384V 100Ah LiFePO4 UPS Battery for a Data Center in South Africa

Battery ESSUPSData Center

The Challenge

The Solution

VRLA Baseline vs LiFePO4

ItemTypical VRLA bankLiFePO4 as supplied
ChemistryVRLA lead-acidLiFePO4
Nominal voltage384V DC384V DC (120S LiFePO4)
Capacity100 Ah100 Ah (38.4 kWh per string)
Deep-cycle enduranceDegrades with repeated deep dischargeSeveral times the deep-cycle throughput of VRLA
Recharge behaviourLong absorption tailHigh charge acceptance, ready again sooner
Communication with UPS hostTerminal voltage sensing onlyModbus over RS485, CAN optional

Results

  1. Commissioned in April 2024 with a total installed capacity of 60 kWh.
  2. The bank is sized to the operator's required runtime. Ride-through is set by site load — and because the UPS now receives real state-of-charge data, that runtime figure is calculated rather than estimated.
  3. Charge acceptance is what matters between events: the pack returns to readiness during short grid windows instead of trailing a long absorption phase.

Related Product

Case Study FAQ

Why does frequent grid outage favour LiFePO4 over VRLA?

Deep cycling is what wears a battery out. VRLA tolerates a limited number of deep discharges before capacity drops off; LiFePO4 delivers several times that cycle count, so a site that discharges often wears through a VRLA bank far faster.

Does the UPS need to be replaced as well?

Not normally. If the nominal voltage band, charge behaviour and communication protocol are matched, the LiFePO4 string drops into the existing UPS. We configure the BMS to the host's protocol rather than asking the customer to change equipment.

How fast does the battery recharge between outages?

Much faster than VRLA. LiFePO4 accepts high charge current through most of the charge curve, so the pack returns to full readiness during short grid availability windows instead of waiting out a long absorption phase.

What data does the UPS receive from the battery?

State of charge, state of health, individual cell voltages, temperatures and alarm states — enough for accurate runtime prediction and for the monitoring platform to raise faults before they become outages.

More Reading

Need a Custom Power Solution?

37 senior R&D engineers. OEM/ODM from 500 units. Samples in 3-7 days.

Get a Quote WhatsApp Us