Self-consumption percentage is the single most important variable in commercial solar economics — and the one that estimates get most wrong. Every kWh of self-consumed solar saves the full retail electricity tariff (24-32p in 2026), while every exported kWh earns only the SEG rate (4-15p). A 10-percentage-point swing in self-consumption on a 250 kW system moves annual benefit by around 12,000 pounds. The only way to nail self-consumption is to overlay your real half-hourly load curve against a properly modelled solar generation profile. That is what this page is about — what HH data is, why it matters, how we extract it, what we do with it, and what the deliverable looks like.
What half-hourly data actually is
A half-hourly (HH) meter records electricity consumption in 30-minute increments — 48 readings per day, 17,520 readings per year. It is the standard metering approach for UK commercial sites consuming above 100,000 kWh per year and for all sites that switched to P272-class half-hourly settlement in 2017 under Ofgem reforms. Sub-100,000 kWh sites are typically on non-half-hourly (NHH) profile-class meters that record monthly or bi-monthly aggregates only.
The HH data flow goes meter, to data collector (DC), to data aggregator (DA), to supplier, to customer. The customer is the legal owner of the data and any UK supplier must release it on request. The standard format is a CSV file with 17,520 rows per year of consumption, each row containing a date, a half-hour timestamp, and a kWh value.
Why the load curve drives self-consumption
Solar generation has a fixed daily shape — zero at sunrise, building to a peak around solar noon, declining to zero at sunset. The peak hours are roughly 10:00-15:00 in summer and 11:00-14:00 in winter. The shape of the customer load curve relative to this generation profile determines self-consumption.
An office that operates 08:00-18:00 weekdays sees solar output during 100 percent of its operating hours. But weekends are export. Summer holidays are export. Lighting on overcast days still imports. Self-consumption typically lands at 55-65 percent for offices.
A 24/7 warehouse with chillers, lighting, and material-handling equipment sees solar output against a constant baseload. Self-consumption typically lands at 70-80 percent.
A school with August-September long summer holiday sees solar peak during the worst-case empty period. Self-consumption typically lands at 45-55 percent.
A care home with 24/7 occupied baseload, HVAC, and laundry sees the highest self-consumption — typically 85-95 percent.
Without HH data, an installer guesses these figures from sector typology. With HH data, the figures are calculated exactly. The accuracy gain is 15-25 percent of modelled IRR.
Sector load curves — what they actually look like
From our 2024-2026 install database of HH analyses, the typical sector load curve shapes:
Office (commercial 5-day week): Sharp ramp 06:30-09:00 from 25 percent baseload to 100 percent peak. Flat 09:00-17:30 around 95-100 percent. Sharp drop 17:30-19:00. Overnight baseload 20-30 percent. Saturday and Sunday at 25-35 percent baseload.
Warehouse and logistics (24/7 operation): Constant 60-80 percent baseload. Slight ramps for shift changes. MHE charging often nighttime which reduces solar self-consumption ratio. HVAC peaks summer afternoons aligned with solar peak.
Factory and manufacturing (single shift): Sharp ramp 06:00-08:00 from 20 percent to 100 percent. Flat 08:00-17:00. Sharp drop 17:00-19:00. Overnight baseload 15-25 percent for cooling and standby. Weekends typically 20 percent baseload.
Hotel: Two daily peaks — morning 06:00-10:00 (showers, breakfast service, laundry start) and evening 17:00-22:00 (dinner service, room occupation). Daytime mid-load 11:00-16:00 around 60 percent. Overnight baseload 35-45 percent. Weekend pattern varies — leisure hotels peak weekends, business hotels mid-week.
Care home: Constant 75-90 percent baseload all day. Minor morning peak for laundry and breakfast. Minor evening peak for dinner. Heating dominates winter, cooling and lighting dominate summer. Highest self-consumption sector we model.
School (term-time): Sharp ramp 07:00-08:30 from 30 to 90 percent. Flat 09:00-15:30. Sharp drop 15:30-18:00 to 30 percent. Overnight baseload 20-25 percent. Weekends and school holidays at 20-30 percent baseload — August summer holiday is the worst-case for self-consumption.
Restaurant: Lunch peak 11:30-14:30. Evening peak 18:00-22:00. Daytime mid-trade lower. Overnight baseload 15-25 percent for refrigeration and cooling. Self-consumption depends on lunch-versus-evening trading split.
Data centre: Near-constant 90-98 percent baseload. Slight diurnal HVAC variation summer. Highest self-consumption of any commercial sector.
