A commercial solar system should be sized by overlaying modelled hourly generation against the business's own interval consumption data, then testing several capacities to see which produces the best result once self-consumed, exported and curtailed energy have been separated and valued differently.
Annual electricity consumption, roof area and monthly bills are all inputs to that process. None of them is sufficient on its own, and each of them is regularly used as a shortcut that produces the wrong answer.
This article sets out the workflow, the data sources, the physical and electrical constraints that limit what is possible, and a worked comparison showing why the largest system a roof can hold is not automatically the best investment.
Why annual consumption is not enough
A business consuming 600,000 kWh a year cannot size a system from that figure, because the number contains no information about timing.
Solar generates during daylight. Its value depends entirely on whether the business is consuming at the same moment. Two sites with identical annual totals can have completely different overlaps: a bakery running from three in the morning, a hotel with an evening peak, a logistics depot working shifts, a manufacturer on a five-day pattern with a summer shutdown. The annual figure is the same. The proportion of solar output they can use is not.
The common shortcut is to divide annual consumption by an assumed annual yield per kWp. For a site consuming 600,000 kWh at an assumed 1,500 kWh per kWp, this gives 400 kWp. That calculation does something useful but limited: it estimates the capacity that would generate roughly the same quantity of energy over a year as the business consumes. It does not establish the economically correct system size, because it says nothing about whether generation and demand coincide, and therefore nothing about how much of that output would be self-consumed rather than exported or curtailed.
Sizing is a timing problem, not an arithmetic one.
Five capacity figures that get confused
Proposals move between these without always distinguishing them, and they are not interchangeable.
| Measure | Unit | What it describes |
|---|---|---|
| Installed capacity | kWp | Peak direct-current rating of the panel array under standard test conditions |
| Annual generation | kWh | Electricity the system is modelled to produce over a year |
| Inverter capacity | kW | Maximum alternating-current output the inverters can deliver |
| Maximum demand | kW | Highest recorded power drawn by the site, measured in intervals |
| Contracted capacity | kW | Capacity contracted with the supplier in each tariff period, billed regardless of use |
Installed capacity and inverter capacity are frequently different by design, and the relationship between them is a design decision rather than a fixed rule. Maximum demand and contracted capacity are properties of the site's electricity use, not of the solar installation, and neither is automatically changed by installing panels.
When comparing proposals, confirm which figure each supplier is quoting. A "300 kW system" may mean 300 kWp of panels, 300 kW of inverters, or something in between.
Self-consumption rate and solar coverage rate
These two ratios are the core outputs of a sizing exercise, and they are routinely mixed up.
Self-consumption rate = solar generation used on site ÷ total solar generation
This asks how much of what the system produces the business actually uses. It falls as the system grows, because additional capacity produces surplus at times when demand is already met.
Solar coverage rate = solar generation used on site ÷ total site consumption
This asks how much of the business's electricity demand solar meets. It rises as the system grows, but with diminishing returns.
The two move in opposite directions, which is why a single headline percentage is meaningless without knowing which one is being quoted. A proposal claiming "80%" could be describing a small system using nearly all its output, or a large system meeting most of the site's demand while wasting a great deal.
Both should be stated. Neither should be optimised in isolation.
The sizing workflow
Obtain interval consumption data
Twelve months minimum, covering a full seasonal cycle including the site's quietest and busiest periods. Shorter datasets miss the shutdown, the summer peak or the winter heating load that determines whether a system works.
Interrogate the data before using it
Raw consumption data contains events that will distort a model if carried through unexamined. Look for shutdown periods, gaps where the meter did not report, months affected by unusual operations, and any equipment that has since been removed or added. Decide explicitly whether each is representative of the future or should be adjusted.
A three-week August shutdown is not an anomaly to be smoothed away. It is a real feature that will produce surplus every year, and the model should reflect it.
Account for the operating pattern
Separate weekdays from weekends and holidays. Identify shift patterns, seasonal variation and any load that runs continuously, such as refrigeration or process cooling. Continuous base load is the most valuable demand a solar system can meet, because it is present whenever the sun is.
