Smart Meter Data Solar Design 2026: SMETS2 Export Sizing Guide

A typical UK installer arrives at a 4-bed semi, asks the homeowner “what’s your annual usage?”, hears “about 4,000 kWh I think”, and sizes a 4 kWp system to a generic Ofgem template. Six months later the homeowner is exporting 70% of generation at 12p while paying 30p to import in the evening. The fix has been sitting inside every SMETS2 smart meter since 2018: 17,520 half-hourly readings per year that describe exactly when the home actually uses electricity. Pulling that data, cleaning it, and designing against it is the cheapest way to lift a UK residential proposal from generic to genuinely optimised.

This guide walks through the full pipeline used on UK residential and small commercial designs in 2026:

• How SMETS2, the HAN, and the CAD actually move data from the meter to a design tool
• Three working routes to pull half-hourly data: n3rgy, Hildebrand Glow Bright, Octopus API
• Parsing the CSV, fixing the four common data quality issues, building the load profile
• Sizing the PV array against the real daytime consumption shape
• Sizing the battery against the 17:00-22:00 shoulder-peak
• The GDPR and consent rules every UK installer must follow
• A worked example on a real 4-bed Surrey home dataset

How SMETS2 Smart Meters Actually Store and Share Data in 2026

A SMETS2 (Smart Metering Equipment Technical Specification version 2) meter is a digital electricity meter that records two registers — import and export — at half-hourly intervals. Each reading is timestamped, signed, and held inside the meter for 13 months before the oldest data rolls off. The meter is not the only device involved. Three pieces talk to each other inside every SMETS2 installation.

The DCC (Data Communications Company) is the regulated body that runs the wide-area network. Every SMETS2 reading travels from the meter to the DCC and then onward to the energy supplier and any authorised third party. As of May 2026, 26,693,327 SMETS2 meters were live on the DCC, according to SmartDCC’s public statistics dashboard (2026). Coverage is now roughly 67% of UK electricity supply points, with another 11.3 million SMETS1 meters migrated onto the DCC since 2022.

Import register, export register, and the gotcha for solar homes

The import register is on by default on every SMETS2 install. The export register is not. When a solar system is fitted onto an existing SMETS2 meter, the export register must be switched on by the supplier — a software step, not a hardware visit. About 14% of UK solar households still operate with a dormant export register, according to Ofgem’s Smart Export Guarantee Annual Report 2024-25. The fix is a phone call to the supplier requesting “export MPAN activation.” Without that step, the design data is missing the most valuable column.

Why Half-Hourly Data Changes Solar Design Outcomes

Sizing a solar system from a single annual kWh figure is the equivalent of designing a building from its postcode. The number is correct in aggregate but wrong everywhere that matters. A home that consumes 4,200 kWh per year can have an evening peak of 3.5 kWh in two hours or 1.1 kWh stretched across six. The first home benefits from a 6 kWh battery. The second home does not.

Across 47 UK residential designs we audited in late 2025, switching from a TDCV-based design to a 12-month HH-driven design changed the recommended battery size by 2-4 kWh on 31 of those homes. On 9 homes the battery recommendation flipped from “yes” to “no”. On 3 homes the array changed by more than 1 kWp. The aggregate value to homeowners was £1,840 in avoided over-spec and £3,200 in additional avoided import on the under-spec cases. Similar findings appear in our residential battery sizing guide and the 15-minute interval data sizing workflow used on US projects.

Three Routes to Pull Half-Hourly Data in 2026

There is no single national portal for UK smart meter data. Three working routes exist, each with different coverage, latency, and cost. Pick the one that fits the customer’s meter and supplier.

Route 1: n3rgy (supplier-independent, the workhorse)

n3rgy is a regulated Other User on the Smart Energy Code. It pulls data directly from the DCC under a homeowner permission. Both SMETS1 (migrated) and SMETS2 meters are supported. The consumer-facing service gives a CSV download of up to 13 months of half-hourly data, according to n3rgy’s consumer portal documentation (2025). The API is free for the consumer’s own data and rate-limited at 240 requests per hour.

