String or central inverters: which architecture suits your commercial array?

The two designs differ in how they are physically laid out and how the array connects to them.

• String inverters. A string is a series-connected run of panels. A string inverter is a modular unit that takes several of these strings, each on its own tracker, and there are several units spread across the site. Commercial units commonly carry two to four trackers, with some newer models carrying more, and unit ratings run up to a few hundred kilowatts. Because the units are distributed, the DC cabling runs are shorter and the array is broken into independent sections.
• Central inverters. A central inverter is one large unit, typically rated in the megawatts, that takes the whole array, or a large block of it, through a small number of trackers. It sits next to the medium-voltage transformer, with long DC runs feeding it from across the array. It assumes the array in front of it is uniform, because its few trackers cannot follow many different sections independently.

That single difference, many small distributed units against one large central unit, is what drives every other trade-off below: how the array copes with mixed orientations and shade, what happens when a unit fails, what it costs per kilowatt, and how it is serviced over a twenty-year life.

A maximum power point tracker, or MPPT, continuously finds the voltage and current that extract the most power from the strings connected to it. The number of independent trackers is the single most important difference between the two architectures on a commercial roof.

Panels on one tracker should share orientation and tilt, because the tracker can only find one optimum operating point. Mix a south-facing run and an east-facing run on the same tracker and it settles on a compromise that short-changes both. A string design carries several independent trackers, so each roof plane, each orientation and each separately shaded section runs on its own tracker without dragging down the rest. A central inverter has only a handful of trackers for the whole array, so it works best where the array is one large uniform plane and there is nothing to isolate.

This is the engineering reason a string design suits the typical commercial building: multiple pitches, a north light or two, plant and parapets that throw shade across part of the roof through the day. A central design suits a flat, open, single-orientation ground-mount where every panel sees the same sun. Independent trackers reduce the mismatch loss that mixing orientations causes; the size of that gain on your roof is what the PV*SOL model returns, rather than a figure carried in from another site.

The two architectures fail and are repaired in very different ways, and that difference matters more on a commercial asset than the headline kilowatt cost.

With a single central inverter, one failure can take the whole block, or the whole site, offline until it is repaired. With distributed string inverters, a failed unit removes only its share of the capacity. As an illustration, one large central unit failing can mean total shutdown, whereas on a site built from twenty smaller string units, one failure leaves the great majority of production still running while the unit is swapped. For a business that depends on the generation, that difference in single-point-of-failure exposure is often decisive.

Serviceability follows the same pattern. A string unit is a small, comparatively inexpensive item that can be swapped without specialist plant, so a fault means limited downtime and a fast, contained replacement. A central inverter is built for a long design service life and is repaired in place by specialists, but a fault carries a larger production impact while it is out. Over a twenty-year life, spare-part availability is a real consideration for both, since a discontinued platform has to be worked around; a discontinued string model is usually cheaper and simpler to design around than a discontinued central platform. We weigh this against the maintenance plan for your site rather than treat the inverter as a fit-and-forget item.

Central inverters are cheaper per kilowatt at scale. Fewer units, less wiring complexity and economies of scale all pull in their favour, and that advantage grows as the system gets larger and the array more uniform. A string design costs more per kilowatt at the equipment-and-install level, though the gap is modest, and the redundancy and serviceability it buys often justify it on a commercial roof. There is no clean megawatt figure where one becomes cheaper than the other; it depends on the array's uniformity, the roof's complexity and how the project is phased, so we frame the decision by site type and model the cost for your project rather than quote a crossover threshold.

Phasing is where string architecture earns its place on many commercial buildings. A string design scales incrementally: units are added as roof areas come on line or as the array is built out over several phases, which suits a rooftop that grows with the business. A central inverter is sized once for a fixed block and suits a single large build that is commissioned in one go.

