Application
Diesel Generators for Data Centers
A data centre generator is not sized from the IT load alone. It is sized from the load steps that the UPS, the chillers and the pumps present after a utility failure, from the redundancy the hall was designed to, and from how long the site must run before fuel arrives. In a colocation hall the generator is part of the topology that the customer is buying, so the ratings and the number of sets are fixed by the design, not by the equipment budget.
Standby generator sets for data centres are normally quoted at 500 kVA and above, arranged in N+1 or 2N groups, with paralleling controls, automatic transfer at each level of the distribution and on-site fuel for the autonomy the site declares to its clients.
Requirements
What a data centre specification has to state
The rows below are the items that change the machine. Two halls of the same IT capacity can need very different generators, depending on the answers.
| Item | Typical requirement | Why it changes the set |
|---|---|---|
| Redundancy | N+1 as a minimum for concurrently maintainable operation; 2N for fault-tolerant halls | Fixes how many sets, and whether paralleling control is required |
| Duty rating | Standby (ISO 8528-1 ESP, or DCP where the duty is declared as data centre power) | ESP and PRP are different numbers from the same hardware; in regions with a weak grid the duty becomes closer to prime power |
| Load acceptance | The declared step-load curve for the set, against the largest step the hall can present | Chillers and large UPS rectifiers are the steps that decide the rating |
| Transfer time | UPS carries the IT load; the generator must be at voltage and frequency before the UPS battery is exhausted, and before mechanical cooling restarts | Sets the allowable start sequence, and whether a black-start sequence is needed |
| Autonomy | Commonly 12-24 h of on-site fuel at full load; longer where fuel delivery is unreliable | Drives bulk storage, day tanks, transfer pumps and bunding |
| Ambient | 45-50 °C design ambient with the radiator sized for it | A set quoted at 40 °C is derated above it, and a roof-level radiator sees the highest temperature on site |
| Noise | Typically 65-75 dB(A) at the nearest receptor, not at the set | Decides the enclosure and whether a wall or a roof screen is needed |
| Harmonics | The alternator sized for non-linear loads from UPS and drive equipment, with a stated sub-transient reactance | Harmonic current heats the alternator beyond its kW rating |
| Paralleling and control | Load sharing, reverse power protection, synchronising, and reporting to the building management system | Determines the control panel scope and the interface to the hall monitoring |
| Testing | Witnessed load bank test at commissioning, and periodic testing afterwards | Rated load data has to be demonstrated, not assumed |
Configurations
Configurations we quote most often
Each rating below links to the power range page, which carries the specification envelope, the fuel figures and the options for that class.
| Rating per set | Typical arrangement | Fits | Range page |
|---|---|---|---|
| 500 kVA / 400 kW | Two sets, N+1, one bus | Small hall or an edge site with a single mechanical plant | 500 kVA |
| 800 kVA / 640 kW | Two or three sets, N+1 | Mid-size hall, split UPS and mechanical buses | 800 kVA |
| 1000 kVA / 800 kW | Two groups of two, 2N | Fault-tolerant hall where each path must stand alone | 1000 kVA |
| 2000 kVA / 1600 kW | Multiple sets in parallel, N+1 or 2N | Large campus or a hall with central chilled water plant | 2000 kVA |
Reference configuration
A worked example
This is an illustrative configuration, not a delivered project. It shows how the requirement table above turns into a specification. Replace the figures with your own load list before pricing.
| Parameter | Example value |
|---|---|
| Critical IT load | 800 kW, delivered through two UPS systems |
| Mechanical load | Two chillers at 300 kW each plus pumps and fans |
| Redundancy | 2N on the generator bus, so two independent sets per path |
| Rating selected | Four sets at 1000 kVA / 800 kW, standby rating, paralleled within each path |
| Largest load step | One chiller started with soft start, about 300 kW running, plus UPS recharge |
| Autonomy | 24 h at full load on bulk storage with a day tank and transfer pump |
| Enclosure | Containerised sets at roof level, or open sets in a ventilated generator hall |
| Control | Paralleling panels, load sharing, reverse power protection, BMS reporting |
| Testing | Load bank test at commissioning with the mechanical plant running, then periodic testing |
Questions
Questions we are asked about data centre sets
What is the difference between N+1 and 2N for a generator plant?
N+1 means one set more than the load needs, so a single set can be taken out for service without interrupting the hall. 2N means two complete independent supplies, each able to carry the whole load on its own, which is what fault tolerance requires. 2N needs twice the sets of N, doubles the switchgear and needs two separate fuel or cabling paths to mean anything.
How fast does a data centre generator have to start?
The generator does not have to beat the IT load, because the UPS battery does that. What it has to beat is the UPS battery autonomy and the cooling plant restart window. Common design targets are the generator at voltage and frequency within about 10-15 seconds, ready to accept the mechanical plant load shortly afterwards. Where the whole load is on battery, the actual constraint is how long the batteries last, which is usually measured in minutes.
How do you size a generator for a chilled water plant?
Start from the largest single step, not the total. A chiller with a soft starter or a variable speed drive presents a much smaller starting step than one started direct on line, and starting the machines in sequence rather than together often lets a smaller set carry the same plant. The declared load acceptance curve for the set, at the site ambient, is the document that proves whether the step is acceptable.
Should data centre sets be containerised or installed in a hall?
Both are common. Containerised sets at roof or yard level save building volume and are easy to replace, but they need clear air paths and they count against the site’s roof loading. A ventilated generator hall gives better access for major maintenance and better acoustic control, at the cost of building area. The mechanical and electrical design of the site usually decides it.
How much fuel autonomy is normal?
Twelve to twenty-four hours at full load is the common range, and 24 h is the figure most often written into client commitments. The volume that implies is large: at 0.24-0.27 litres per kWh delivered, 400 kW of load needs roughly 2 300-2 600 L for 24 h, per running set. That is why the bulk tank, the bunding and the transfer pump are part of the generator scope rather than a site afterthought.
Related
Related applications and ranges
-
Diesel generators by application
How the requirement changes between data centres, hospitals, hotels, sites and oil and gas.
-
Hospitals
Essential supplies, imaging loads and long autonomy.
-
Containerized sets
The usual form for a roof-level or yard installed data centre plant.
-
Open frame sets
Generator halls, ducted radiators and load step data.
-
1000 kVA class
800 kW per set, the common building block of a 2N hall.
-
Request a quote
Send the load steps, the redundancy target and the autonomy, and we size the plant.
Specifying standby plant for a data hall?
Send the critical load, the largest step, the redundancy target and the fuel autonomy. Those four answers fix the number of sets and the rating of each one.