Direct answer: Data center battery storage should be designed as part of a coordinated critical-power system, not as a larger substitute for a UPS. The UPS protects the IT load through short disturbances and transfer events. A BESS can add peak shaving, energy shifting, and longer backup support, but only when the electrical topology, reserve policy, transfer sequence, protection settings, and failure modes are engineered together.
For data center owners, EPCs, and facility engineers, the important question is not simply “How many kilowatt-hours do we need?” It is: Which loads must never see an interruption, which loads may be shed, how long must each load be supported, and what happens at every handoff between utility, UPS, BESS, and generators?
UPS, BESS, and backup generation solve different problems
A UPS is primarily a power-continuity and power-quality device. The U.S. ENERGY STAR program describes data center UPS batteries as commonly supporting loads for seconds to tens of minutes—long enough for generators to start or for equipment to shut down correctly—while also correcting sags, surges, and frequency distortion. The international performance and test framework for complete UPS systems is covered by IEC 62040-3:2021.
| Function | Primary design objective | Energy decision | Control priority |
|---|---|---|---|
| UPS ride-through | Maintain acceptable power to critical IT loads through disturbances and source transfers | Required ride-through duration at the protected load, including conversion losses and aging margin | Continuity and power quality |
| BESS peak shaving | Limit facility demand during tariff or grid peaks | Discharge power multiplied by the expected peak window, plus operating constraints | Cost reduction without consuming protected reserve |
| Extended backup | Support selected critical loads during a longer outage | Critical-load energy over the target outage, coordinated with generation and load shedding | Resilience and recovery |
| Generator bridge | Hold critical loads until standby generation is stable and accepted | Worst-case start, synchronization, transfer, and retry sequence | Successful handoff under fault conditions |
A single battery plant may support more than one function, but those functions compete for state of charge, inverter capacity, thermal headroom, and maintenance availability. A system dispatched aggressively for demand savings may not have enough reserve when an outage begins. The operating policy must therefore protect resilience before optimizing cost.
Three practical architecture patterns
1. Dedicated UPS plus a separate facility BESS
The UPS remains on the critical IT path, while the BESS connects at the facility or microgrid level. This creates a clear separation of responsibilities: the UPS handles no-break continuity, and the BESS manages demand, energy, and longer-duration support. It is often the easiest arrangement to explain, test, maintain, and operate because a peak-shaving control error does not directly redefine the UPS reserve.
2. Coordinated UPS and BESS with a shared reserve strategy
Here the EMS coordinates an external BESS with UPS status, generator availability, tariff signals, and the critical-load forecast. The design can use more battery value during normal operation, but only if the reserve floor, communication-loss behavior, maintenance mode, and manual override are defined. The UPS must still meet its continuity requirement independently of optimistic dispatch assumptions.
3. BESS-centered microgrid for critical and noncritical loads
A microgrid can coordinate BESS, utility supply, standby generation, renewable generation, and controllable loads. This is useful when the project needs both resilience and grid interaction. It also increases commissioning scope: islanding, black-start sequence, grid-forming or grid-following behavior, protection coordination, load shedding, generator minimum loading, and reconnection all require validation. NREL’s data center work emphasizes system integration, resilience planning, and real-world validation rather than treating any one technology as a complete answer.
Size power, energy, and reserve separately
Do not begin with a catalog capacity. Begin with measured load data and operating scenarios. The following equations are scoping tools; detailed design still requires electrical studies and vendor performance curves.
- Protected power: identify the highest coincident kW or kVA on each critical path, including expected growth and step-load behavior.
- Ride-through energy: protected kW × required minutes ÷ 60, then adjust for usable state-of-charge window, conversion efficiency, temperature, aging, and the required contingency margin.
- Peak-shaving power: baseline site demand minus the target grid-import limit. Confirm whether the tariff is based on 15-minute, 30-minute, or another demand interval.
- Peak-shaving energy: required discharge kW × expected peak duration. Repeat across the actual interval load profile rather than using one monthly maximum.
- Extended-backup energy: critical-load kW × outage duration, reduced only by generation that is demonstrably available in that scenario.
- Protected reserve: the minimum state of charge that remains unavailable to economic dispatch. Define how the EMS raises this floor before storms, grid events, maintenance, or generator outages.
Our commercial battery storage sizing guide explains the difference between power and energy in more detail. For a live data center project, the model should also include PUE, cooling continuity, UPS efficiency by load factor, auxiliary consumption, and the load that can be shed without violating service requirements.
Write the operating sequence before selecting equipment
A credible specification describes what the system does in each state. At minimum, define the following sequence:
- Normal grid operation: UPS online mode, BESS charging window, reserve floor, and generator readiness.
- Demand peak: BESS discharge limit, target meter, ramp logic, and conditions that suspend economic dispatch.
- Power-quality disturbance: UPS response, BESS behavior, and protection settings that prevent conflicting actions.
- Utility outage: event detection, islanding decision, protected load list, and immediate load shedding.
- Generator start and acceptance: start command, voltage/frequency qualification, transfer, warm-up, minimum loading, and retry logic.
- Extended outage: battery dispatch, generator optimization, fuel constraints, renewable input, and staged load curtailment.
- Grid return: stability timer, synchronization, reconnection, battery recharge rate, and avoidance of a new demand spike.
- Communication loss: local fallback setpoints, fail-safe reserve, alarm routing, and manual control.
