Solar Energy for Manufacturing Plants and Factories in Colombia: How to Size It Against the Load Profile

A manufacturing plant usually meets the three conditions that make a commercial solar project strong: it consumes while the sun is up, it consumes almost every day and it has large, unobstructed roofs. Why that shortens the payback compared with a home is covered in the article on why a business recovers the solar investment sooner; here the question is how it plays out in a specific factory.

The first clarification is about scope. A photovoltaic plant produces electricity and only replaces electricity, which in Colombian industry is the smaller part of the energy used. According to UPME's National Energy Plan 2024-2054, with data from the Colombian Energy Balance, the energy sources most used by industry are natural gas (29%), mineral coal (24%), bagasse (19%) and electricity (15%). The same document, citing the first Useful Energy Balance (2018 data), splits industry's end uses as follows: indirect heat, such as steam, 44%; direct heat, such as furnaces, 43%; motive power, including refrigeration, 11%; and other uses such as lighting and IT, 2%.

That has two consequences for the project. Gas- or coal-fired furnaces and boilers do not consume less because of panels: the solar plant acts on the electricity bill, not the gas bill. And any percentage a proposal promises must be expressed over electrical kWh, never over the factory's total energy.

In a factory, monthly consumption says little. What decides how much of a solar plant's output is used is the shape of the hourly curve: when each shift starts and ends, how much stays on outside shifts and what happens on Saturdays, Sundays, public holidays and scheduled shutdowns.

Each process leaves a different mark on that curve. Production motors (conveyors, mills, extruders, pumps, fans) follow the shift; in a grid-connected plant their starts and peaks are supplied by the grid, and the solar plant does not need to cover them. Compressed air often stays partly on outside shifts, because of leaks and equipment that remains pressurised, and it is one of the base loads most worth measuring. Furnaces and process heat count for solar only if they are electric (induction or resistance), in which case they are often among the plant's largest loads. Process cooling, cold rooms and chillers run day and night: solar feeds them during the day and at night they stay on the grid.

The practical way to see this is to ask the retailer or the network operator for twelve months of hourly load data and, if possible, to measure the large processes circuit by circuit for a few weeks. That separates three quantities the sizing needs: the shift load, the off-shift base load and the non-production days. If the company is planning a new shift, a new line or the electrification of a thermal process, the profile changes: the solar plant is sized for the load of the coming years, not only for that of the last twelve months.

To give orders of magnitude, we modelled a standard plant with four schedules and PVGIS hourly production for Bogotá, Medellín, Cali and Barranquilla, using 19 years of data (2005-2023). It is our own calculation, not a measurement of real plants; the assumptions are listed in the next section.

The solar plant was sized at 1.25 kWp for each kW of shift load, the load-cap criterion from the article on roof area: since a solar plant rarely delivers more than 80% of its kWp in an hour (in PVGIS hourly data the maximum ranges from 0.81 to 0.87 kW per kWp depending on the city), at that size it barely exports during shift hours. Each range goes from the city with the least sun (Cali) to the one with the most (Barranquilla).

Three readings. The percentage falls as shifts are added not because the solar plant produces less, but because consumption grows in hours without sun: in kWh, the three-shift plant uses more solar energy than the one-shift plant. A 6:00 to 14:00 shift coincides with 77% to 79% of the day's solar production; the afternoon, when that schedule only serves the base load, and Sunday are what generate surplus. And in the continuous process a solar plant of this size does not export: in the model it could reach 1.5 kWp per kW of load while exporting at most 1% of what it produces, so load stops being the limit and roof area, budget or the 1 MW threshold take over.

Off-shift consumption matters for the one-shift plant. If it is 15% of the shift load instead of 25%, solar covers somewhat more of the consumption (38% to 45%) but exports between 18% and 21% of what it produces; if it is 40%, it covers 30% to 36% and exports 5% to 8%.

The example is modelled, not a real project. A fictitious two-shift plant in Bogotá, with 400 kW of shift load and 100 kW off-shift, consumes about 2,379 MWh a year (about 198 MWh a month). With 500 kWp, PVGIS gives production of about 661 MWh a year: about 616 MWh are self-consumed, 25.9% of consumption, and about 45 MWh are exported, 6.7% of production, practically all of it on Sundays.

