- The roof is a structure with a maintenance path
- Daytime load is the reason the building was a candidate
- Inverter choice follows the roof, not a default
- Interconnection and shutdown are part of the building permit
- Frequently asked questions
- Is a commercial roof array still “distributed” if it is over a megawatt?
- Will the roof cover the building’s whole bill?
- Can the facility electrician maintain the DC side?
- Sources and references
A distributed solar system on an industrial or commercial building works when the roof, the parking area, or the yard can carry it, and when the building actually uses electricity while the sun is up. The array is still distributed if it sits on the customer side of the meter and serves that site or the local distribution network. A large warehouse roof does not become a central power station because the kilowatts are large. Pacific Northwest National Laboratory’s size note is the electrical hinge: arrays under about 50 kilowatts may land on the building service, and larger ones may need a transformer to reach grid voltage. Their 480-volt-to-15-kilovolt step-up is an example of that larger case, not a rule that every factory uses 15 kilovolts.
The energy hinge is the load shape. A cold store, a day-shift factory, and an office that empties at 5 will not see the same fraction of solar used on site. Exporting the rest is fine if the tariff pays for it and the interconnection allows it. It is a different project from “the roof will run the plant.”
The roof is a structure with a maintenance path
PNNL describes roof arrays as typically fixed, anchored to the structure or ballasted, and often smaller because area and structure run out. Maintenance cost follows access. A membrane roof that cannot be walked, a high bay that needs a lift, and a production line that cannot stop for a penetration repair are commercial constraints a house does not have. The inspection items do not change: penetrations, loose mounts, corrosion, water in equipment, signs of arcing. Modules stay live in daylight. The person who may open DC wiring is still a trained maintainer, even when the building has its own electricians for other work. Those electricians need the photovoltaic shutdown procedure, AC disconnect then DC, posted at the inverter.
Soiling on a low, flat commercial roof in a dry industrial area is the case PNNL flags for more frequent washing, including a possible monthly summer wash. A steeper, rainy site may never need it. Flat roofs collect dust and they also collect ponding. A wash plan that ignores drainage can put water where the membrane is already tired. Follow the roof maker and the module maker. Production, not the appearance of the glass, decides whether the wash is earned. PNNL treats roughly 10 percent variation in sunlight as ordinary.
Daytime load is the reason the building was a candidate
Match the solar day to the shift. A plant that runs one weekday shift will use midday solar and will export or curtail on Sundays if the interconnection says so. A plant that peaks in the evening will buy power then unless storage is allowed to discharge into that peak. Storage is a second project, with its own power rating and its own interconnection settings. It is not included because the modules are on the roof. The jobs storage can do, and the jobs it cannot, are separated in what solar storage is actually for.
Monitoring has to be good enough to defend the bill. A single daily total cannot show that one inverter on a large roof stopped at 10:40 while irradiance was normal. PNNL’s minimum is a monthly subtraction from the production meter if no software exists. A commercial roof with several inverters outgrows that minimum quickly, because the failed unit is a fraction of the total until the fraction gets large. What to specify in the monitor is in solar monitoring systems.
Inverter choice follows the roof, not a default
PNNL already lists the arrangements a building might use: one large central inverter, several, or many microinverters. Large central units on that page include internal transformers. Smaller inverters and microinverters typically do not. A 2017 guide for utility-scale plants, which is a different size class but the same architecture choice, notes that string inputs can track irregular arrays more finely and that a central failure takes a larger block offline. A commercial roof with many orientations, vents, and a tree line is the irregular case. A single large unshaded warehouse roof may not need a tracker per few modules. The 2017 capacity bands in that guide, up to about 4 megawatts for central units and about 25 to 125 kilowatts for string units, are that year’s table. Nameplates have moved. Use the datasheet. Do not use 2017 prices, which the guide itself expected to change.
Heat in a rooftop inverter, or in an electrical room below a hot roof, shows up as afternoon derating. The Sandia and NREL thermal study tied component temperature to dissipated power, season, and wind. A roof parapet that blocks airflow is a commercial detail with a yield cost. Filter cleaning stays on the manufacturer’s interval. It is not optional because the building is large.
Interconnection and shutdown are part of the building permit
The array is not installed when the modules are on the roof. It is installed when the utility has accepted the settings and the disconnects are where the code at that site requires them. Under IEEE 1547-2018, as the NREL primer states it for US interconnection, the inverter must stop energizing an island within 2 seconds, including with support functions enabled. A factory that wanted the roof to carry the line through an outage needs a designed island, not the default grid-tied mode. Fire access and roof pathways are local. This article does not substitute for the fire marshal. It does say that a distributed system which cannot be shut down in the order PNNL describes is not finished, however good the production model looked.
Where else a distributed system can sit, if the roof is the wrong surface, is in where distributed solar can be installed. A yard or a canopy can be the better structure. They do not remove the need to match the load.
Frequently asked questions
Is a commercial roof array still “distributed” if it is over a megawatt?
Yes, if it serves the site or the distribution network rather than operating as a central plant on the transmission system. PNNL’s distinction at that scale is practical: larger arrays may need a transformer, and the example step-up is not a universal voltage.
Will the roof cover the building’s whole bill?
Only if the building uses power when the array produces it, and if export rules and tariffs do the rest. An evening peak is not served by midday solar unless storage or another supply is there. Compare a year of interval data with the production shape before treating a roof area as a bill eraser.
Can the facility electrician maintain the DC side?
Only if that person is qualified for the photovoltaic voltage and follows the posted shutdown, AC then DC. PNNL states that modules are live in daylight regardless of disconnect status. Familiarity with the building’s AC panels is not that qualification by itself.
Sources and references
Service size, transformers, roof mounting, soiling, shutdown, and inverter arrangements: Pacific Northwest National Laboratory. Dated central and string architecture, not as a 2026 price: ICF, 2017. Rooftop inverter temperature drivers: Thermal study of inverter components. Cease-to-energize: NREL/Sandia IEEE 1547-2018 primer.