A data centre needs a reserve that can be maintained without losing its agreed availability. Battery capacity alone cannot deliver this. Supply paths, switching, module redundancy and planned-service behaviour are equally important.
Start with the load architecture
Identify dual-fed racks, the supply to network equipment and the cooling backup arrangement. N+1 or 2N must refer to a defined system boundary, rather than being used as a general label for the entire site. Request a description of the individual failures through which the critical load remains supplied. Include shared auxiliaries that could defeat apparently independent paths.
A battery does not replace a topology
A BESS can feed a UPS group and extend autonomy. Changing the battery technology within the UPS requires a compatible configuration permitted by its manufacturer. Matching voltage alone does not make arbitrary LiFePO4 modules a replacement for VRLA: charging algorithms, BMS communications and protections also need to fit.
Ask for the service scenario
What happens during battery-string replacement, maintenance bypass operation or cooling failure? Which loads lose redundancy? The supply documentation should answer these questions. Before modelling economic dispatch, establish the emergency reserve that must remain untouched. Only the remaining usable capacity can be considered for other duties.
Trace the supply path to each rack group
Begin purchasing with a diagram showing distribution equipment, cable paths and load inputs as well as UPS units and batteries. Two server power supplies do not create two independent paths if both depend on the same upstream component. Mark those common points and identify which event could remove power from a rack. This helps assess the value of additional equipment before paying for it.
Single-input devices need separate attention. They can change switching requirements and acceptable transition behaviour. Do not automatically transfer the behaviour of dual-supply servers to networking devices, storage systems or auxiliaries. The specification should identify the loads supported by each path and the assumptions made about the receiving equipment's internal arrangement.
Give N+1 an explicit boundary
If three power modules are needed for the load, a fourth may provide one-module redundancy in the agreed duty. That does not establish redundancy of a common controller, busbar, bypass or battery system. Request a list of single failures and the system response to each. N+1 is useful only when the supplier states which components it covers and how much usable capacity remains.
Consider a hypothetical 180 kW load and four 60 kW modules. Removing one module leaves an arithmetic total of 180 kW: exactly the load, with no growth allowance. Next, verify rating conditions, battery-path availability and whether the particular system permits the duty. If demand is expected to reach 210 kW, the same arrangement no longer supports that demand after one module is removed.
Treat maintenance as its own operating scenario
Ask what happens during power-module replacement, battery-string work, a common-input repair and operation through a maintenance bypass. The maintenance state may offer less protection than normal operation. It matters not only whether work is possible, but which additional failures the system can tolerate at that time. A statement about maintenance without interruption needs a defined boundary and conditions.
A useful comparison matrix lists the component withdrawn, remaining power, remaining runtime, unprotected loads and required operator actions. It reveals differences between packages with equal nominal ratings. In some arrangements, removing a common dependency may offer more benefit than adding another module. That conclusion requires the actual diagram, not a count of cabinets.
Keep planned maintenance and unplanned failure separate in the discussion. A controlled procedure can prepare loads and confirm switching conditions; an unexpected fault does not provide that opportunity. A proposal should not use success under a planned procedure as blanket proof of fault tolerance. Request the evidence relevant to each claim and retain it with the accepted configuration.
A BESS upstream of UPS units supplies their input demand
When storage extends the runtime of a UPS group, it supplies more than useful IT output. UPS losses, battery recharge after a transition and other loads on the same backup circuit must be included. For illustration, 100 kW output at 95% UPS efficiency requires about 105.3 kW input for the load alone. Charging demand is additional; this example does not determine its magnitude.
Include cooling from its actual electrical arrangement and operating duty. Multiplying server power by annual-average PUE does not establish precise emergency demand. The relationship may change during transitions or partial operation. Use measurements and data for the systems that remain active: pumps, fans, controls, networking and other dependencies.
Keep the energy reserve genuinely available
For a BESS serving several purposes, define protected reserve and EMS priorities. If it reduces peaks, show the reserve remaining at the end of that duty. A financial model cannot use the full capacity for savings while simultaneously promising it all for an outage. Compare cases at minimum permitted SOC, not only at the start of a day after full charging.
Battery resource must also be connected to the required function. UPS batteries need runtime evidence at the specified load, temperature and resource condition. BESS evaluation needs usable energy and power at an agreed boundary. These are different performance measures. Comparing kWh alone does not establish that two packages provide the same continuity of service.
Questions that make proposals comparable
| Boundary | Question |
|---|---|
| Rack | Do its inputs actually use independent agreed supply paths? |
| UPS | What power remains with one module unavailable? |
| Battery | Is required runtime retained while one string is serviced? |
| BESS | What reserve remains after an economic operating cycle? |
| Cooling | Which auxiliary systems are needed for the agreed IT load? |
| Monitoring | Are logs and warnings available without external communication? |
IEC 62040-3 addresses UPS performance and testing requirements. A reference to that standard on a datasheet does not establish the availability of the complete data-centre infrastructure. The actual package needs manufacturer documentation, agreed scenarios and verification results. Do not apply availability tiers or certification labels without evidence for the particular facility and claim.
Purchase with the next expansion stage in view
State present demand, agreed growth and when that growth is expected. Ask what can be added without replacing the main system and where busbars, switching, batteries or licences need expansion. Space for another power module does not prove that every associated component has adequate capacity. Obtain a view of the next-stage cost before the first delivery.
Include spare-parts and support questions in that comparison. Establish which parts are replaceable in the supplied configuration, how faults are diagnosed and what information is required for a support request. A promised response time is different from a confirmed replacement-part delivery time. The commercial terms should make that distinction visible without assuming that a local stock holding exists.
Send USK Energy the supply-path diagram, rack loads, runtime requirements and maintenance matrix. These inputs support comparison of online UPS units, battery cabinets, BESS and associated components. The final specification should identify exact configurations, redundancy boundaries, documents and support terms. This ties the purchase to continuity for a particular load group instead of a general promise of reliability.
Two checks before selecting a UPS
Change the inputs and save the result to discuss the equipment package.
S = P / PF; E = P × minutes / 60. Compare both kW and kVA with the UPS ratings. The calculated kWh is load energy, not the required VRLA battery capacity. Select batteries using manufacturer discharge tables, including end voltage, temperature, losses and ageing.
Before requesting a quotation
Mark questions for which you already have a confirmed answer. Save the list and send unresolved points to the supplier.
What must remain powered during an outage?
Start with the loads that cannot stop. A server, a conveyor controller and the conveyor motor have different requirements: continuous power, retained process state and starting current. A single site power figure cannot describe all three. Separate loads into those requiring uninterrupted supply, those tolerating a brief interruption and those that can be shed during an outage.
A UPS bridges the transition for critical loads; a BESS provides stored energy; a generator can extend backup during a long outage. Check the UPS input requirements against the backup source output. Include battery recharge power after supply returns: restoring backup capacity must not overload the source.
Further verification
From calculation to equipment
Preparing an equipment purchase? Send the load, backup duration or specification. USK Energy will clarify the configuration, compatibility and delivery terms.