Three practical consequences. With 400 kW of demand and 198 MWh a month, the company meets the threshold to be treated as an unregulated user (0.1 MW or 55 MWh a month, measured at a single individual delivery point, under Resolución CREG 131 de 1998); if it contracts as an unregulated user, the value of each avoided kWh comes from its contract and not from the regulated unit cost, as the article on the company bill explains. The 45 MWh on Sundays are surplus from a plant of more than 100 kW, with the charges per swapped kWh explained in the article on surplus sales. And 500 kWp on a flat roof need, using the criteria in the article on roof area, roughly 2,500 to 2,600 m² of array field at 10° (about 3,200 using NREL's 70%), before subtracting edges, walkways and equipment.

Capacity is chosen with the three caps from the article on roof area —the energy to be covered, the usable daytime load and the available area— and the lowest one wins. In a factory, the load cap is calculated with the load that holds during the central sunny hours of production days, neither the peak nor the daily average; the afternoon, if the shift ends before sunset, and the non-production days tell how much is exported, as the model shows.

Those days govern the surplus. A Sunday is one seventh of the year's solar production (14.2% to 14.3% in the PVGIS data); if the plant also stops on Saturday, it is two sevenths, plus public holidays and shutdowns. If the base load on those days is low, a good part of that production goes to the grid: in the model, about half of what is produced on Sunday with a 25% base load (47% to 55% depending on the city) and about two thirds with 15%.

There are three ways to handle it, and the proposal must say which one it uses: accept it as surplus, at the value given by the project's regulatory category (energy credit for a small-scale self-generator, a contract with a representative for a large-scale one); curtail it with zero export, with a simpler connection procedure in exchange for losing that energy; or move shiftable loads, such as charging forklift batteries or pumping to tanks, into the sunny hours of those days. Which is best is a calculation, not a rule, and the article on surplus sales explains it.

The value of each avoided kWh comes from the plant's own bill, not from a published tariff: the voltage level changes the unit cost, and an industrial user has not paid the 20% solidarity contribution since 2012 if its main activity is listed in Decreto 2860 de 2013 and it has applied to its retailer for the exemption. Decreto 2860 is still in force, but the exemption is under discussion: the Finance Ministry's draft decrees (August 2025 and July 2026) would keep it for manufacturing (CIIU codes 101 to 332) and remove it from other activities that have it today, and as of 1 October 2026 a bill to abolish it for all industrial users was on record in the Senate. That is why the project should be assessed with and without it, as the article on the company bill explains.

Motors, compressors and cooling equipment consume reactive energy as well as active energy. Resolución CREG 101 035 de 2024 requires the inductive power factor of a user's installation to be 0.90 or higher, determined hourly (per billing period if the meter does not record hourly); whoever exceeds the limits pays for transporting the excess reactive energy under Chapter 12 of Resolución CREG 015 de 2018. How it appears on the bill is covered in the article on the company bill.

A solar plant operating at unity power factor reduces the active energy the factory draws from the grid, but not the reactive energy. Our own calculation shows it: a plant with 400 kW of load and a power factor of 0.92 draws about 170 kvar; if mid-morning the solar plant delivers 200 kW, the grid now supplies 200 kW with the same 170 kvar, and the power factor metered in that hour falls to 0.76. Because the rule is hourly, the problem appears precisely in the sunniest hours. And in the hours when the plant exports, such as Sundays, the grid delivers no active energy: all the inductive reactive energy recorded in that hour is then charged (literal b of Chapter 12 of Resolución CREG 015 de 2018).

There are three answers, and they are defined in the design. Compensation with a capacitor bank, whose controller should be reviewed because solar changes what it measures, without overshooting: capacitive reactive energy has its own limit (a capacitive power factor of 0.90 at levels 1 and 2, 0.95 at level 3 and 0.98 at level 4, per billing period, under Resolución CREG 101 035 de 2024). The inverter itself: commercial inverters allow the power factor to be adjusted (Huawei's SUN2000-150K-MG0, for example, between 0.8 leading and 0.8 lagging), but RETIE requires the inverter to keep a power factor above 0.9 when it delivers more than 50% of its rated power (literal f of section 2.3.23.1 of Book 2 of Resolución 40284 de 2026), which limits how much reactive power it can supply in the hours of highest output. And the regulatory one: Decreto 929 de 2023 exempts users with small-scale self-generation from non-conventional renewable energy sources (FNCER), such as solar, from reactive-energy charges. If the retailer applies it, neither the export-hour charge nor the capacitive one should reach that self-generator; before counting on it, ask the retailer how it applies it, and remember that it does not reach a plant of more than 1 MW, which falls outside that exemption.