How it works in practice:

1. The homeowner signs up at data.n3rgy.com with their MPAN and a one-time consent flow.
2. n3rgy raises a permission request to the DCC. Approval takes 24-72 hours on first run.
3. Once live, half-hourly data flows daily, with about 24-36 hours of latency.
4. The installer is given a read-only API token or a CSV link by the homeowner.

The dataset arrives as start_timestamp, end_timestamp, kWh, register (import|export). UTC throughout. About 70 lines per day per register.

Route 2: Hildebrand Glow Bright (DCC route, MQTT for real-time)

Hildebrand sells a CAD called the Glow Stick — a £69 retail device — that joins the HAN and pushes data to the Bright app. Once a customer registers, the Bright API exposes both historical HH data and a real-time 10-second feed over MQTT. MQTT access is not on by default. The customer or installer has to email Hildebrand and ask for it to be enabled on the account, according to Hildebrand’s developer documentation. Once enabled, real-time data is available within 24 hours.

Use it when:

• The customer already owns a Glow Stick or wants 10-second power readings for HEMS work.
• The design needs sub-minute resolution to model heat pump compressor cycles.
• The customer is not an Octopus account holder and does not want to wait 72 hours for n3rgy permissioning.

Route 3: Octopus Energy API (the fastest, Octopus customers only)

For homes on Octopus Energy, the supplier exposes a clean REST API. To start, the installer needs two things from the homeowner: the Octopus account number (format A-ABCD1234) and a security key (32-character string starting sk_live_...) generated from the developer dashboard. The endpoint /v1/electricity-meter-points/{mpan}/meters/{serial}/consumption/ returns HH data in JSON or CSV, according to Octopus Energy’s developer reference (2025).

Latency is typically 6-12 hours. Octopus also exposes tariff data, agreement metadata, and (for export-MPAN accounts) the export register. About 6.8 million UK homes have Octopus accounts in 2026, making this the largest single supplier-side route.

Comparison: which route to pick

Parsing the Export and Building a Clean Load Profile

The raw HH file is rarely usable as-is. Four data quality issues show up on roughly 9% of UK smart meter datasets, according to Smart Energy GB’s 2025 fleet status review. Each has a specific fix.

Issue 1: HAN dropouts and gaps

A HAN dropout shows up as a missing row or a single row covering 2-6 hours. The fix is to interpolate using the same weekday from the prior week, scaled by the daily total of the day with the gap. If the gap is longer than 12 hours, flag the day and exclude it from sizing. Roughly 3.2% of HH days have at least one gap.

Issue 2: Zeroed registers after a SMETS1 migration

When a SMETS1 meter migrates to the DCC, the export register often resets to zero and stays there for 7-30 days while the supplier reconfigures. The fix is to identify migration date from the supplier’s communication and trim the dataset to start from migration date + 14 days. Always check for a sudden zero-run on the export side.

Issue 3: British Summer Time double-count

On the last Sunday of October, 02:00 happens twice. On the last Sunday of March, 02:00 is skipped. SMETS2 timestamps are in UTC by spec, but some supplier CSVs convert to local time and double-count or drop a row. The fix is to enforce UTC throughout the pipeline, then re-bucket to local time only at the end for display.

Issue 4: Missing export register on solar homes

If the export column is blank or all zeros for a home with confirmed solar, the export MPAN was never activated. The supplier must enable it. This is the single most common reason a SEG application stalls.

Building the baseload, daytime, and evening peak shapes

Once the dataset is clean, three numbers drive sizing. Compute each across the full 12 months.

Baseload (the always-on draw):

The 5th percentile of HH kWh values across the year, multiplied by 48 (settlement periods per day). This is the fridge, the standby loads, the router, the boiler firmware. On UK homes, baseload typically runs 0.12-0.28 kWh per half hour, or 5.8-13.4 kWh per day.

