Two UPS Modules in N+1 Against an 18-Month Growth Plan
A 600 m² enterprise data centre runs two 500 kW UPS modules in parallel N+1. The measured IT load at the UPS output is 430 kW at a load power factor of 0.95 lagging. An approved business plan adds 260 kW of IT load in the same room within 18 months. The growth case was sold internally on the basis that installed UPS capacity is 1,000 kW against a forecast 690 kW load - '69% loaded, plenty of headroom'. The building standby generator is a single 1,000 kVA / 800 kW prime-rated set feeding the whole building through an automatic transfer switch. Mechanical cooling on the same board draws 190 kW and restarts in two stages. During review a colleague notes that the 6-pulse UPS rectifiers produce significant harmonic current and that the generator has never been load-bank tested above 400 kW. The facility manager wants to sign off the growth plan this week.
1. Test the '69% loaded' claim. Does N+1 redundancy survive the growth plan? Show the arithmetic and the correct denominator for a redundancy decision.
Read the worked answer
Convert everything to the same units before accepting a percentage. Measured 430 kW at 0.95 PF is 452.6 kVA; forecast 690 kW at 0.95 PF is 726 kVA.
The '69%' figure divides forecast kW by the sum of both module ratings. That is the correct denominator for utilisation but the wrong one for a redundancy decision, because redundancy is proven one module at a time.
- Normal operation after growth: 690 kW spread across two modules = 345 kW each, 69% each. Comfortable.
- One module out: the survivor must carry the full 690 kW, which is 138% of a 500 kW module. It cannot. The load transfers to static bypass (unprotected) or the module trips on overload. In kVA terms 726 kVA against a 500 kVA module is worse still. Redundancy is gone.
- Even today: at 430 kW the survivor would run at 86% of rating - inside the nameplate but above the 80% design ceiling normally used to leave room for transients, battery recharge and unforeseen growth.
Two further points the growth case omits. First, battery autonomy falls roughly in proportion to load: a 10-minute autonomy quoted at design load may be around 6 minutes at 690 kW, which shortens the window between a utility failure and generator availability. Second, heat rejection from the UPS grows with load and must be fed into the cooling calculation.
Conclusion: N+1 does not survive the growth plan. Either add capacity (a third 500 kW module gives a genuine N+1 at that forecast, with the survivor at 69%) or impose a hard load cap until capacity is added.
2. Assess the generator against the post-growth building load, including step-load and harmonic effects, and state what must be verified rather than assumed.
Read the worked answer
Build up the load on the generator from the load list, not from the UPS rating:
- UPS input: 690 kW output ÷ roughly 0.96 efficiency ≈ 719 kW, plus battery recharge current for a period after transfer. Budget 750–800 kW peak.
- Cooling on the same board: 190 kW, restarting in two stages.
- Lighting, controls, losses: realistically 40–60 kW.
- Total ≈ 980–1,050 kW against an 800 kW prime rating. The set is undersized before any derating. In kVA terms the alternator also sees roughly 1,030 kVA plus harmonic current against a 1,000 kVA rating.
Two effects must be verified from manufacturer data, not assumed:
- Step load. On a utility failure the UPS holds the load on battery and the rectifiers re-connect after a walk-in delay, so the generator sees the rectifier block load arriving in one or two steps while the chillers restart in two more. The set must be checked against its step-load acceptance curve - typically something like 50–60% of rating in one step followed by timed increments - at site ambient temperature with the manufacturer's altitude/temperature derating applied.
- Harmonics. Unfiltered 6-pulse rectifiers draw high THD-i, and alternators must be derated for non-linear load; use the manufacturer's derating curve against the applicable generating-set standard rather than relying on the kW figure alone.
The 'never load-banked above 400 kW' note is a verification gap, not a reassurance. Without a site test at the real step profile, every number above is an estimate.
3. What is your recommendation on the sign-off, and what interim operating rule would you impose?
Read the worked answer
Do not sign off. The growth plan is technically unsupportable on two independent grounds - redundancy lost at the UPS layer, and a generator that cannot carry the post-growth building load even at nameplate, let alone after derating.
Options, in order of preference:
- Defer the load and cap the hall until a third UPS module and a correctly sized generator (or a second set in parallel) are installed. Cheapest in capital, most expensive commercially - be explicit about that trade-off.
- Fund the upgrade now: third module plus generator replacement, with the electrical design showing the one-out case and the step-load case.
- Accept partial growth with a documented, tested load-shedding scheme and a generator that has passed a site step-load test at the real profile.
Interim operating rule to write into the capacity register: the maximum IT load at which N+1 is retained equals one module rating times the agreed utilisation ceiling - for example 500 × 0.9 = 450 kW, or 400 kW at an 80% ceiling. The N+1 claim must be restated against that number, not against installed kW. Add three mandatory actions before the next peak: a site load-bank and transfer test at the real step profile, correction of the published battery autonomy to the actual load, and a monthly report of UPS load versus the cap. Record all of it so the next person inheriting the facility cannot repeat the '69% loaded' error.