A plant that receives power at medium voltage, usually at level 2 (1 kV to less than 30 kV, under Resolución CREG 015 de 2018), and has its own transformer distributes at low voltage inside the factory. The level changes the price of the avoided kWh, as explained in the article on the bill; in the design it also raises two questions.

The first is which switchboard each inverter connects to, and at what voltage. RETIE sets 120 V phase-to-ground and 208 V phase-to-phase as the standard voltages for general-use three-phase low-voltage networks, and warns that using equipment at other voltages requires special transformers (Title 9 of Book 3). But that is the general-use network: a plant with its own transformer may distribute at another voltage, such as 440 or 480 V, and what counts is the voltage of the switchboard each inverter connects to. Inverters are also built for different voltages: Huawei's 150 kW SUN2000-150K-MG0 runs at 380, 400 or 480 V, and its 80 kW SUN2000-80K-MGL0 at 220 V. If the switchboard voltage does not match the inverter's, a coupling transformer is needed, with its cost, losses and footprint, and the proposal should say so from the start.

The second is capacity. For the 1 MW limit of small-scale self-generation, what counts is the sum of the inverters' rated AC capacities (Article 3 of Resolución CREG 174 de 2021, which refers to Resolución UPME 281 de 2015). If the plant exports surplus and connects at voltage level 1, before applying for connection you must check on the network operator's website that the circuit has availability (Article 6 of the same resolution); self-generators that do not export are not subject to those limits. And the plant's own transformer should be reviewed, because exported energy flows through it in reverse. Which regulatory category fits a large plant is covered in the article on self-generation and distributed generation.

It is the question asked most often in a plant, and it deserves a straight answer: a grid-connected solar plant does not keep the factory running during an outage. RETIE requires the grid-connected inverter to have islanding protection, so that the generator stops energising the local network operator's grid within no more than 2.0 s of losing grid voltage (item 1 of literal j of Article 3.17.21 of Book 3). It is a protection for people working on the grid, and that is why a conventional grid-tied inverter disconnects when the grid goes down and comes back when it is restored.

The question has context in 2026. In response to the 2026-2027 El Niño event, CREG adopted transitional measures to secure supply (Resolución CREG 101 127 de 2026) and, with Circular CREG 343 de 2026, started the supply-shortage risk period on 5 September; the circular itself states that this does not, by itself, mean scheduled rationing. A solar plant reduces what the factory draws from the grid during sunny hours, but it does not protect it from an outage.

For solar to sustain loads during an outage, a different architecture is needed: inverters able to operate islanded from the grid, storage, and a means of separating the installation from the grid before feeding the loads. It means more investment and more paperwork: when storage is included, Resolución CREG 101 113 de 2026 adds technical requirements and tests to the connection procedure of Resolución CREG 174 de 2021. The sensible starting point is to identify which loads are critical and for how long they must be sustained; when batteries are justified is covered in the guide on panels with batteries.

If the plant already has an emergency generator, the design must define how the two coexist: what the solar system does when the generator starts and how it is coordinated with the transfer system. It is a question for the electrical designer, and the answer should be written into the proposal.

Roof mounting and cable runs can progress while the plant is producing, but two moments affect operations: work on roofs above process areas and the final connection to the switchboards, which requires de-energising them. It is best to schedule that connection during a shutdown that is already planned —a Sunday, a public holiday, a maintenance stop— and to ask the installer for a written outage plan.

Roof work is work at height, regulated by Resolución 4272 de 2021, and the contracting company also has verification duties; the article on choosing a provider details what to ask for. RETIE also requires the design to define the means of access and movement for installing, inspecting, operating and maintaining the panels, and the inspection body must be able to verify them in the finished installation (literal o of Article 3.17.23 of Book 3).

The timeline also depends on procedures with their own deadlines: the connection with the network operator and, if the company is going to use the tax incentives, the certification with UPME, which is best requested early so as not to advance the VAT on the equipment, as the article on tax benefits explains.

With this data, a proposal for a manufacturing plant comes from calculations rather than assumptions, and it should return consumption coverage and surplus by month and by type of day, how the power factor is handled, the voltage and connection point of each inverter, and what happens during an outage.

If you want your plant assessed with these criteria, you can request a commercial quote through Nodosolar. Whoever prepares it, it is worth arriving with the following: