Industrial Solar Self-Consumption: Sizing PV and Storage from the Load Curve

Sizing an industrial self-consumption system well does not start on the roof or in the panel catalogue: it starts with the plant's load curve. This guide explains how to use the real hourly profile to decide the photovoltaic capacity, how to balance coverage against export, when batteries genuinely add value and when they are an expensive indulgence, with a complete worked example for a single day-shift plant.

The load curve rules, not the roof

The question that kicks off most industrial self-consumption projects is the wrong one: how many square metres of roof do I have and how many panels will fit. The roof is a constraint, not a criterion. The criterion is how much energy the plant consumes, at what hours it consumes it and with what regularity. Two identical 8,000 m² buildings can justify very different installations if one runs a morning shift with process machinery and the other runs three shifts with furnaces that never stop.

The technical starting point is the hourly, or better still quarter-hourly, load profile covering at least twelve months. In Spain, any supply point with a smart meter has that data available through the distribution company, and plants with their own monitoring can go into far more detail. With that time series you can answer the three questions that govern sizing: what is the base load that never drops (the night-time and weekend trough), what is the typical consumption in the middle hours of the day, and how does all of it vary between seasons and between working days and holidays.

Without that profile, any capacity figure is a gamble. With it, sizing becomes an exercise in superposition: laying the solar production curve over the demand curve and measuring how much energy coincides. Everything else (structure, inverters, permits) comes afterwards.

The mistake of sizing by fashion

In recent years a recognisable pattern has spread: management decides to install self-consumption because a competitor has done it, quotations are requested from three installers, and the one offering the most kilowatt-peak per euro wins. The usual result is an installation sized by the available surface, which in plants with large roofs and moderate consumption produces far more than the factory can absorb in real time.

The problem is not cosmetic. Every kilowatt-hour the plant does not consume at the moment it is generated is exported to the grid, and that export is remunerated well below the price at which the plant buys energy. An oversized installation has a somewhat lower cost per kilowatt-peak (economies of scale in the works), but a much worse value per kilowatt-hour, because a growing fraction of its production is paid at surplus prices. The tell-tale indicator is the self-consumption rate: if the preliminary study projects that only 50 or 60 per cent of the generation will be consumed on site, the installation has been designed for the roof, not for the plant.

The practical rule is the opposite of the commercial intuition: in industry it is usually better to fall slightly short than to overshoot. An installation matched to the minimum daytime consumption guarantees that almost every kilowatt-hour generated replaces one that would have been bought, which is where the real saving lies. Expanding later, if consumption grows or a process is electrified, is almost always possible; trimming an oversized installation is not.

Coverage and export: the two faces of sizing

Two ratios summarise the behaviour of any self-consumption system, and it is worth having them clear before talking numbers:

The two ratios pull in opposite directions. A small installation achieves self-consumption rates close to 100 per cent (everything it produces is consumed) but covers little of the demand. A large installation covers more demand but exports ever more. Optimal sizing maximises neither: it looks for the point at which the last kilowatt-peak added still generates more value than it costs. That point depends on the price of purchased energy, on the remuneration of the surplus and on the shape of the load curve, which is why there is no universal ratio of kWp per square metre or per contracted kilowatt that works for every plant.

One practical consequence: the marginal economic return declines. The first kilowatt-peaks are the most profitable, because their production coincides almost entirely with real consumption. The last are the least, because most of their production ends up exported. Any serious study should show this marginal value curve, not just the aggregate of the proposed installation.

Solar profile, plant profile and their overlap

Photovoltaic production has a familiar shape: a bell curve centred on solar noon, taller and wider in summer, lower and narrower in winter, with cloudy days clipping it irregularly. The plant's profile is what it is: it depends on shifts, processes and the calendar. Sizing consists of measuring the overlap between the two curves, hour by hour, over a full year.

Three industrial archetypes illustrate how different that overlap can be:

Special attention must be paid to weekends and August shutdowns. A plant that closes on Saturdays and Sundays leaves around 100 hours a week of potential generation without consumption, many of them in broad daylight. Over the annual total, those days weigh heavily: an installation that self-consumes 95 per cent on weekdays can end up at 75 or 80 per cent for the year purely through the calendar effect. The study must be run over the complete year, with the plant's real working calendar, not over a typical week.

Sizing method, step by step

This is the procedure we follow in a sizing study:

1. Obtain the hourly or quarter-hourly series of consumption for the last twelve months, and clean it: identify extraordinary stoppages, changes in production regime and loads that will disappear or appear (a new compressor, a line being relocated).
2. Build the generation profile per kilowatt-peak for the real location, orientation and tilt of the roof, using irradiation data from reputable public sources (for example the PVGIS tool of the European Commission's Joint Research Centre).
3. Simulate the superposition for a range of capacities (for example from 100 to 1,000 kWp in 50 kWp steps), calculating for each size the energy self-consumed, exported and residually purchased, hour by hour.
4. Value each scenario with the real price of purchased energy by tariff period and a prudent assumption for surplus remuneration, and build the marginal value curve per kilowatt-peak added.
5. Choose the size where the marginal value falls below the marginal cost, and check it against the physical constraints (usable area, structure, capacity at the connection point) and administrative ones.
6. Repeat the exercise adding batteries only if the chosen scenario leaves significant export or there are capacity charges to shave, as explained in the next section.

Note the order: batteries are assessed last, on top of an already optimised photovoltaic size, not as part of the initial package. Mixing the two decisions is the quickest route to justifying a battery that in reality only compensates for oversized panels.

Storage: when it makes sense and when it does not

An industrial battery does, in essence, three jobs: it shifts surplus solar energy into hours without sun, it shaves demand peaks that inflate the capacity term of the bill, and it provides back-up against micro-outages in sensitive processes. Its economic fit depends on how many of those jobs it can do simultaneously in a given plant.

Storage tends to make sense when several of these conditions hold:

• The plant has significant consumption outside the solar window (afternoon or night shift) and the photovoltaic installation, even well sized, exports systematically in the middle hours.
• The spread between the price of energy bought in expensive hours and the value of the surplus is wide and stable, so that each charge and discharge cycle captures a real margin.
• There are short, predictable demand peaks (start-ups, furnaces, compressors) whose shaving allows contracted capacity to be reduced or penalties avoided.
• There are processes where a micro-outage is expensive (extrusion, injection moulding, lines with product in progress), so the back-up value adds to the energy value.

And it tends not to make sense when:

• The photovoltaic installation is well matched and annual export is small: there is no cheap energy to shift, and the battery would only be moving purchased energy around.
• The plant runs a single day shift and stops in the afternoon: the shifted energy would have no consumption to serve until the next morning, when the sun is back anyway.
• The business case depends on a single revenue stream. The batteries that pay off in industry almost always stack two or three uses; with only one, the payback stretches beyond the expected life of the cells.

It is worth saying plainly: in a day-shift plant with well-sized photovoltaics, the battery is usually the worst investment of the three (panels, demand management, battery). That can change if cell prices keep falling or if the plant's profile changes, which is why the reasonable recommendation is to design the installation ready to incorporate storage (space, a hybrid inverter or the electrical room for it) without buying it on day one.

Surplus compensation in industry

The Spanish self-consumption framework allows, under certain schemes, exported surpluses to be offset against the energy term of the bill. Without going into the regulatory detail, which changes and must be verified case by case with the retailer and the rules in force, there are three conceptual ideas an industrial manager should hold on to.

First: simplified compensation is designed for small installations and has capacity limits that exclude a good share of industrial systems. Above those limits, the surplus is sold as energy on the market, with the associated obligations, or through a market representative. The applicable regime shapes the value of the exported kilowatt-hour and, therefore, the optimal size.

Second: even where compensation applies, the value of the compensated surplus is lower than that of self-consumed energy, because it only offsets the energy term and with monthly caps. Compensation cushions the cost of export; it does not turn it into a business. Sizing on the assumption that the surplus "comes back on the bill" is repeating the fashion mistake with a different argument.

Third: the design variable remains the same. The better the generation fits the consumption, the less the treatment of the surplus matters. Compensation is a cushion for the unavoidable mismatches (weekends, August, low-production days), not a pillar of the business case.

Worked example: a day-shift plant

Take a machining plant with a single shift from 7:00 to 15:30, Monday to Friday. Its load profile, built from a year of quarter-hourly data, summarises as follows: a night-time and weekend base load of 60 kW (metrology room air conditioning, compressed air holding pressure, services), consumption during productive hours oscillating between 320 and 420 kW with an average of 380 kW, and a total annual consumption of approximately 1.45 GWh. The roof can physically hold up to 900 kWp.

If sized by roof, the 900 kWp installation would produce around 700 to 800 kW in the middle hours of a clear June day, twice what the plant consumes in its best hour. The hourly simulation of that option yields a self-consumption rate of around 55 per cent: almost half of the annual production would be exported, much of it on summer weekends with the plant shut.

The simulation across a range of capacities tells a different story. Around 400 kWp, peak-hour production sits just above the average productive consumption, annual export falls to figures in the region of 15 to 20 per cent (concentrated in weekends and holidays, which no capacity choice avoids) and the coverage rate of annual consumption lands near 30 per cent. The marginal value curve shows that beyond that point each additional 50 kWp block delivers less and less real saving, because a growing share of its production ends up exported.

And the battery? In this plant, the shift ends at 15:30: the afternoon sun generates a surplus with no consumption to serve until the next morning, when the sun is available again. Shifting that energy with batteries creates no value, because there is no significant night-time demand to cover (the 60 kW base load barely justifies cycling). The only use with any fit would be shaving the start-up peaks first thing in the morning, and that is solved earlier and more cheaply with demand management: sequencing start-ups and scheduling movable loads within the solar window. The study's conclusion: 400 kWp, no battery, with the electrical room prepared to add one if the plant moves to two shifts.

The specific numbers will vary from plant to plant, but the mechanics are always the same: the answer was in the load curve from the start, and the role of the study is to make it visible before anything is signed.

Indicators for taking the decision

A sizing study fit to present to management should include, as a minimum, this set of indicators for each capacity scenario:

• Annual self-consumption and coverage rates, and their monthly breakdown.
• Annual exported energy and its distribution (weekdays versus weekends).
• Estimated annual saving with real prices by tariff period, separating the value of self-consumed energy from the value of the surplus.
• Marginal value curve per block of capacity added.
• Sensitivity to the two assumptions that move the result most: the future price of energy and the evolution of the plant's consumption.

And once the installation is running, the work does not end: the real self-consumption rate only holds if someone measures it and acts on it, shifting loads into the solar window and detecting deviations. That operating layer is what connects self-consumption with the plant's connected energy management, and it is usually worth several points of coverage a year without adding a single panel.

Frequently asked questions on industrial self-consumption

At Captia Energy we approach industrial self-consumption in this order: first the curve, then the capacity, storage last. If you are considering an installation for your plant, our industrial solar solution always starts from the hourly simulation on your real data, and the connected energy management layer makes sure the projected self-consumption rate still holds in year two. Explore the full approach at Captia Energy.