Three-Year-Old 1 MW/2 MWh Site: Rack Imbalance and Rising Temperatures
A 1 MW / 2 MWh LFP system (ten 100 kW / 200 kWh liquid-cooled cabinets, 20 racks of 16 cells each) has operated for three years at a data-center campus, cycling once daily between 15% and 95% SoC. The annual capacity test shows 87% of beginning-of-life. Cabinet 7, rack 14 has a cell voltage spread of 180 mV at top of charge (35 mV at commissioning), runs 6 °C above the cabinet average, and contains one cell whose self-discharge is three times the rack average. Loop supply pressure fell from 22 psi to 14 psi while the pump VFD now runs at 92% speed, and container ambient is 32 °C. The BMS has raised no alarm. The warranty requires an annual capacity test and written notice within 30 days for any rack exceeding 100 mV spread, and guarantees 70% capacity at year ten or 3,650 equivalent cycles.
1. Triage the findings: what is the most likely root cause, what is the more serious finding, and what immediate actions and notifications are required?
Read the worked answer
Immediate notification. The 180 mV spread exceeds the 100 mV warranty threshold, so written notice is due within 30 days regardless of root cause. Do not delay it while diagnosis continues.
Likely root cause of the heat. Low loop pressure combined with the pump running at 92% speed points to a flow restriction on the branch serving rack 14 - a kinked hose, a partly closed quick-disconnect, a clogged strainer, or an air lock. Reduced coolant flow raises cell temperature, accelerates aging, and widens cell spread, which is self-reinforcing.
More serious finding. The cell with 3× self-discharge suggests a soft internal short, lithium plating, or contamination. That is a thermal-runaway precursor if it progresses.
Immediate actions: open the DC contactors and LOTO the cabinet; verify zero volts at the rack; then perform an ultrasonic flow check on the branch, a thermal scan under controlled load if safe, and a rack-level capacity test. Correct the flow problem before any rebalancing. Do not run a passive balancing cycle first - balancing masks the symptom and delays the warranty claim.
2. Describe the maintenance and safety plan for accessing and repairing rack 14, including the documentation the warranty claim needs.
Read the worked answer
Safety. Racks in this class sit near 800–1,000 V DC. Use a CAT III 1,000 V meter, insulated tools, arc-rated PPE sized from the site incident-energy study, and the one-hand rule. LOTO the cabinet, open the rack disconnect or fuse, wait the manufacturer's specified time, and verify zero volts at the terminals before touching busbars. A swollen or off-gassing cell means evacuate and call the fire brigade - it is not a wrench problem.
Repair sequence:
- Replace the suspect hose, quick-disconnect, or strainer, bleed air, and verify flow and coolant ΔT against the commissioning baseline.
- Replace the suspect module or cell group. Do not replace a single cell inside a welded module unless the vendor permits it.
- Run a rack capacity test after the repair, then a full-system test at the next window.
- Add a monthly rack-level spread and temperature-delta report with an alarm at 80 mV so the next event is caught before the warranty threshold.
Claim package: spread trend, temperature trend, self-discharge data, flow and pressure logs, corrected repair records, and evidence that the 32 °C ambient and the 95% top-of-charge were within the warranty's operating envelope. Vendors routinely deny claims where the site exceeded ambient or SoC limits, so document both.
3. Is 87% capacity at year three a defect? Recommend operating changes that reduce further degradation.
Read the worked answer
Is it normal? One cycle per day for three years is about 1,095 cycles. The warranty floor is 70% at 3,650 cycles or ten years, so 87% at 1,095 cycles is on the high side of expected but not clearly a defect. The likely accelerators are the 32 °C ambient and the daily 95% top-of-charge. Calendar aging alone at 32 °C can cost 1–2% per year, and holding at high SoC increases it.
Recommendations:
- Lower top-of-charge to 90% and keep the low limit at 15–20% where dispatch allows; shift charging to the coolest part of the night.
- Improve thermal management: verify the container HVAC setpoint and coolant supply temperature. Roughly, every 10 °C reduction in average cell temperature substantially slows Arrhenius aging.
- Trend rack-level capacity, not just system capacity - a few strong racks can mask a weak one.
- Track both energy throughput and equivalent full cycles, because warranty claims are judged against the specific metric named.
Conclusion. The 87% system number alone probably does not justify a whole-system claim, but the rack 14 evidence supports a component-level claim. Fix the rack, then re-baseline the whole system.