In short. Above 2 MVA the question “which generator” is the last one to answer. The plant is several paralleled sets on a common bus, usually at 11 kV, with synchronising controls, breakers, protection, transformers and a power management system that decides how many sets run and which loads are shed. Load flow, fault level, protection, motor starting and earthing studies define the equipment; a catalogue kVA chosen before them is usually wrong in a way that costs more to fix later.
What a multi-megawatt plant contains
A requirement above 2 MVA is met by a single very large unit only in unusual cases. The normal arrangement is three to eight sets of 1000 to 2500 kVA each, synchronised onto a common bus, with enough units that the loss of one leaves the critical load covered. Around those sets sit generator breakers, a synchronising bus, bus-couplers and incomers, step-up or step-down transformers, MV and LV switchgear, a protection scheme, an earthing system, bulk fuel infrastructure and a supervisory control system. The plant is described by its single-line diagram, its redundancy level and its control philosophy rather than by kVA. This is the territory of data centre campuses, steel and cement plants, refineries, large hospitals and mines in Telangana and Andhra Pradesh.
The architecture decisions that come first
- Generation voltage, almost always 11 kV at this scale because a 2.5 MVA set at 415 V produces about 3500 A and paralleling several such sets drives LV fault levels beyond what switchgear can withstand.
- Number and size of sets, set by the redundancy requirement, the load profile and the largest step the plant must accept.
- Standby, prime or continuous duty, which changes the engine rating, the fuel system and the maintenance regime.
- Utility interaction: islanded operation only, momentary paralleling for seamless transfer, or sustained parallel operation for peak shaving, each with different protection and DISCOM approval requirements.
- Load management: which loads are shed, in what order, and how fast, so that the loss of one set does not collapse the bus.
The master controls page describes the power management layer that holds these decisions together, and the generator to grid paralleling page covers the utility interface.
Engineering studies and what each decides
| Study | Question it answers | Equipment it fixes |
|---|---|---|
| Load flow | Do voltages and loadings stay within limits in every operating combination? | Transformer ratings, cable sizes, tap settings |
| Fault level | What is the maximum and minimum short-circuit current at each bus? | Switchgear withstand ratings, relay settings |
| Protection coordination | Does the nearest device clear each fault first, on every source? | Relay types, CT ratios, time-current settings |
| Motor starting | Can the largest motor start with the sets that will be online at that moment? | Starting method, minimum sets online, alternator size |
| Harmonic assessment | Will drives, UPS and rectifiers distort the bus beyond limits on generator supply? | Filters, alternator derating, transformer vector groups |
| Earthing | How is each neutral earthed and are touch and step voltages safe? | Neutral earthing resistors, earth grid, earth fault scheme |
These are the studies listed under electrical studies, and the plant’s single-line diagram is both their input and their output.
Fault current from paralleled generators
Isc ≈ Irated / X″d
Here Isc is the initial symmetrical short-circuit current one generator feeds into a close three-phase fault, Irated is its rated current and X″d is its sub-transient reactance in per unit on the machine base. A 2500 kVA set at 11 kV has a rated current of about 131 A; with X″d of 0.15 it contributes roughly 875 A initially, and four such sets in parallel contribute about 3.5 kA before the utility source is counted (illustrative). At 415 V the same four sets would contribute roughly 93 kA, which few LV switchboards can withstand, and that single arithmetic result is why multi-megawatt plants generate at MV. The contribution decays over a few cycles as the machine moves through its transient and synchronous reactances, so making capacity, breaking capacity and relay operation must each be checked at the right instant. The fault current page covers the decrement and the contribution of motors and inverters.
Controls, sequencing and load shedding
A power management system starts sets on demand, synchronises each to the bus, shares kW and kVAr between them, stops sets when load falls so that the rest run efficiently, and sheds pre-agreed loads in milliseconds if a set trips. The sequence in which sets start and the minimum number kept online are set by the largest motor start and the step-load capability of each engine, not by average load. Dead-bus logic must prevent two sets closing onto a dead bus together, and sync-check relays must back up the synchroniser. For plants with battery storage, a BESS can supply the first second of a block load or a motor start while engines respond, and can absorb excess generation so that fewer sets idle; the generator and BESS page sets out the screening.