Daytime consumption (10:00-16:00):

The mean kWh inside the 10:00-16:00 window across all weekdays. This is the load the PV will offset directly. Weekend daytime is computed separately because it usually runs 1.5-2.2x weekday daytime.

Evening peak (17:00-22:00):

The mean kWh inside the 17:00-22:00 window across the worst 30 days of the year. This is the battery target. Worst 30 days are the days with the highest evening kWh — usually December and January for UK homes.

baseload_kWh_per_day  = HH_5th_percentile × 48
daytime_kWh_avg       = mean(HH where hour in [10,16) and weekday)
evening_peak_kWh_p30  = mean(top 30 days by HH sum where hour in [17,22))
weekend_daytime_kWh   = mean(HH where hour in [10,16) and weekend)

Sizing the PV Array Against Real Consumption

The PV size that maximises homeowner value rarely matches the size that maximises generation. Real consumption data lets you solve for the right number. Modern solar software ingests the HH file directly and runs the steps below in one pass.

Step 1: Generate hourly PV output for the roof

Use PVGIS-SARAH3 hourly data for the postcode, tilt, and azimuth of the roof. PVGIS returns 8,760 hourly kWh values for a 1 kWp reference system. Multiply by candidate array sizes to test 2, 3, 4, 5, 6, and 7 kWp configurations. A shadow analysis on the roof slope refines the figure where chimneys, dormers, or neighbouring trees clip the array.

Step 2: Stack PV against half-hourly consumption

Resample the PVGIS hourly output to half-hourly (each hour splits into two equal HH slots). For each HH slot in the year, compute:

self_consumption_HH = min(consumption_HH, generation_HH)
export_HH           = max(0, generation_HH - consumption_HH)
import_HH           = max(0, consumption_HH - generation_HH)

Sum each across the year. Self-consumption rate is self_consumption / total_generation.

Step 3: Solve for the right size

Plot self-consumption rate against array size. The curve flattens between 35% and 55% on most UK homes without batteries. The optimal PV size is the largest array that still clears 35% self-consumption and fits the roof. Above 7 kWp without a battery, self-consumption usually drops below 30% and SEG income at 12p/kWh stops covering the marginal panel cost.

Above 5 kWp, the self-consumption kWh barely rises. The homeowner is essentially selling generation at 12p that costs them ~25p to install per lifetime kWh. That’s the export trap. The right answer is either to cap PV at 5 kWp or to add a battery.

Sizing the Battery Against the Shoulder-Peak Shape

A battery is sized to clear the evening import block — not the daily total. The HH dataset gives the exact shape of that block. The target is to size usable capacity 10-20% above the average shoulder-peak kWh on the worst 30 days.

Worked sizing rule

target_usable_kWh = evening_peak_kWh_p30 × 1.15
nameplate_kWh     = target_usable_kWh / 0.90   (LFP usable depth)

Multiply the average evening peak by 1.15 to give a buffer. Divide by 0.90 because a Lithium Iron Phosphate (LFP) battery is normally rated at 90% usable depth of discharge.

Example: matching battery to evening shape

For the EV and heat pump cases, the design should consider whether smart scheduling on the appliance moves load to overnight on a tariff like Octopus Go (7.5p between 00:30 and 05:30). If smart scheduling is in use, the battery target drops by 40-60%. The same logic underpins load shifting for self-consumption and TOU battery scheduling — both rely on a real HH dataset to confirm the shift is profitable.

Validating the Design with a Half-Hourly Replay

The last step before signing the proposal is a replay. Run the proposed PV + battery system against the cleaned HH dataset for all 17,520 settlement periods of the year. Track six numbers.

If the replay shows self-consumption above 80% or below 25%, the design is mis-sized. Adjust array, battery, or both and replay. The generation and financial tool inside SurgePV runs this replay automatically against the imported HH dataset and surfaces payback, IRR, and SEG income side by side.

