What is a unit substation?

A unit substation is not one device. It is three pieces of equipment bolted and bussed together as a single tested assembly:

• Primary section. Where medium-voltage service lands. It contains either a fused load-interrupter switch or a draw-out circuit breaker, plus incoming cable or bus termination.
• Transformer section. The step-down transformer itself, throat-connected to the sections on either side — no field cable between them.
• Secondary section. Low-voltage switchgear or a switchboard that breaks the transformer output into feeders, with metering and protective relays.

The defining word is close-coupled. The transformer bushings enter the adjacent sections through a flanged throat with a flexible bus connection, so the whole line-up is one continuous, factory-terminated electrical path. That is the difference between a unit substation and three separately-purchased components that happen to sit near each other.

This naming trips up nearly everyone the first time. The prefix describes the secondary voltage, not the incoming one.

• A secondary unit substation steps medium voltage down to utilization voltage — typically 13.8 kV or 12.47 kV in, 480Y/277 V out. Its secondary is below 1000 V. This is what almost everyone in a data center means when they say "unit sub."
• A primary unit substation steps a higher medium or sub-transmission voltage down to a distribution medium voltage — 34.5 kV in, 13.8 kV out, for example. Its secondary is still above 1000 V and feeds further downstream gear.

A large campus usually has both, in series: a primary unit substation at the site boundary establishing the campus MV distribution voltage, then a secondary unit substation per data hall or pod turning that into 480 V.

On the secondary side, 480Y/277 V is the common North American choice. Some operators specify 415Y/240 V instead, which lets server power supplies run at 240 V line-to-neutral and removes one transformation step at the rack PDU.

The single biggest cost-and-capability fork in a unit substation is what sits in the primary section.

A fused load-interrupter switch is an air-break switch rated to make and break load current, backed by current-limiting power fuses that handle fault current. It is compact, inexpensive, and mechanically simple. The trade: a fault blows fuses that have to be physically replaced, you cannot reclose remotely, you get no adjustable time-current curves to coordinate with, and there is no relay package to trend or interrogate. Fuses also blow single-phase, which can leave a three-phase load single-phased if the design does not catch it.

A draw-out vacuum circuit breaker in metal-clad construction gives you relay-based protection with settable curves, a resettable device after a fault, remote operation, and — the operational argument that usually decides it — the ability to rack the breaker out of its cell and work on one feeder while the main bus stays energized. Grounded metal barriers and automatic shutters make that safe to do.

For a facility that has to be maintained live, breakers win. For a small, non-critical, or seldom-touched load, a fused switch is proportionate. Data centers land on breakers for anything in the critical path and often accept fused switches only on house or non-critical loads.

The transformer section usually decides where the line-up can physically go.

• Liquid-filled transformers are efficient and thermally forgiving, but mineral oil brings NEC vault and containment requirements indoors. Less-flammable ester fluids relax some of that.
• Cast-coil transformers encapsulate the windings in epoxy — mechanically robust, moisture-tolerant, well suited to indoor and harsh environments, and the common choice inside a data hall.
• VPI dry-type transformers are vacuum-pressure impregnated, lighter and lower first cost than cast coil, but less tolerant of humidity and contamination.

The impedance you specify (5.75% is a common default) sets the available fault current on the secondary bus, which in turn sets the interrupting rating you must buy in the secondary section. Choosing it casually is how a project ends up with under-rated gear. Forced-air cooling (an AA/FA rating) gives extra capacity above the self-cooled base rating, which is often how N+1 headroom is bought.

Enclosures follow the location: NEMA 1 for indoor line-ups, NEMA 3R or a walk-in weatherproof housing for outdoor pads. Outdoor placement moves the transformer fire and containment problem outside the building envelope, which is why so many campus designs put the unit substation on a pad next to the hall rather than inside it.

A single-ended unit substation has one primary, one transformer, one secondary bus. Simple, and a single point of failure.

A double-ended unit substation puts two primaries and two transformers on either end of a secondary bus that is split by a tie breaker. Each transformer normally carries half the load. If one source or transformer is lost, the tie closes and the surviving transformer carries the whole bus — which means each one has to be sized (or force-cooled) for the full load, not half of it.

Double-ended is the distribution-level expression of N+1, and it is why you see two feeders, two transformers, and a main-tie-main arrangement repeated across data-center one-lines. The operational catch is that closing the tie momentarily parallels two sources unless the scheme is open-transition, so the control logic — and whether it is manual, automatic, or fast-transfer — belongs in the specification, not in the field.

You can buy a switch, a transformer, and a switchboard separately and connect them on site. Large operators generally do not. The reasons are practical:

• One party owns coordination. Short-circuit withstand across the throat, bus bracing, and ground-bus continuity are engineered once, by one vendor, and tested as an assembly.
• Fewer field terminations. Cable terminations made in a trailer at 2 a.m. are a leading source of commissioning defects. A close-coupled throat eliminates most of them.
• One schedule. Piece parts arrive on separate lead times, and the last one to show up sets the energization date. An assembly arrives matched.
• Repeatability. A data-center campus is the same pod copied many times. A standard line-up drawing that has already been built and tested is worth more than a marginally cheaper bespoke arrangement.

The schedule argument has become the dominant one. OEM new-build lead times for the transformer and switchgear inside a unit substation currently run years, not months, which is why in-stock and surplus assemblies get specified into projects that would once have ordered new. Line-ups are available through Pantheon under electrical equipment.