Fuel, cooling and civil works at plant scale
A 5 MVA plant at full load consumes in the region of 900 to 1000 litres of diesel an hour, so a day of autonomy is a tank farm of 20000 litres or more with bunding, fire separation, tanker access, duty and standby transfer pumps, day tanks and a polishing system (illustrative). Heat rejection of several megawatts needs remote radiators or cooling towers designed for all sets at full load in peak summer. Exhaust stacks follow the CPCB height requirements for the plant capacity, and engines above 800 kW fall under the emission norms for large generators rather than CPCB IV+, to be confirmed per engine. For temporary multi-megawatt requirements, the same architecture is built from synchronised rental modules; see multi-megawatt rental power.
Procurement path
The sequence that works is requirement, then engineering, then equipment confirmation. The requirement is a load list with criticality, a redundancy level, an autonomy and a utility interaction mode. The engineering is the single-line diagram, the studies above and the control philosophy. Only then are generator ratings, alternator reactances, breaker ratings and transformer impedances fixed and sent to manufacturers for confirmation against current product data. Requesting a price for “2500 kVA sets” first locks in reactances and ratings that the studies may later show to be unsuitable. The 1 to 2 MVA page covers smaller plants where some of these questions simplify.
Common mistakes
- Paralleling several sets at 415 V because the existing switchboard is LV. The bus fault level exceeds the switchgear withstand and the whole board must be replaced.
- Sizing the plant on average load and keeping too few sets online at night. A single compressor start pulls the bus down and trips every set.
- Omitting fast load shedding. When one set trips, the remaining sets are overloaded within a second and the bus collapses.
- Using a decades-old DISCOM fault level letter for the utility contribution. Network reinforcement has since raised the fault level and the switchgear is under-rated.
- Designing the fuel farm after the electrical plant is fixed. The tanks end up far from the sets, pipe runs are long and the bund does not fit the site.
What to send us
Send the load list with criticality and the largest motors, the redundancy requirement, the intended utility interaction mode, the preferred generation voltage if already decided, the required autonomy, the site plan with available area and tanker access, and any existing single-line diagram or DISCOM fault level data. Submit it through the application engineering enquiry form and we will propose an architecture, a study scope and a sequence for confirming equipment with manufacturers.
Questions engineers ask
Why not buy one 3000 kVA generator instead of three 1000 kVA sets?
One unit means no redundancy, a very large block of capacity that runs lightly loaded for most hours, and a single point of failure for the whole site. Three sets allow N+1 coverage, efficient part-load operation and maintenance without losing backup, at the cost of synchronising controls.
Do I need DISCOM approval to parallel generators with the grid?
Yes. Sustained or even momentary paralleling with the utility requires the DISCOM’s consent, a protection scheme that meets its interconnection requirements, including anti-islanding and reverse power protection, and compliance with the applicable CEA regulations.
How fast must load shedding act in a multi-set plant?
Fast enough to prevent the remaining sets tripping on overload or underfrequency after one set is lost, which in practice means within a few hundred milliseconds. Breaker-status based shedding is faster than frequency-based shedding and is preferred at this scale.
Which studies are mandatory before energising a 2 MVA plus plant?
At a minimum a fault level study, a protection coordination study and an earthing design, because these decide whether the switchgear is safe and whether faults are cleared. Load flow, motor starting and harmonic studies follow where the load profile makes them relevant.
Generators by rating
- Under 15 kVA generators
- 15–50 kVA generators
- 50–125 kVA generators
- 125–250 kVA generators
- 250–500 kVA generators
- 500–1000 kVA generators
- 1–2 MVA generators
Technical content reviewed by Srinivas Reddy, Managing Partner, Reddy Generators. Last updated 19 October 2026.