What we do with the data
Once we have the HH CSV, the analysis runs in five stages.
Stage 1: Data normalisation. CSV format varies between suppliers. Octopus typically delivers in a clean two-column format. E.ON delivers an Excel workbook with sheets per month. SSE delivers a row-and-column matrix with dates as rows and half-hours as columns. Our Python tooling normalises all variants to a single 17,520-row format and runs sanity checks (no missing intervals, no kWh values that exceed plausible peak demand, no negative values).
Stage 2: PVSyst hourly simulation. The proposed solar array is modelled in PVSyst commercial 8.x for the customer postcode using TMY irradiance data, the panel-and-inverter specification, the orientation and tilt, and the array shading model. PVSyst returns 8,760 hourly generation figures for a typical meteorological year.
Stage 3: Hourly matching. The 17,520 half-hour load values are summed pairwise to 8,760 hourly load values. Hour by hour, generation and load are compared. Where generation is less than or equal to load, all generation is self-consumed. Where generation exceeds load, the surplus is exported. The hour-by-hour matched and exported totals are summed to annual figures.
Stage 4: DUoS-band overlay. The DNO-published DUoS Red, Amber, and Green band tables are applied to the HH consumption profile. Red Band weekday 16:00-19:00 carries the highest unit charge (sometimes 12-25p per kWh on top of the energy unit cost in some DNO areas). Solar generation that fires through the Red Band offsets exactly these expensive units, generating outsized financial benefit relative to a flat-rate average.
Stage 5: Financial modelling. Self-consumed kWh times retail tariff plus exported kWh times SEG tariff yields annual benefit. Capital cost yields simple payback. Annual benefit modelled forward 25 years with degradation (0.45 percent per year) and inflation (3 percent per year baseline) yields IRR and NPV.
What we deliver
The deliverable is a 12-15 page PDF deck containing:
- Cover summary — system size, modelled annual generation, modelled annual self-consumption percentage, modelled annual financial benefit, modelled simple payback, modelled IRR, modelled NPV at 5 percent and 8 percent discount rates
- Customer load profile analysis — annual consumption, peak demand, average demand, shape factor, daily mean, daily standard deviation
- Summer versus winter consumption split (typically 45 percent summer, 55 percent winter for non-cooling-led businesses)
- Weekday versus weekend delta
- DUoS Red Band exposure — kWh consumed in Red Band per year, financial value
- Hour-by-hour solar generation versus load — visualised as a heat map for a typical year
- Self-consumption breakdown by month and by day-type (weekday vs weekend, term-time vs holiday for schools)
- Export profile — month-by-month export volumes
- Optimal battery sizing analysis (if battery storage is being considered) — kWh battery capacity at which incremental cost equals incremental benefit
- Sensitivity analysis — how IRR moves with electricity-price inflation, SEG rate changes, system cost, and self-consumption variance
- 25-year cash flow with degradation, inflation, and discount rates
The financial impact in real numbers
Across our 2024-2026 install database we have HH-analysed and quote-vs-delivered numbers on roughly 80 commercial systems. Typical findings:
- Sector-typical self-consumption estimates land within plus or minus 10 percentage points of HH-modelled actual on around 70 percent of cases
- The remaining 30 percent of cases see HH-modelled self-consumption diverge from sector-typical by more than 10 percentage points — and almost every one of those would have led to a materially mispriced quote
- Average HH-corrected modelled annual benefit on a 250 kW system: 47k pounds versus 39k pounds on a sector-typical estimate — 21 percent uplift
- Average IRR uplift: 2.8 percentage points versus sector-typical estimate
- Investment-committee-grade rigour: most plc-grade procurement processes will not approve a solar capex paper without HH analysis
Tools we use
PVSyst commercial 8.x for the hourly PV simulation. Industry standard, used by every utility-scale solar developer.
Helioscope for the panel layout, shading model, and string design. Best-in-class for fast layout iteration.
Custom Python tooling for HH data normalisation, hourly matching, and DUoS-band overlay. Maintained internally.
UKPN, NPg, ENWL, SSEN, WPD, SP Energy Networks DUoS band tables sourced from each DNO published charging schedule.
What HH data tells us about battery storage
Battery storage adds capacity to shift exported solar into evening self-consumption. The battery economics work hardest where the export-vs-retail spread is large — high retail tariff and low SEG rate. HH analysis is the only way to size the battery correctly. Typical findings: a battery sized at around 1 kWh per 2 kW of solar PV (so a 100 kW system pairs with a 50 kWh battery) captures the lion share of evening self-consumption uplift. Larger batteries hit diminishing returns. See battery storage for the wider battery proposition.