Include planned new loads, with evidence
Fleet charging, heat pumps, additional machinery and expansion all raise daytime demand and improve the case for a larger system. They belong in the model when they rest on committed plans with dates and specifications. They do not belong in it when they exist to justify capacity the business does not currently need.
Where future load is genuinely uncertain, model both cases and see how much the answer moves.
Model hourly generation for the actual site
Generation must be modelled for the specific coordinates, roof orientation, tilt and shading, not taken from a regional average. The generation and consumption series must then be aligned to the same timestamps and interval length before they are compared.
Test several sizes and overlay
Run the overlay at several capacities rather than one. Each hour, compare generation against demand and record which part was consumed on site, which part was surplus and, under a zero-export configuration, which part was curtailed.
Separate the streams and value each correctly
Self-consumed electricity is valued at the cost avoided in that tariff period. Exported surplus is valued under the applicable compensation or sale arrangement. Curtailed energy is worth nothing. Applying a single price to total generation will overstate every option, and will overstate the largest option most.
Getting consumption data in Spain
Datadis is the platform operated jointly by the Spanish distribution companies, giving supply point holders access to their own consumption data using the CUPS reference from their electricity bill.
According to the platform's own documentation, the data supplied by the distributors covers the supply point's NIF, CUPS, distributor, connection voltage, access tariff, province, municipality, postcode, contracted capacity by period, contract dates, current supplier, maximum demand and load curves.
Interval resolution is not uniform. Depending on the meter installed, the nature of the supply and the records held by the distributor, the consumption curve available may be hourly or quarter-hourly, and the depth of history retrievable also varies. A business should check what is actually available for its own CUPS rather than assume a particular dataset will be there. Where the resolution turns out to be hourly, that is generally sufficient for sizing work, though quarter-hourly data gives a better view of short demand peaks.
Access can be granted to a third party. The holder authorises the adviser or installer through the platform, which is generally cleaner than exporting files by email and avoids the version confusion that follows when several parties hold different extracts.
The distributor's own portal is the alternative route. i-DE, e-distribución, UFD, Viesgo and the other distributors each provide a portal for the CUPS holder, normally requiring a digital certificate or electronic ID.
Where data cannot be obtained at all, temporary metering can be installed to record a representative period. That is slower and less complete than twelve months of history, and it is better than sizing from a bill.
Modelling generation with PVGIS
PVGIS, the Photovoltaic Geographical Information System, is maintained by the European Commission's Joint Research Centre and provides free solar radiation and photovoltaic performance data for almost any location worldwide. It requires no registration and offers an API for automated access.
Two versions are relevant. PVGIS 5.3 is the established, stable service, and the Spanish-language interface and the typical meteorological year output are currently identified by the JRC as PVGIS 5 features. PVGIS 6 is a beta release, introducing a new interface, a new API web service and a rebuilt Python backend, with a dataset covering 2013 to 2024 that the JRC describes as essentially the same underlying data as PVGIS 5.3 in an updated format and handling process.
For commercial work the practical implication is procedural. Use whichever version suits the task, but record which one produced the figures, together with the radiation database selected, the date the model was run and every input applied. A generation estimate that cannot be reproduced is difficult to defend when a system underperforms, and a beta service may change while a project is in progress.
For sizing purposes the relevant output is the hourly time series, which can be exported and aligned directly with the consumption curve. The inputs to set are the coordinates, the installed capacity, the mounting configuration, the azimuth, the tilt angle and the system loss assumption.
What PVGIS does not do matters as much as what it does. It models expected generation from meteorological data. It does not assess whether the roof can carry the load, whether nearby structures cast shadows that the horizon data does not capture, whether the electrical infrastructure can accept the output, or how the array should actually be laid out. A PVGIS figure is an input to a design. It is not a design, and it does not replace a structural survey, a site-specific shading study or detailed engineering.
Orientation, tilt, shading, temperature and losses
Orientation and tilt determine both the annual total and the daily shape of generation. In Spain a south-facing array at a tilt broadly in the region of the site's latitude tends to maximise annual output, but on a commercial building the array usually follows the existing roof pitch, and the shape of the production curve matters more than the annual peak.
Shading is the constraint most often underestimated. Roof plant, adjacent buildings, parapets, chimneys, trees and neighbouring structures all cast shadows that vary through the day and the year. Modelled horizon data captures distant terrain, not the air handling unit ten metres away. Near shading requires site assessment, and its effects can be disproportionate where module strings are poorly arranged.
