In a photovoltaic plant the transformer works in reverse of the classic case: it steps the voltage up to inject. That inversion, plus a few PV particularities, genuinely changes the specification.
Panels (DC) → inverters (LV AC) → step-up transformer LV → MV → delivery / evacuation substation → public grid. Up to a few MW, evacuation runs at MV on the distribution network; large parks step up to HV towards the transmission operator via a dedicated HV substation. The step-up transformer itself lives at the foot of the inverters, often in compact outdoor substations (IEC 62271-202) spread across the field.
Utility-scale inverters rarely output 400 V: depending on the model, the nominal voltage typically ranges from 400 to 800 V (600 V, 660 V, 690 V, 800 V and others). The transformer is designed exactly for your inverters' voltage: the first piece of information to lock down, inverter datasheet in hand. It is the perfect example of custom-built: there is no generic "solar transformer".
A great PV specificity: the transformer with two separate LV windings ("dual secondary", a Dy11y11-type vector group or equivalent). Each secondary feeds its own inverter group, electrically independent of the other, which limits circulating currents between inverters and saves one transformer in two at plant scale. Declare it at enquiry stage: it changes the design in depth.
The solar profile is unique: zero at night, a ramp in the morning, a plateau at noon, every day. That daily thermal cycling fatigues a transformer differently from a constant industrial load, and inverters inject their share of harmonics (milder than servers, but real). The design accounts for it: temperature rises computed on the real profile, and at night the unit stays energised paying only its no-load losses, one more reason to specify them low.
Injecting is not the same as drawing: a producer's connection follows a specific system-operator procedure (study, connection agreement, metering), and the delivery substation carries a decoupling protection, the device that disconnects the plant if the grid disappears. These requirements (NF C 13-100 on the substation side in France, operator requirements on the grid side) are handled with your design office; we align the transformer and the substation with the validated diagram.
General design pointers; exact voltages, vector group and protections are set to your inverters, your architecture and the connection agreement, to be confirmed with the design office and the system operator.
Your inverters' voltage, taken from their datasheet: 400, 600, 660, 690 or 800 V depending on the model. It is the first sizing input; the LV voltage field of the configurator accepts all these values via Other.
On a plant of a certain size, yes: one transformer serves two independent inverter groups, which reduces the number of units, substations and MV connections. It simply has to be planned from the design stage, not as a modification.
For pure injection, what matters is a correct step-up design: taps, losses, temperature rise on the solar profile. True bidirectional capability becomes decisive as soon as battery storage joins the plant; state it in the request.
The system operator (in France, Enedis or a local DSO for MV; RTE for HV) through a study, a connection agreement and the metering. The delivery substation carries the decoupling protection; we align the transformer and the substation with the validated single-line diagram.