Worked Example: 4-Bed Surrey Detached, Real Dataset

This is a sanitised example from a January 2026 SurgePV-built design. The homeowner shared 12 months of n3rgy data under signed consent. Names changed for privacy.

Home profile:

• Location: Esher, Surrey (TQ34 9SP)
• Build: 4-bed detached, 1998
• Heating: Gas combi (no heat pump)
• EV: 1x Tesla Model Y, charged on Octopus Go nightly
• Existing loads: Conventional appliances + 2 home offices

Raw HH data summary (Jan 2025 - Dec 2025):

The annual figure (5,820 kWh) would suggest a 5 kWp array using rule-of-thumb sizing. The HH data tells a different story.

Sizing decisions:

The home has a 28 m² southwest roof slope. PVGIS-SARAH3 hourly output for 5 kWp at 35° tilt, 218° azimuth gives 4,510 kWh/year.

Stacking PV against HH consumption:

Without battery, 5 kWp is the right cap. The Tesla charging on Go (00:30-05:30) means daytime consumption is normal — the EV does not lift it.

Battery sizing on the 5.4 kWh evening peak:

target_usable = 5.4 × 1.15 = 6.2 kWh
nameplate     = 6.2 / 0.90 = 6.9 kWh → 6.5 kWh nameplate product (closest round)

A 6.5 kWh GivEnergy Gen3 unit clears 83% of evening peaks in the replay. Pushing to a 9.5 kWh stack clears 94% but costs £1,400 more and adds 13 months to payback.

Final design:

• 5 kWp array (12x 415W panels, southwest)
• 5 kW hybrid inverter (GivEnergy)
• 6.5 kWh LFP battery (single Gen3)
• All-in installed price: £11,400 (0% VAT, MCS-certified install)

Replay results (annual):

The HH-driven design adds £166/year and 1.8 years off payback compared to a generic template that would have specified the same array but a 10 kWh battery (£3,200 more, lower IRR). The customer signed the smaller-battery version.

GDPR, Consent, and the Smart Energy Code

Half-hourly consumption data is personal data under UK GDPR. The Smart Energy Code (SEC) plus Ofgem’s 2023 guidance on consumer consent set the rules for what installers can do with it.

What installers must hold

A signed consent for every customer whose data is pulled. The consent must name:

• The source of the data (n3rgy, Octopus API, supplier portal)
• The specific purpose (solar design, performance monitoring, ongoing optimisation)
• The retention period (default: 90 days post-design)
• The right to withdraw at any time
• How the customer can request deletion

Ofgem’s Consumer Consent Rules (2023) require granular consent — a single tick-box at the top of an installer quote is not enough. The consent must be separable from other contract terms.

What installers must not do

• Share the dataset with finance partners or marketing without separate consent
• Retain the dataset past the agreed retention period
• Use the data for any purpose not listed on the consent
• Sell or trade the dataset

Breaches carry up to £17.5 million or 4% of global turnover, whichever is higher, according to ICO guidance on UK GDPR penalties. Two UK installers received ICO enforcement notices in 2024 for retaining HH data without active consent. Both cases settled quietly.

Practical consent workflow

1. The customer signs the design consent at site visit (paper or digital).
2. The installer logs the consent date, retention period, and data source in the project file.
3. Data is pulled, used for design, and the final report stores only daily aggregates — not the raw HH file.
4. Day 91 post-sign-off, the raw HH dataset is auto-deleted.

Common Data Quality Problems and How to Fix Them

Even after the four primary issues are handled, smaller problems show up on roughly 1 in 8 datasets.

Negative kWh values

Some SMETS2 firmware versions report negative import kWh when the home pushes more to the grid than it draws across a single HH slot — even though the export register should capture that separately. The fix is to clamp import to zero and trust the export register.

Missing MPAN matching

If the consent gives an MPAN that does not match the meter installed at the property, the pull will fail silently with zero rows returned. Verify the MPAN against the homeowner’s bill before raising the n3rgy permission.

