A centralized solar inverter and a string inverter do the same conversion, DC from the modules into AC for the grid, at different scales and with a different failure. In the 2017 ICF guide that is still the clearest public comparison of the two plant layouts, central units of that year were described up to about 4 megawatts AC, and string units at about 25 to 125 kilowatts AC. Those bands are a 2017 catalog. Nameplates have moved, so use the datasheet in front of you for kilowatts. Use the guide for the layout differences that were true of the architectures, and check each one against the current manual before you treat it as still true of a 2026 product.
Pacific Northwest National Laboratory describes the same family without those kilowatt bands: one large central inverter, several central inverters, or many microinverters. Microinverters are a third arrangement, at the module, and they are not what this comparison is about. Large central units, on PNNL’s page, use internal transformers. Smaller inverters typically do not.
How the wiring is organized
Both layouts start with modules in series strings. The ICF guide then splits them. Central plants combine those strings in DC combiner boxes and bring the DC to one large inverter, often on a skid that may also hold a medium-voltage transformer. String inverters take strings more directly, often without those DC combiners, and their AC outputs are collected and then stepped up. The guide’s loss note follows from that: central layouts tend toward shorter AC runs, string layouts toward shorter DC runs. Which loss dominates is a layout question, not a moral one.
Installation, in that same 2017 table, was a crane and a pad or skid for the central unit, and a rack that two people could lift for the string unit. Commissioning was described as heavier on the central side. Those are logistics. They do not decide energy yield.
What happens when one unit fails
This is the difference that still organizes the choice. If a central inverter stops, the guide says a large share of the plant stops with it, until that unit is repaired. If a string inverter stops, a small group of arrays stops. Availability then depends on spares and on how fast a qualified person can act. The guide describes central units as field-serviceable, with fans, filters, and power modules that are maintained, and with owners commonly planning a major refurbishment around year 10 and year 20. It describes string units of that period as typically not field-repaired: the unit is replaced, so the spare is a whole inverter, and there is no filter to clean because the guide’s string units were convection cooled.
Read the last sentence with the date on it. A 2026 string inverter may have a fan. PNNL’s instruction, clean intakes on the manufacturer’s timing, applies to the unit that has an intake. The architectural point to keep is the spare strategy. A plant that chose many small inverters and stocked none has traded a large outage for a long tail of small ones. A plant that chose one central stage and stocked no power module has the outage the guide described.
| Question | Central layout, as the 2017 guide described it | String layout, as that guide described it |
|---|---|---|
| DC collection | Combiner boxes, then one inverter | Strings into the inverter inputs |
| Trackers | Single or multiple | Multiple, so unlike arrays can be separated |
| Cooling in that table | Air or liquid, with filters to maintain | Convection, no fan or filter in that description |
| After a failure | Field repair; a large block is down | Replace the unit; a small block is down |
| Design-life column | 10 to 25 years with refurbishment | 20 years or more, limited field service |
The design-life row is the author’s planning column, not a measured survival study. Warranties in the same table were about 5 to 10 years for central units and about 10 years for string units, again as of that paper. PNNL’s later best-practice page still describes typical inverter warranties as about 5 years with extensions to 10 or 15, without splitting the architectures. Use the contract you are offered.
Where the extra trackers earn their place
The ICF guide’s performance claim is narrower than advertising. More trackers let each group of strings sit at its own maximum when shade, tilt, or a damaged module makes the groups unlike. On a uniform field the gain may not exist. On irregular ground it might. The guide does not give a percentage you can paste into a model. It says evaluate the site. Potential-induced degradation is the other check it insists on for transformerless, floating-ground designs: ask for the module’s test, because the guide’s view was that slowing the mechanism is not the same as immunity.
Both layouts still have to meet the grid code. IEEE 1547-2018’s cease-to-energize limit of 2 seconds, described in the NREL primer, does not care whether the inverter is large or small. A central unit and a string unit can both ride through and both trip. The settings file is per the interconnection, not per the marketing name. Monitoring is where the layouts feel different. A string layout produces more separate power readings. That is useful only if someone looks at them. What those readings need to include is in solar monitoring.
How to choose without a 2017 price
Ask what share of the plant stops when one inverter stops, and what spare you will actually stock. Ask whether this model is field-repaired or replaced, and whether the cooling parts exist. Ask how many trackers you need for the shade and the tilt you really have. Ask for the weighted efficiency and the continuous rating at the DC voltage you will use, which is a datasheet question either architecture must answer. Ignore any cost gap that cites a year before the project. The 2017 guide expected string and central equipment prices to move closer and said balance-of-plant cost could reverse a simple comparison. It did not publish a 2026 price, and neither does this article.
Frequently asked questions
Are string inverters always more efficient?
The 2017 guide does not say that. It says a possible energy gain from finer tracking on irregular sites, which has to be modeled, and it keeps conversion efficiency as a separate datasheet item. A peak number is not the comparison.
Do central inverters last longer?
The guide’s design-life column shows an overlap, not a winner: central units with planned refurbishment, string units with replacement rather than repair. PNNL’s typical warranty language is about 5 years unless extended. Measure the offer, not the category.
Is a string inverter a microinverter?
No. A string inverter takes a series string, or several. A microinverter, in PNNL’s list, sits at the module. Both are smaller than a central stage. They are not the same spare part.
Sources and references
Layout, 2017 capacity bands, trackers, cooling, failure domain, and the design-life column: ICF, A Guide to Choosing Central Versus String Inverters, 2017. Central, string-scale, and microinverter arrangements, and typical warranties: Pacific Northwest National Laboratory. Grid behavior required of either size: NREL/Sandia IEEE 1547-2018 primer.