The industry direction reflects this. String architecture is now competitive well into the megawatt range, and so-called string power stations, racks of string units in an enclosure beside the medium-voltage transformer, are increasingly used even at larger scale, combining the redundancy and serviceability of string with a single centralised maintenance location. Central inverters still dominate the very large, uniform plants, and where central architecture and deep grid-side control are in play we often reach for the SMA commercial inverter range, which pairs a central platform with the SMA Power Plant Controller for export and grid management. Whichever architecture suits your site, the inverter needs G99 compliance evidence for the exact model, because the choice between string and central does not change the G99 obligation. The current edition of G99 grants Fully Type Tested status only to units of 50 kW or less, so a larger unit, including any central inverter, normally connects on a partial type test or manufacturer's information, with the rest demonstrated through the application and commissioning tests on site. We verify that certification for the specific unit before it is specified.

The architecture choice changes the DC side of the system, and the DC side is where the recognised rooftop PV fire risk sits. A string of modules carries several hundred volts in daylight, present continuously with no zero-crossing, so a loose, corroded or badly made DC connection can strike an arc that does not self-extinguish the way an AC fault does. RC62, the joint fire-and-rescue and industry guidance, and the wider RISCAuthority fire-safety literature are written around this DC hazard, and the inverter architecture decides how much DC cabling exists and how far it runs.

A string design distributes the inverters across the roof, so the high-voltage DC runs are short and local to each unit and the array converts to AC close to the panels. The trade is a larger count of DC string connections and connectors, each a potential fault point that has to be made with one connector type, mated like-for-like, with factory-style crimps. A central design has far fewer connections, but it pulls long high-voltage DC runs across the whole array to one unit beside the transformer, so a greater length of the system stays live at high DC voltage and further from the point of isolation. Neither is safer in the abstract. They move the risk to different places, and the survey decides which profile suits the building, the roof construction and the access a fire service would need.

This is also where isolation and shutdown differ in practice. Because daylight keeps the DC side energised, a fault cannot be cleared by switching the building off, so clear, labelled DC isolation matters under either architecture. The wiring-regulation basis for those terminations and isolation points is set out in our guide to BS 7671 and solar, and the fire-safety detail of the DC side sits in our note on RC62 and rooftop fire risk. We design the DC layout, the connector regime and the isolation strategy to the same standard whichever architecture the roof calls for, because the choice changes where the DC risk lives, not whether it has to be controlled.

A commercial array is a twenty-year asset, and how you see a fault and how you fix it differ sharply between the two architectures. A string design gives granular visibility: each unit and each tracker reports separately, so a drop in one section is localised to a unit or a string rather than hidden in a single aggregate figure. That granularity is what turns a vague "the site is down on yield" into a specific string to inspect, which is the difference between a same-day fix and weeks of lost generation. A central design reports at the level of one large unit, so a partial fault, a single failed combiner or a shaded sub-array, is harder to isolate from the monitoring alone and usually needs an on-site inspection to locate.

That visibility feeds directly into the operations and maintenance plan. PAS 2038, the British Standards Institution specification for commercial building retrofit, frames the case for designing the monitoring and the maintenance regime in from the start rather than bolting them on, and the inverter architecture sets what that regime can detect. A string fleet lends itself to remote, per-unit fault-finding and a stock of swap-out spares; a central platform leans on a planned, specialist service contract and in-place repair. Our performance audit explains how that monitoring data is read, and our commercial maintenance guide sets out the service regime each architecture implies.

Warranty and long-term support follow the same split. String inverters typically carry a standard manufacturer warranty with an option to extend, and because the units are inexpensive and interchangeable a failed one is a contained replacement against held spares. A central platform is a larger capital item with a longer design service life, usually supported by a manufacturer service agreement rather than a shelf spare, so its warranty and support contract carry more weight in the decision. On a system outside MCS, none of this is assured by a domestic badge: the support path is whatever the manufacturer agreement and your maintenance contract actually say. We check the warranty terms, the spare-part position and the service route for the specific unit before it is specified, and we weigh that against bankable-brand support and the build cost modelled for your project.