Power conversion is the plant’s heartbeat, and topology decides whether you have one industrial-grade heart or dozens of small ones beating in formation. Solar fought this same war a decade ago; storage is fighting it now with its own twists — because in a battery plant, the inverter isn’t just converting, it’s managing strings of electrochemistry with personalities.
For a visual companion, explore the interactive MV Skid Structure diagram on BESS.Engineer.
The two topologies
Central PCS. A few large inverters — megawatt-class units, typically skid- or container-mounted with their MV transformer — each serving many battery racks paralleled onto a common DC bus. The classical utility architecture: fewest boxes, familiar to grid engineers, economical iron.
String PCS. Many smaller inverters, each dedicated to one battery container, rack group, or string, aggregating on the AC side. The distributed philosophy: granular control, modular failure, factory-style repetition.
The availability argument — string’s headline
When a central inverter trips, its entire slice of the plant — often a double-digit percentage — goes dark until repair. When one string unit fails, you lose a small percentage of output and keep operating; modular designs are marketed on precisely this arithmetic, with a single unit’s failure costing a few percent of plant output rather than a block. For revenue models built on availability guarantees and capacity commitments, graceful degradation is money. String units are also swap-repairable by smaller crews, versus central units whose repairs are events with cranes in them.
The battery-management argument — string’s quieter advantage
Here’s the storage-specific twist solar never had: batteries on a shared DC bus must electrically agree. Parallel many racks behind one central inverter and mismatched strings — different ages, temperatures, or augmentation vintages — fight, circulate current, and force conservative operating limits. Give each container its own conversion and every battery group gets its own operating point: individually optimized charge profiles, cleaner SoC accounting, and — decisively for long-lived plants — the freedom to add new-generation containers later without negotiating voltage compatibility with a decade-old DC bus. String topology is, quietly, an augmentation strategy.
The central case — still standing
Central PCS answers with economics and grid muscle: fewer, larger power stages historically deliver lower $/kW conversion cost, fewer network nodes, fewer things to firmware-manage, and big machines with the electrical authority grid operators like for fault current and grid-forming duty. For very large plants with homogeneous batteries and strong service organizations, central remains a rational default — and hybrid layouts (central conversion, sectionalized DC) split the difference.
The market’s direction, though, has been visible for years: as containers standardized and their integrated electronics matured, string-style and container-level conversion kept gaining share, and the AC-block product philosophy is essentially string topology sold as a product. The cost gap narrowed; the granularity advantages didn’t.
Choosing like an adult
Weight the decision by what your project actually fears: if availability LDs and augmentation dominate your risk register, granularity pays; if $/kW and grid-services muscle dominate, central defends itself. And whichever you choose, the plant-level controller inherits the job of making many heartbeats look like one plant at the POI — the integration task where, in my experience, more commissioning schedules die than anywhere else.
FAQ
Which is more efficient? Peak efficiencies are comparable in modern units; real-world differences come from partial-load behavior — many small units can shut segments down and keep the rest at sweet-spot loading.
Is string PCS just the solar playbook repeated? The economics rhyme, but the battery-management dimension — per-string operating points and mixed-age fleets — is storage’s own argument, and often the deciding one.
Can topologies mix? Yes — large sites increasingly zone their architecture, and repowering projects graft string conversion onto legacy layouts. Consistency per zone matters more than site-wide purity.
PCS selection, protection coordination, and the control integration behind it all get dedicated modules in my Grid-Scale BESS: Complete Guide.