Temperature reduces output. Panels are rated at 25 degrees, and cell temperatures on a Spanish industrial roof in July run well above that. Losses from temperature are real, predictable and should be visible in the model rather than absorbed silently.
System losses cover cabling, inverter conversion, soiling, mismatch and availability. PVGIS applies a default that can be adjusted, and the figure used should be stated in any proposal rather than left at whatever the tool offered.
Why east-west can sometimes beat south
Many Spanish industrial buildings have a duo-pitched roof running east to west, giving two slopes facing east and west rather than one facing south.
A south-facing array of the same capacity will usually generate more electricity over a year. An east-west array produces a flatter, wider curve: earlier generation in the morning, later generation in the afternoon, a lower midday peak.
For a business whose demand starts at six or seven in the morning and continues into the evening, that flatter curve can align better with actual consumption. More of the output arrives when the business is using it, and less arrives as a midday surplus that has to be exported or curtailed. The self-consumption rate rises even though total generation falls.
East-west layouts also allow closer module spacing, since the panels do not shade each other in the way tilted south-facing rows on a flat roof do, so more capacity often fits in the same area.
Whether this outweighs the lower annual yield depends entirely on the consumption curve. It is a question the overlay answers and that a rule of thumb cannot.
Roof constraints: usable area, not total area
Total roof area and usable roof area are different numbers, sometimes by a wide margin.
Deductions to expect include perimeter and edge zones kept clear for safety, fire access routes and any separation required between arrays, rooflights and smoke vents which cannot be built over, existing plant such as air handling units, extractors, condensers and their maintenance access, walkways, and areas shaded by adjacent structures.
Beyond area, three physical questions govern feasibility.
Structural capacity. The roof has to carry the additional dead load, plus wind uplift and, on flat roofs, any ballast. This requires assessment by a qualified professional, not an assumption.
Roof condition and remaining life. Fixing a twenty-five year asset to a roof with eight years left in it creates a problem that will have to be paid for twice. Where the roof is nearing the end of its service life, the sensible sequence is to address the roof first.
Asbestos-containing materials. Older Spanish industrial roofs may contain fibre cement with asbestos. These cannot simply be drilled or fixed into, and the presence of such materials changes the mounting approach, the cost and the programme. This must be established during the survey rather than discovered during installation.
Electrical constraints
The electrical infrastructure often limits system size before the roof does.
Connection voltage determines the equipment and the procedures involved. A low-voltage connection and a medium-voltage connection are different projects, with different protection, metering and engineering requirements.
Transformer and switchboard capacity. The existing transformer, main switchboard and available ways all constrain how much generation can be connected and where. A site with a fully loaded board may need switchgear work before any capacity can be added.
Export limitation. Where export is not available or not wanted, an anti-export system holds inverter output below site demand. This changes the sizing calculation fundamentally, because surplus is not exported at low value but curtailed at zero value. Under curtailment, oversizing is a straightforward loss.
Distributor requirements. Installations without surplus are exempt from grid access and connection permits under the Spanish self-consumption framework established by Real Decreto 244/2019. Exporting installations above the exempt thresholds are not, and available capacity at the connection point can cap the system regardless of what the roof and the load would support.
Inverter sizing and the DC to AC ratio. Arrays are commonly specified with more panel capacity than inverter capacity, since the array rarely reaches its peak rating. The appropriate ratio depends on the site's irradiation profile, the orientation, whether the array is split across orientations, the inverter's own limits and whether export is constrained. An east-west array with its flatter curve supports a different ratio from a south-facing one. There is no single correct figure, and a supplier applying the same ratio to every project is applying a habit rather than a calculation. What matters is that the chosen ratio is stated, justified and reflected in the modelled output including any clipping.
Tariff periods and weekends
Spanish commercial tariffs divide the year into six periods, and generation should be valued according to the period it displaces rather than against an annual average price.
The weekend effect deserves specific attention during sizing. Saturdays, Sundays and national holidays fall entirely within P6, the cheapest period, in every season. For a business operating Monday to Friday, weekend generation arrives when on-site demand is at its lowest and when whatever is consumed displaces the least valuable electricity of the week.