If the same BESS also performs peak shaving, the EMS control logic should preserve a backup reserve and avoid charging at the wrong time. The related backup-reserve Q&A lists practical reserve controls.
Reliability questions procurement teams should ask
| Buyer question | Evidence to request | Red flag |
|---|---|---|
| Which loads receive no-break power? | Single-line diagram, A/B path definition, UPS topology, transfer-time requirement, and protected-load schedule | “The battery backs up the whole site” without a load hierarchy |
| Can peak shaving reduce backup readiness? | EMS reserve logic, minimum SOC, forecast method, communication-loss fallback, and operator override | No protected reserve or reserve defined only as an informal operating preference |
| What happens if one subsystem is unavailable? | Failure-mode matrix covering UPS module, PCS, battery rack, EMS, switchgear, generator, and communications | Availability claim without maintenance or degraded-mode behavior |
| How is performance accepted? | Factory and site acceptance procedures, step-load tests, outage sequence, transfer records, alarms, and recovery tests | Commissioning limited to energization and basic communication checks |
| Can the supplied system be audited later? | Cell/module/rack genealogy, firmware versions, test records, alarm history, and data export method | Serial numbers without traceable production and test evidence |
For the final item, use our battery passport and cell traceability checklist to define the record package expected from the supplier.
Safety and compliance are architecture inputs
Applicable codes depend on the country, jurisdiction, building, chemistry, installation location, and system configuration. For North American stationary ESS projects, NFPA 855 addresses installation hazards. UL 9540 covers the safety of complete energy storage systems and equipment, while UL 9540A is a test method used to evaluate thermal-runaway fire propagation behavior. A test report is not interchangeable with a site approval, and a component certificate does not automatically certify the assembled system.
Early design should therefore reserve space for required separation, ventilation, detection, emergency response access, electrical protection, and maintenance. It should also identify who will coordinate with the authority having jurisdiction, insurer, fire consultant, utility, and data center operations team. The choice between an indoor rack battery system and an outdoor cabinet or container is a site decision, not merely a capacity decision.
Common design mistakes
- Using monthly utility bills instead of interval data. The bill shows the outcome, not the shape or duration of the peak.
- Assuming nameplate kWh equals usable backup energy. Reserve windows, conversion losses, degradation, temperature, auxiliary loads, and end-of-discharge limits all matter.
- Counting a BESS as a UPS without validating continuity. Inverter start, mode transition, protection, and transfer performance must match the protected load.
- Ignoring cooling and support loads. Servers are not the only loads required to keep the data hall operating safely.
- Optimizing savings before defining failure behavior. A dispatch model is incomplete until it covers communication loss, maintenance, and component outages.
- Commissioning only the normal sequence. The project must test rejected transfers, generator failure, low SOC, alarms, and controlled recovery.
Data center battery storage RFQ checklist
- 15-minute or finer interval load data, plus identified critical-load feeders
- Existing and future IT load, PUE assumptions, cooling load, and growth phases
- One-line diagrams, UPS topology, A/B distribution, and generator configuration
- Required ride-through time, target outage duration, and acceptable load-shedding stages
- Utility tariff, demand interval, interconnection limits, and export restrictions
- Required redundancy level and allowed maintenance states
- Battery location, environmental conditions, access, fire strategy, and AHJ requirements
- EMS/BMS/UPS/generator communication protocols and cybersecurity ownership
- Factory acceptance, site acceptance, integrated systems testing, and handover documents
- Warranty boundaries, service response, spare parts, remote diagnostics, and data ownership
FAQ
Can a data center BESS replace the UPS?
Usually not by assumption. A BESS may be part of a no-break architecture, but the complete system must meet the load’s continuity, power-quality, transfer, protection, redundancy, and test requirements. Keep a dedicated UPS function unless the alternative architecture has been specifically engineered and validated.
Can one BESS provide peak shaving and backup?
Yes, if the EMS protects a minimum reserve and the PCS, battery, and thermal design can support both duties. The economic model must account for energy held in reserve, degradation, charging constraints, and periods when resilience takes priority over savings.
How much backup time should a data center battery provide?
It depends on the outage scenario and the rest of the power system. Define the critical load, generator start and retry sequence, fuel strategy, allowable load shedding, grid-event history, and recovery target. Then calculate energy for each scenario rather than choosing a generic duration.
What should we send a BESS supplier for a first proposal?
Send interval load data, one-line diagrams, critical-load list, UPS and generator details, tariff, site conditions, target applications, backup requirement, project location, schedule, and expected expansion. A proposal based only on total site MW and desired hours is not enough for a dependable design.
Prepare a data-center-specific design brief
XenPai supports commercial energy storage scoping from load data through battery configuration, controls, documentation, and commissioning requirements. Review our commercial energy storage solutions and commercial BESS buyer’s checklist.
For an initial review, send: the facility one-line diagram, 15-minute interval load data, critical-load schedule, UPS/generator topology, required backup duration, site location, tariff, available footprint, and project stage. Request a project assessment.
Technical references
- IEC 62040-3:2021 — UPS performance and test requirements
- ENERGY STAR — Reduce Energy Loss from Uninterruptible Power Supply Systems
- NREL — Smart Infrastructure: The Foundation of High-Performance Data Centers
- NFPA 855 — Standard for the Installation of Stationary Energy Storage Systems
- UL Solutions — Energy Storage System Testing and Certification