In short. A data centre generator project is judged by whether the whole power chain survives a utility failure: generator start, transfer, UPS behaviour, cooling restart and fuel autonomy. A generator-only test does not prove that. A credible data centre project defines the supported load and topology, tests each operating mode under controlled load, and records the results so the facility can show what was verified and what was not.
How data centre generator projects differ
In most buildings a generator project ends when the set starts on mains failure and the lights come back. In a data centre the generator is one link in a chain that also includes the DISCOM supply, transformers, transfer switchgear, UPS modules and batteries, cooling plant and redundant A and B distribution. The IT load never sees the generator directly; it sees a UPS output that must remain stable while the generator starts, synchronises or transfers, and while the UPS rectifiers draw recharge current. A data centre project therefore needs a defined boundary from the start. Is the scope the generator alone, the generator and its transfer panel, or the full continuity sequence? The answer decides what can be tested and what can be claimed afterwards. Facility-level design is covered in data centre power and data centre microgrids.
Scoping the project
| Scope item | Question to settle | Why it matters |
|---|---|---|
| Supported load | IT load, mechanical load, recharge and growth, in kW and kVA | Sets generator rating and step-load capability |
| Topology | N, N+1 or 2N; single or synchronised sets | Decides how maintenance is done without losing redundancy |
| UPS compatibility | Rectifier type, input harmonics, walk-in time | Prevents voltage and frequency instability on generator supply |
| Transfer | Open or closed transition, timing, retransfer logic | Determines what the UPS must ride through |
| Fuel | Autonomy at actual load, refuelling access, fuel polishing | Runtime is usable fuel at real load, not tank volume |
| Testing | Load bank type, modes to be tested, acceptance criteria | Defines what the handover evidence will prove |
Generator and UPS coordination
Double-conversion UPS rectifiers present a non-linear load to the generator. With older six-pulse rectifiers, harmonic currents heat the alternator and distort the voltage the AVR senses, sometimes causing hunting or UPS bypass alarms. Modern active front-end rectifiers are far gentler, but they still need a soft walk-in so the full rectifier load is not applied as a single step. Sizing must also include the recharge of batteries that have just discharged through the transfer gap. A practical check is the ratio of generator kVA to UPS input kVA, together with the alternator’s harmonic capability from manufacturer data. Where the UPS supplier and the generator supplier are different, both should sign off the interface.
t_run = (V_usable × ρ_fuel) / (SFC × P_load)
Diesel generator calculator
Fuel use, runtime, tank size and cost
How this is calculated
Real power is kVA × power factor. Fuel use is estimated as L/h ≈ 0.04 × rated kW + 0.22 × load kW, which follows the usual shape of modern diesel gensets: about 0.26 L per kWh near full load, rising towards 0.38 L per kWh at a quarter load because the engine's fixed losses are spread over less output.
If you enter the full-load figure from your engine datasheet, the curve is scaled to match it. Runtime is usable fuel ÷ L/h. Tank size is L/h × hours, plus the reserve margin, divided by the usable fraction.
These are planning estimates, typically within 10 to 15 percent of a healthy engine. Actual use depends on the engine model, its condition, altitude, ambient temperature and how the load varies. For a figure you can rely on, use the manufacturer's fuel curve or a metered test.
Estimates only, worked out in your browser. For a checked selection, send us your load list and site details or call 098852 01608.
t_run is runtime in hours, V_usable the usable fuel volume in litres above the suction point, ρ_fuel the fuel density in kg per litre, SFC the specific fuel consumption in kg per kWh at that load from manufacturer data, and P_load the actual generator load in kW. Worked example (illustrative): 2,000 litres usable, density 0.84 kg/l, SFC 0.215 kg/kWh at 75 percent load and a 600 kW load give 1,680 / 129 ≈ 13 hours. The brochure figure for tank volume is not the answer; usable volume and real load are. See fuel runtime calculation.
Testing that proves the chain
A data centre acceptance test should exercise each operating mode the facility depends on, not only the generator. Live IT load is rarely available or safe for testing transients, so load banks are connected to represent it. Resistive banks prove kW capability and cooling; resistive and reactive banks at 0.8 power factor prove the alternator and AVR at rated kVA. The test record should note load steps, voltage dip and recovery time, frequency dip, temperatures, and the UPS behaviour seen at each step. Load bank sizing explains how to rate the banks for the test.
Test sequence we would expect to see recorded
- Utility failure simulation with generator start, time to rated voltage and transfer time.
- Block load acceptance in the steps the UPS and cooling will actually apply.
- Extended run at the design load with temperature and fuel consumption recorded.
- Failure injection, such as one generator failing to start in an N+1 plant.
- Retransfer to utility with the cooldown run and return to standby.
If the existing generator must be out of service during upgrade work, data centre rental power can hold the critical bus temporarily, which itself needs a defined test before the permanent set is disconnected.
Confidentiality and what we publish
Data centre operators rarely allow site names, layouts or capacity figures to be published, and we respect that. No approved data centre case is presented here at present. When a case is published it will describe load, topology and test conditions in enough detail to be useful while omitting what the operator classifies as sensitive. Relevant experience can be discussed in confidence once we understand your requirement.
Common mistakes
- Accepting a generator load test as proof of facility continuity. The UPS, transfer and cooling interfaces remain untested until a real outage tests them.
- Sizing the generator on IT load and forgetting UPS recharge and chiller restart. The set is overloaded in the minutes after transfer when the facility is most exposed.
- Quoting runtime from tank volume. Unusable fuel below the suction and higher consumption at real load cut the actual runtime short.
- Reusing the existing generator cables after an upgrade without a cable check. The new rating overheats the old cable or fails voltage drop limits.
What to send us
Send the single-line diagram, UPS make and rating with rectifier type, IT and mechanical load figures, existing generator ratings and controller details, fuel tank arrangement, and the redundancy level you need. Recent event logs from a utility outage help greatly. Send your application details. Other project categories are listed under verified projects.
Questions engineers ask
Should a data centre generator be tested on live load or a load bank?
Both have a place. Load banks allow controlled steps and full rated load without risking the IT load. Live-load transfer tests prove the real sequence but must be planned carefully, with the UPS healthy and a rollback plan agreed.
What generator size is needed for a given UPS rating?
It depends on rectifier type, recharge current, mechanical load on the same bus and the alternator’s harmonic capability. Older six-pulse UPS often needed generous oversizing; modern active front-end units need much less. Confirm with both suppliers’ data.
How much fuel autonomy should a data centre hold?
It depends on the operator’s continuity target and refuelling arrangements. Many facilities in India specify between 12 and 48 hours at design load, with a refuelling contract for longer outages. The figure should be stated in the design basis.
How often should the integrated sequence be tested?
At acceptance, after any significant change, and on a planned schedule, often annually. Monthly generator runs verify the engine, not the full transfer and UPS interaction.
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
- 2 MVA and above generators
Technical content reviewed by Srinivas Reddy, Managing Partner, Reddy Generators. Last updated 19 October 2026.