A sizing model built on annual totals will not reveal this. One built on interval data will show it immediately, and it usually argues for a smaller system than the roof would accommodate.
What are you optimising for?
Different objectives produce different answers from the same data, and the objective should be agreed before capacity is proposed.
| Objective | Tends to produce | Trade-off |
|---|---|---|
| Maximum self-consumption rate | A smaller system with little or no surplus | Leaves available savings unrealised |
| Maximum annual savings | A larger system capturing more total benefit | Lower return per euro invested |
| Best NPV or IRR | A system at the point where incremental return still exceeds cost of capital | Neither the smallest nor the largest option |
| Fixed roof area | Whatever fits | Ignores the consumption profile entirely |
| Decarbonisation target | Sized to a coverage percentage | May accept a weaker financial return deliberately |
| Future electricity demand | Capacity ahead of current need | Carries surplus until the load materialises |
Most commercial decisions should be made on NPV or IRR. Where a business has a reporting commitment or a corporate target, a coverage-driven size may be justified, but it should be a stated choice rather than an unexamined default.
Worked example: three sizes for the same business
The following compares three capacities for one site. Every figure is illustrative and internally consistent, so a reader can substitute their own inputs and follow the same method.
The business. A food processing facility operating Monday to Friday, 06:00 to 18:00, with refrigeration running continuously. Annual consumption 600,000 kWh. Roof capable of accommodating up to 500 kWp. Export permitted and surplus sold under an applicable market or representation arrangement.
Shared assumptions. Modelled yield of 1,500 kWh per kWp, illustrative rather than modelled for a specific location. Self-consumed electricity valued at €0.155 per kWh, weighted across the tariff periods displaced. Exported surplus valued at €0.040 per kWh net. Maintenance at €12.50 per kWp per year. Costs excluding IVA, reflecting the usual decline in cost per kWp as capacity rises.
| Measure | 150 kWp | 300 kWp | 450 kWp |
|---|---|---|---|
| Installed cost, excluding IVA | €120,000 | €210,000 | €288,000 |
| Cost per kWp | €800 | €700 | €640 |
| Modelled annual generation | 225,000 kWh | 450,000 kWh | 675,000 kWh |
| Self-consumption rate | 90% | 66% | 50% |
| Electricity self-consumed | 202,500 kWh | 297,000 kWh | 337,500 kWh |
| Surplus exported | 22,500 kWh | 153,000 kWh | 337,500 kWh |
| Solar coverage rate | 33.8% | 49.5% | 56.3% |
| Value of self-consumed electricity | €31,388 | €46,035 | €52,313 |
| Value of exported surplus | €900 | €6,120 | €13,500 |
| Less annual maintenance | −€1,875 | −€3,750 | −€5,625 |
| Net annual benefit | €30,413 | €48,405 | €60,188 |
| Simple payback | 3.9 years | 4.3 years | 4.8 years |
Reading the comparison
The 450 kWp system delivers the largest annual benefit in absolute terms, nearly double that of the 150 kWp system. It also has the longest payback and the lowest return per euro committed.
The reason is visible in the self-consumption rate, which falls from 90% to 50% as capacity triples. The first 150 kWp lands almost entirely on demand that exists. The last 150 kWp lands substantially on hours when demand is already met, so half its output leaves the site at roughly a quarter of the value.
The incremental view makes this explicit.
| Step | Additional cost | Additional annual benefit | Incremental payback |
|---|---|---|---|
| 150 kWp to 300 kWp | €90,000 | €17,993 | 5.0 years |
| 300 kWp to 450 kWp | €78,000 | €11,783 | 6.6 years |
Each increment is less attractive than the one before. Whether the second step is worth taking depends on the business's cost of capital and on what else it could do with €78,000. A business with cheap capital and no competing uses may reasonably take it. A business with constrained capital and a better internal use for the money should probably not.
The roof capacity sets the ceiling rather than selecting the optimum. This roof can hold 500 kWp, but that fact does not establish which of the three options produces the best use of the business's capital.
Common sizing mistakes
Dividing annual consumption by an assumed yield. Produces a number quickly and ignores the entire question of timing.