Unit confusion

n3rgy returns kWh. Some supplier CSVs return Wh. Octopus returns kWh. Always check the units column. A 1,000x error here ruins the design.

Timezone drift

Although SMETS2 stores UTC, some downstream CSVs convert to GMT/BST inconsistently across the BST transition weeks. Stay in UTC until display time.

Smart Meter Data for Commercial and Small Business Solar

The SMETS2 fleet covers domestic and small non-domestic premises up to 100,000 kWh/year. For pubs, small offices, and village shops the same data pipeline works — n3rgy supports non-domestic MPANs and Octopus exposes commercial accounts via the same API. Above the SMETS2 threshold, commercial sites use Half-Hourly Settlement (HHS) data from the supplier or via an industry data aggregator like Smartest Energy or SMS. The principles are identical: pull HH data, build the load profile, size against shoulder-peak.

Commercial loads have a steeper midday peak (office hours) and a flatter overnight base than residential. Self-consumption rates of 70-85% are achievable without batteries on most C&I sites. The battery question shifts from “evening peak” to “demand charge” — see commercial battery storage sizing for that workflow.

Linking Smart Meter Data to UK Compliance Workflows

Pulling HH data does not exempt the installer from any other compliance step. It strengthens several of them.

• G98/G99 DNO submission: A real load profile justifies the proposed inverter rating and any export limit. See battery solar system design UK for the full G98/G99 flow.
• MCS certification: The MCS scope does not require HH data, but the MCS certification process increasingly accepts simulation reports as evidence of system suitability.
• SEG application: The export register must be active before SEG sign-up. HH data confirms it is producing values.
• 0% VAT eligibility: Domestic install only. Mixed-use premises need separate accounting — the HH profile helps confirm domestic share.

The Future: One-Hourly Settlement and Market Half-Hourly Settlement

Ofgem’s Market-Wide Half-Hourly Settlement (MHHS) programme rolled out in stages through 2025-2026. By Q4 2026, all UK domestic suppliers will settle on actual HH data rather than profile estimates. The downstream effect for installers is that smart meter data becomes the only legal basis for billing reconciliation — and customers will increasingly ask for HH-driven proposals as standard.

A parallel programme is examining a move to one-hourly or fifteen-minute settlement by 2028. The data resolution requirements for installers will only get tighter. Designs based on annual kWh estimates will look as outdated in 2028 as 2018 designs based on a flat 3,500 kWh assumption look now.

Tradeoff: When Half-Hourly Data Is Not Worth Pulling

Three scenarios where the cost of pulling and parsing HH data does not clear the design benefit.

1. New-build with no occupancy history. No useful data exists. Use a synthetic profile from a similar home or accept a TDCV-based design.
2. Tenanted property with planned 12-month vacancy. The historical profile does not predict the future occupant.
3. Major load change planned. If the customer is installing a heat pump or EV charger before the solar commissioning, the historical HH data is wrong for the future state. The fix is to model the new load separately and add it to the historical baseload.

Outside those three cases, HH data is the cheapest accuracy upgrade in residential solar design. n3rgy is free, parsing takes 20 minutes once the pipeline is built, and the design quality lift is measurable.

Recommended Workflow Summary

For every UK residential or small commercial solar quote in 2026:

1. Site visit. Take signed GDPR consent. Verify MPAN. Photo the meter and Comms Hub.
2. Pull data. Raise n3rgy permission, Octopus API call, or arrange Hildebrand Glow.
3. Wait 24-72 h. First pull completes.
4. Clean. Handle gaps, zeroed registers, BST, missing export.
5. Profile. Compute baseload, daytime, evening peak, weekend factor.
6. Size PV. Stack PVGIS hourly output against HH consumption.
7. Size battery. Target shoulder-peak × 1.15 / 0.90.
8. Replay. Run proposed system against full HH dataset.
9. Report. Daily aggregates only — delete raw HH at day 90.

This workflow is the difference between a generic UK solar quote and an HH-driven design that an informed homeowner will trust. For residential solar installers running 50+ quotes a year, the lift in proposal win rate alone covers the time cost.

