar changes apply equally: today Initial Settlement lands 16 working days after the settlement day, and at the M16 timetable cutover due early in July 2027 that compresses to about seven working days with final settlement moving from roughly fourteen months to about four. Export positions firm up on the same schedule as import positions.

Storage is where this gets genuinely interesting, because a battery is the only common asset that changes direction on purpose. A battery charging from the network is import. The same battery discharging to the network is export. Over a day it advances both registers, and its pattern is dictated by prices and instructions rather than by weather or occupancy. That makes storage a much harder forecasting subject than solar, and it makes the data trail more important: you cannot reconstruct what a battery did from a daily total, only from the half-hourly sequence.

Co-location sharpens the question further. Put a battery and a solar array behind a single grid connection and the metering arrangement decides what can be seen and what can be paid. If only the connection point is metered, the data shows a single net flow and there is no way to tell solar output from battery discharge, or to tell whether the battery charged from the panels or from the network. If each asset is metered separately, the two can be settled and rewarded on their own terms, at the cost of more metering and more registration. This is an open design question in the industry rather than a solved one, and it is a live example of a rule that decides in advance what analysis will ever be possible.

A related point applies at the boundary between microgeneration and the wholesale market. Larger embedded generation connected to the distribution network is visible to the distribution network operator through its Embedded Capacity Register, which records connected and accepted capacity. That register tells you what is installed. It does not tell you what any of it produced in a given half hour. Capacity data and output data are different data products, and confusing them is one of the more common errors in embedded generation analysis.

Everything so far has assumed a meter. Most GB solar generation is not metered in a way that reaches the system operator. Rooftop arrays sit behind supply points on the distribution network, and their effect on the transmission system is to reduce measured demand rather than to appear anywhere as measured generation. NESO cannot balance the system on a quantity it cannot see, so it estimates it.

PV_Live is the service that does this. Sheffield Solar runs it with NESO. It takes readings from a sample of monitored solar sites, combines them with what is known about installed capacity across the country, and produces an estimate of GB solar output resolved to GSP level and to the half hour. The output is used operationally: it feeds the demand picture that balancing decisions are made against, and it appears in the generation mix figures that downstream services build on.

The reason to teach this in a data course is the principle rather than the service. Where meters end, models begin, and both are data products. The difference is what travels with them. A settled export volume carries an identifier, a validation history and a settlement run that can be re-run. A modelled solar figure carries an estimation method, a sample, an assumed capacity base and an error band. Treating the second as though it were the first is how forecasts get over-trusted and how reconciliation exercises fail in ways nobody can explain afterwards.

The estimate also has known weak points that are worth naming. It depends on the accuracy of the installed capacity picture, which lags real installation activity. It depends on the sample being representative of the fleet, geographically and by orientation. It degrades in conditions where the sample and the fleet behave differently, such as patchy cloud across a region. None of that makes the estimate wrong. It makes it a measurement with uncertainty attached, which is exactly what it is published as.

There is a direction of travel here worth holding on to. As more export becomes metered, registered and settled half-hourly through MHHS, the share of generation that has to be modelled shrinks. Estimation does not disappear, because behind-meter self-consumption stays invisible to every register by construction. But the balance between the measured and the modelled shifts, and the quality of the national picture depends on the unglamorous work of getting export MPANs registered and export registers read correctly.