Sizing from the electricity bill alone. Bills show totals and costs. They do not show the load curve, and the load curve is the thing being matched.
Filling every available square metre. Roof area is a constraint on the maximum, not an input to the optimum. A proposal whose capacity happens to equal the roof's capacity was sized by the roof.
Ignoring weekends and shutdowns. Both produce systematic surplus at predictable times, and both are invisible in annual figures.
Treating exports like self-consumption. Applying one price to total generation flatters every option and flatters the largest option most, which is precisely the direction of the resulting error.
Assuming future loads without evidence. Planned electrification genuinely improves the case for capacity. Hypothetical electrification is a way of justifying a bigger system.
Adding batteries before sizing the solar case. Storage changes the surplus position, so a battery introduced early obscures whether the solar element works on its own.
Batteries come after, not during
Storage should be evaluated as an incremental decision once the solar system has been sized.
The reason is a matter of sequence. A battery changes what happens to surplus, so introducing it during sizing means the array is being sized against a load profile that includes an asset not yet justified. The array grows to feed the battery, and the battery is justified by the surplus the enlarged array creates.
The correct order is to establish the best solar capacity on its own terms, then examine the residual surplus and the residual evening demand, and ask what storage would be worth against that specific position. That question has its own answer, its own cost and its own return.
Frequently asked questions
How do you size a commercial solar system? By overlaying modelled hourly generation against at least twelve months of the site's interval consumption data, testing several capacities, and separating self-consumed, exported and curtailed energy so that each can be valued correctly.
How many solar panels does a business need? There is no formula based on annual consumption or roof area. The answer comes from the overlap between the site's consumption curve and modelled generation, and two businesses with the same annual consumption can need very different systems.
What is the difference between self-consumption rate and solar coverage rate? Self-consumption rate is solar used on site divided by total solar generated, and it falls as a system grows. Solar coverage rate is solar used on site divided by total site consumption, and it rises as a system grows. They answer different questions and should both be quoted.
Where can a Spanish business get its hourly consumption data? Through Datadis, using the CUPS reference from an electricity bill, or through the distributor's own portal. The curve available may be hourly or quarter-hourly depending on the meter, the supply and the distributor's records, and the depth of history varies by supply point.
Is PVGIS accurate enough for a commercial project? PVGIS is a sound basis for modelling expected generation from meteorological data and provides hourly time series suitable for overlay work. It does not assess structural capacity, near shading from roof plant or adjacent buildings, or electrical constraints, so it supports a design rather than replacing one. Record which version and radiation database produced the figures.
Is an east-west roof worse than a south-facing one? Not necessarily. A south-facing array usually generates more annually, but an east-west layout produces a flatter curve that can match a business's operating hours better, raising the proportion of output consumed on site. Which is better depends on the load profile.
What DC to AC ratio should be used? There is no universal figure. It depends on the site's irradiation profile, orientation, whether the array is split across orientations, inverter limits and whether export is constrained. What matters is that the ratio is stated and that any clipping appears in the modelled output.
Should the system be sized to cover all the business's electricity? Rarely. Covering total consumption would require capacity far beyond what daytime demand can absorb, producing large surpluses at low value. Sizing is normally optimised on return rather than coverage, unless a decarbonisation target justifies a different objective.
Reviewed 19 August 2026. All figures in the worked example are illustrative and internally consistent, not market averages or modelled outputs for any specific location.
Sources used
Primary and official sources:
- PVGIS, European Commission Joint Research Centre, for the tool's scope, hourly time series and typical meteorological year outputs, identified as PVGIS 5 features
- About PVGIS 6, for the beta release, the new API web service and Python backend, and the 2013 to 2024 dataset
- Datadis FAQs, for the fields supplied by the distributors, including contracted capacity by period, maximum demand and load curves, and for third-party authorisation
- Real Decreto 244/2019, for the self-consumption framework, the exemption of installations without surplus from access and connection permits, and the 100 kW limit on simplified compensation
- CNMC Circular 3/2020, for the six-period tariff structure
One item to confirm before publication. Datadis history depth is described as variable rather than fixed, which is the safe formulation; if you want to state a specific maximum, verify it against the platform directly, since integration partners report different limits from the platform itself.