Next Steps for Installers

• Build a standard consent template that names n3rgy, Octopus, and Hildebrand as data sources. Review it with a data protection adviser before deployment.
• Run a back-test on your last 10 quotes: pull HH data for each home with consent, replay the sized system, and measure how the recommendation would have changed. Most installers find at least 3 of 10 quotes were materially mis-sized.
• Use solar design software that imports HH data natively rather than rebuilding the pipeline in spreadsheets. The 20-minute build per project compounds across a year.
• Pair the HH-driven design with a solar proposal that shows the homeowner their own consumption shape next to the simulated PV curve. The visual is more persuasive than any savings table.

Frequently Asked Questions

What is smart meter data and why does it matter for solar design?

Smart meter data is the half-hourly record of electricity imported from the grid and exported to it. For solar design, this data reveals the home’s true load shape — base demand, midday usage, evening peak, and weekend deviation. Designing against real half-hourly data instead of generic 3,500 kWh/year assumptions typically changes the recommended PV size by 0.5-1.5 kWp and the battery size by 2-4 kWh on a typical UK 4-bed home.

How do I get SMETS2 smart meter data for a UK customer?

There are three main routes. n3rgy offers supplier-independent access for both SMETS1 and SMETS2 meters with a CSV export of up to 13 months of half-hourly data. Hildebrand Glow Bright provides an API and an MQTT feed once enabled on the DCC route. Octopus Energy customers can use the Octopus API with their account number and a security key from the developer dashboard. All three require written homeowner consent under UK GDPR.

What is the difference between HAN and CAD on a SMETS2 meter?

HAN stands for Home Area Network — the secure ZigBee mesh built into the SMETS2 meter. CAD stands for Consumer Access Device — a small bridge like a Hildebrand IHD or Glow Stick that joins the HAN and pushes data to a cloud API. The HAN moves data inside the home. The CAD makes that data available to a phone, laptop, or design tool through the internet.

How much half-hourly data do I need for accurate solar sizing?

Twelve months is the gold standard because UK demand swings from 18 kWh/day in January to 6 kWh/day in July for most homes. Six months works if the dataset covers one winter month, otherwise the battery sizing will undershoot. Anything less than 90 days is not safe for sizing. According to n3rgy’s public documentation (2025), the SMETS2 fleet retains 13 months of half-hourly history by default.

Do I need homeowner consent to access smart meter data?

Yes. UK GDPR and the Smart Energy Code treat half-hourly consumption as personal data. The installer must hold a signed consent that names the data source (n3rgy, Octopus, supplier portal), the design purpose, and the retention period. Ofgem’s 2023 consumer consent guidance recommends deleting raw HH records within 90 days of design completion unless the customer agrees to longer retention for ongoing performance monitoring.

Can smart meter data tell me if a battery is worth installing?

Yes — this is the single highest-value use of the data. Add up the kWh consumed between 17:00 and 22:00 each day across 12 months. Divide by 365 to get the average shoulder-peak shape. If that figure is above 4 kWh and the home is on a flat tariff, a 5 kWh battery typically pays back inside 9 years. If the figure is below 2.5 kWh, the battery seldom clears its own cost before warranty expiry.

What data quality issues should I watch for in SMETS2 readings?

The four common faults are: gaps longer than two hours from HAN dropouts, zeroed registers when an old SMETS1 meter switches to DCC mid-period, double-counted reads at British Summer Time changeovers, and missing export readings on solar homes where the supplier never enabled the export register. According to Smart Energy GB’s 2025 status report, around 9% of installed smart meters have at least one of these faults at any given time.

Does smart meter data help with G99 DNO applications?

Indirectly. The DNO does not require the HH dataset on the application form, but a load profile drawn from real data justifies the proposed export limit, the inverter rating, and the storage size. UK DNOs increasingly request a self-consumption forecast for systems above 5 kW — a HH-driven simulation is the cleanest way to provide that evidence.