Authority and references
Ofgem half-hourly settlement reform at Ofgem MHHS. ENA standards on commercial connections at Energy Networks Association. UKPN DUoS charging schedule at UKPN charges.
Related decision pages
For the broader site survey see commercial solar survey. For the install timeline see commercial solar installation timeline. For payback methodology see solar panel payback calculator. For the broader business case see are commercial solar panels worth it. For battery sizing see battery storage. For finance and tax see commercial solar finance and capital allowances. For the underlying quote process see get a quote.
Common questions
How do I get my half-hourly meter data from my supplier?
Email your supplier customer service team and request "12 months of half-hourly consumption data in CSV format" for your MPAN. All UK suppliers are obliged to release HH data on request to the metered customer free of charge — it is your data. Octopus, EDF, E.ON, British Gas, SSE, Drax, Total Gas and Power, npower, ScottishPower, and Pozitive Energy all have written processes for this. Typical turnaround is 5-15 working days. The format is normally 17,520 rows for a year (365 days x 48 half-hour periods) with a kWh value per row. Some suppliers email Excel files; some grant a self-service portal download.
What if I am not on half-hourly metering?
If your annual consumption is below 100,000 kWh per year you may be on a non-half-hourly (NHH) profile-class meter rather than HH. In that case the supplier issues monthly or bi-monthly consumption figures rather than 30-minute granularity. You can either (a) request a free meter upgrade to a Smart Meter (SMETS2) which captures HH-equivalent data and have your supplier pull a few months of data, (b) install a third-party CT-clamp logger such as a Carbon Track, GridDuck, or Stark Carbon for 4 weeks, or (c) accept that we use a sector-typical load curve in our analysis and add a 10-15 percent precision uncertainty to the modelled IRR. For most owner-managed businesses below 100 MWh/year, sector-typical curves are accurate enough.
How much more accurate is HH analysis versus an estimate?
Across our 2023-2026 install database, full HH analysis improves modelled IRR accuracy by 15-25 percent versus a sector-typical estimate. The biggest source of estimate error is self-consumption percentage — a "warehouse 75 percent" assumption lands the customer real number anywhere from 62 percent to 89 percent depending on operational pattern, MHE charging, ambient lighting, and HVAC. HH analysis nails the actual figure. The financial impact: on a 250 kW warehouse install, HH-corrected modelling typically moves modelled annual benefit by 8-18k pounds versus a sector-typical estimate. That is meaningful for any investment committee.
Why do most solar installers not do half-hourly meter data analysis?
Two reasons. First, it requires analytical capability that most installers do not have — PVSyst commercial licences cost around 2,500 pounds per year per seat, Helioscope around 4,500 USD per year, and the analyst time per project runs 4-12 hours. Most national solar installers cost-cut these out. Second, it slows the sales cycle — a sales-driven installer wants the quote out within 24-72 hours of an enquiry, while a proper HH analysis needs a fortnight including the supplier data extraction. Cowboy installers default to estimates. Engineering-driven installers do the analysis. Donovan Fawcett and our analyst team run the model on every quote above 50 kW.
What software do you use for half-hourly modelling?
We use PVSyst commercial 8.x for the hourly PV simulation (irradiance, panel temperature, inverter clipping, soiling, system losses), paired with our own Python tooling that ingests the customer HH meter CSV, normalises the format, builds a 17,520-row consumption profile, and overlays the PVSyst hourly output. The result is an hour-by-hour matched-and-unmatched calculation across a typical year. We also use Helioscope for the panel-layout and shading analysis. For DUoS-band overlay we use the published UKPN, NPg, ENWL, SSEN, WPD, and SP Energy Networks band tables and apply them to the HH profile.
What deliverables do I get from the analysis?
A 12-15 page IRR-and-NPV deck containing: shape factor analysis (kurtosis of the load curve, standard deviation across days), summer-versus-winter consumption split, weekday-versus-weekend delta, peak-period (DUoS Red and Amber band) exposure analysis, optimal export-versus-self-consumption ratio, hour-by-hour matching of solar output to load, modelled annual energy saving, modelled annual export income, modelled simple payback, modelled discounted payback at user-specified rate, modelled IRR over 25 years, modelled NPV at 5 percent and 8 percent discount rates, and a sensitivity analysis showing how IRR moves with electricity-price inflation and SEG rate changes. The deck takes 8-12 working days from receipt of HH data.