In short. A data centre already is a critical-power system: utility feeds, UPS, batteries, generators, transfer switchgear and redundant distribution. A microgrid layer, whether BESS, solar or utility-interactive generation, must be fitted inside that architecture and tested against it, never bolted on beside it. The rule is simple: ride-through stays with the UPS, long-duration supply stays with the generators, and any new asset must prove it does not add a shared failure mode.
Start from the architecture that already exists
Most colocation and enterprise facilities in Hyderabad already run an N+1 or 2N arrangement: two DISCOM feeds or a feed and a dedicated transformer, double-conversion UPS with ten to fifteen minutes of battery, diesel generators on automatic transfer, and A and B distribution paths to every rack. Each of those elements carries a defined responsibility and has been accepted through testing. A microgrid proposal should begin by writing those responsibilities down, because the usual failure of a data centre microgrid project is not a bad battery or a weak inverter but an undefined boundary between the old continuity philosophy and the new energy-optimisation philosophy. Which controller decides to open the utility breaker? Who owns the generator start command once a BESS can also carry the load? What happens when the energy management system and the UPS disagree about the state of the bus? Those questions are answered in design or they are answered during an outage.
Which layer owns which timescale
| Time after utility failure | What must happen | Layer that owns it | What a microgrid asset may add |
|---|---|---|---|
| 0 to 20 ms | IT load sees no disturbance | UPS inverter and battery | Nothing; BESS does not replace UPS ride-through |
| Up to 15 seconds | Generators start and reach voltage | Generator controllers and AMF or ATS | Grid-forming BESS can hold the bus and soften block loading |
| 15 seconds to 5 minutes | Transfer, UPS recharge begins, cooling restarts | Transfer switchgear, chiller sequencing | BESS can supply recharge and chiller inrush so generators see a flatter step |
| 5 minutes to hours | Sustained supply, fuel management | Generators, fuel system | Solar reduces fuel burn in daylight if the controller allows it on an islanded bus |
| Normal operation | Demand charge and tariff management | Energy management system | Peak shaving and tariff arbitrage, inside strict state-of-charge limits |
The table makes one point visible: a BESS can perform several roles, but a kWh used for peak shaving at 18:00 is a kWh not available for ride-through support at 18:05. A data centre should reserve a defined state-of-charge floor for continuity functions and allow economic dispatch only above it. DG, UPS and BESS hybrid critical power describes the architecture options in more depth.
Cooling decides whether the servers survive
Servers on UPS stay powered through the transfer gap, but chillers, CRAC units and pumps are usually on the generator bus and stop for the duration. In a dense hall the supply air temperature climbs fast; depending on density and thermal mass, the facility may have only a few minutes before IT equipment throttles or trips on temperature, which means the cooling restart sequence is as much a continuity function as the UPS. A microgrid layer helps here if it shortens the cooling interruption or softens chiller inrush, and harms if its control logic delays generator start while it decides whether the battery can carry the load. Any proposal should state the cooling recovery time before and after the change, with the chiller compressor inrush and the sequencing of CRAC fans included in the generator load step calculation. See data centre power for how cooling and IT loads are classified.
Protection, earthing and common-mode risk
Inverter-based sources behave differently from rotating machines under fault. A diesel alternator delivers several times rated current for the first cycles, which is what lets downstream breakers clear a fault selectively. A BESS inverter typically limits fault current to a little over its rated current, so a bus held by the battery alone may not clear a downstream fault quickly enough, and protection settings proven with the generators may not hold. The earthing arrangement also changes when a transformer-less inverter becomes the only source. Both issues are solvable, but they need a protection study for each operating mode, not a single setting file. The second risk is common-mode failure: a new microgrid controller that sits in series with every source becomes a single point of failure for all of them unless it is redundant and fails to a safe, generator-led mode. Inverter and BESS fault current behaviour and microgrid controller architecture cover both points.
Testing the integrated sequence
Component tests are not enough. The acceptance test for a data centre microgrid should remove the utility, record the UPS bus, time the generator start and transfer, observe the cooling restart, then introduce a generator failure and confirm that the BESS behaves as designed and that the remaining sets are not overloaded. Load banks are needed for this because live IT load cannot be used to prove a transient, and resistive-only banks will not show the reactive behaviour of chillers and UPS rectifiers. Retransfer to utility and the return of the battery to its continuity floor should be part of the same test, with event logs from the UPS, generator controllers and microgrid controller time-synchronised so the sequence can be reconstructed. Sizing the test is covered in load bank sizing, and integrated systems testing sets out the full method. Where the facility cannot free a generator for the test, data centre rental power can hold the bus during the work.
Common mistakes
- Letting the energy management system discharge the battery for tariff savings with no protected floor. The one evening the utility fails at a low state of charge, the microgrid contributes nothing to continuity.
- Treating the BESS as a UPS replacement because both contain batteries. The transfer to battery is not seamless at the IT load and the facility loses the ride-through it was built around.
- Reusing generator protection settings on an inverter-held bus. A downstream fault is not cleared quickly and the whole hall drops instead of one feeder.
- Adding a central microgrid controller without a defined fail-safe mode. A controller fault blocks generator start and the new layer causes the outage it was meant to prevent.
- Accepting the project on individual component tests. The first interaction fault between UPS, BESS and generators is discovered during a real DISCOM outage.
What to send us
Send the single-line diagram with the UPS, generator and transfer arrangement, the IT and mechanical load split with chiller and CRAC starting data, the UPS battery autonomy, the generator ratings and controller models, and the objective of the microgrid layer, whether that is fuel reduction, demand management or longer autonomy. Recent event logs from an actual utility failure are the most useful document you can provide. Send your application details.
Questions engineers ask
Can a BESS replace the diesel generators in a data centre?
Not for long-duration outages at current battery economics. A battery sized for several hours at full hall load is very large, and it cannot be refuelled during the outage. The practical role of a BESS is ride-through support, generator load smoothing, spinning reserve and tariff management alongside generators, not instead of them.
Can solar on the roof supply the data centre during an outage?
Only if the inverters are allowed to run on an islanded bus formed by the generators or a grid-forming BESS, and only up to a share of the online generator load that keeps the sets above minimum load. Standard grid-tied inverters disconnect when the utility fails and contribute nothing during the outage.
Does a microgrid change the generator sizing?
It can reduce the required load step capability if a BESS absorbs chiller inrush and UPS recharge, but it rarely reduces the total installed generator capacity, because the facility must still ride out a long outage with one generator out of service. Size the generators for the worst case without the battery, then let the battery improve the transient.
Who should own the microgrid controller logic?
The party responsible for the facility’s critical-power philosophy, usually the facility electrical engineer with the generator and UPS suppliers contributing. The vendor of the BESS should not be the only party who understands how the sources interact.
How often should the integrated sequence be retested?
At acceptance, after any change to controller firmware or settings, and on a planned cycle agreed with operations, commonly annually. Monthly generator tests prove the generators; they do not prove the interaction with the battery and UPS.
Technical content reviewed by Srinivas Reddy, Managing Partner, Reddy Generators. Last updated 19 October 2026.