Centralized Uninterruptible Power Supply (UPS) systems are becoming a standard feature for new commercial buildings. Usually associated with the mission critical market sector, today, even less-technical buildings, such as office spaces, are including UPS systems to maintain service continuity for their employees and customers.
When designing UPS systems, it is essential to examine the capacity of the UPS closely. In addition to more expensive feeders and distribution equipment, oversizing a UPS results in lost efficiency, causing the overall operating cost to increase also. In cases where redundant UPSs are installed, such as a 2N arrangement, right-sizing is especially important since the design load is 50% or less. A close review of the Day 1 loads, future loads, and spare capacity offers many advantages. When UPSs are aligned with the load power requirements, the design may save floor space in the building, decrease the eventual cost to replace in the UPS(s) in the future, make the DC-link energy storage requirement smaller, and reduce the load on the supporting cooling systems.
Given the importance of right-sizing, this blog post overviews some of the considerations for sizing central UPS systems.
UPS Nameplate Rating
Depending on the type of UPS, the apparent power (kVA) and real power (kW) capacity ratings may not be unity Power Factor (PF) = 1.0. For example, transformer-based UPSs range from 0.8 to 1.0 PF. Transformer-less UPS designs sometimes have oversized internal components that also cause the rated PF to be less than unity. In addition to the UPS itself, consider the loads that will be powered. For instance, a UPS nameplate may list input = 625kVA, output = 625kVA @ 0.9 PF; thus, the UPS is rated to power a total load of 562kVA @ 1.0 PF or 625kVA @ 0.9 PF.
National Electrical Code (NEC)
In most cases, UPS systems are sized based on the connected load. Except in building layouts with many general-purpose receptacles, the NEC permitted demand and diversity factors are not utilized: design load = connected load. The electrical design does have to comply with NEC code minimums. For example, a general-purpose receptacle powered by the UPS must consider at least 180VA per yoke. In installations where redundant branch circuits are supplied, such as to a dual-corded computer power supply, the NEC permits load reduction due to noncoincident loads.
Computer Rooms
Computer rooms, or information technology equipment rooms, vary in physical size from small closets to large scale data centers. The computer rooms contain equipment racks (and/or cabinets) used to power and network information technology equipment. The type of rack equipment varies based on the application and also over time, as rack densities continue to increase and technology changes. The load profile of the UPS will vary accordingly, but the following rack capacities are a good starting place.
Concept Project Phase: UPS and Branch Circuit Design Capacities
| Application | UPS Design Load (kVA, Per Rack) | Branch Circuits (kVA, Per Rack) |
| Small Network Closest | 0.5 to 2.5 | 5 |
| Enterprise Data Center | 5 to 7 | 8.6 |
| Research Data Center | 7 to 10 | 17.2 |
| Co-Location Data Center | 7 to 10 | 17.2 |
| High Performance Computing or HPC (Super Computers) | 40+ | 40+ (multiple circuits) |
Indeed, design capacities may need to be even higher. Note the branch circuit column is different than the UPS design load; this is to allow design flexibility across the racks since the owner equipment will have refresh cycles and equipment may not be uniformly added to the racks. The electrical team should work closely with the owner, mechanical team, and technology designers to refine the power requirements to suit the project. A helpful place to start is by reviewing the Computational Fluid Dynamic (CFD) model (if available) input parameters, which is useful for the cooling design that needs to align with the rack power. Also, consider the specified computer rack-mounted Power Distribution Units (rack PDUs). For example, 17.2kVA equates to a 60A @ 208V/3PH rack PDU (80% due to continuous load, as required by the NEC).
Motor Loads
In general, try to avoid motor branch circuits on Static UPS systems. The UPS capacity must accommodate the motor locked-rotor current (momentary in-rush). Depending on the motor and associated controller, this could be six times the Full Load Current (FLC) of the motor or more, resulting in an oversized UPS.
In some HPC applications that bring a liquid medium to the computer rack for cooling, motor loads on the UPS may be needed to avoid overheating while the system waits for a back-up generator to start. If the application requires uninterrupted power for motor loads, consider if a transformer-based static UPS or industrial UPS for robustness. One alternative to oversizing due to motor loads is to consider a rotary UPS design instead of a static UPS.
Generator-UPS Compatibility
For large campus installations with central plants serving multiple buildings, the size of static UPS system compared to generator capacity is less of a worry. However, in mission critical applications where the UPS supports nearly all of the load, close attention to the UPS and generator sizing is needed. For example, UPS designs that utilize SCR instead of IGBT based rectifiers (front-ends) may require a generator up to 2X the UPS capacity.
At low loading (less than 30%), to avoid reverse power, UPSs with passive input filters may need to be switched offline upon being powered by the generator. Additionally, the filters may increase the Total Harmonic Distortion (THD) of the UPS system, causing alternator heating and insulation stresses in the genset. For designs that utilize tuned filters, switching filters out of the circuit may not be required.
Last, consider the generator sub-transient reactance, as a higher sub-transient reactance (above 12%) equates to a less stiff source for the UPS. The UPS system may go offline (off inverter), leaving the critical loads subject to unfiltered power and possible interruption. Even if the sub-transient value of the generator is compatible, it is common practice to install signal wiring at the UPS for monitoring by the on-board controller. By letting the UPS know about the source change, amongst other benefits, the UPS may be programmed to accept a broader range of input power imperfections and stay online while on generator.
A great start to reviewing generator-UPS compatibility is by using generator manufacturer free commercial software.
Summary
UPS systems are necessarily insurance policies. Even outside of the mission critical market sector, many owners are calculating how expensive downtime is to their organization and electing to have centralized UPS systems installed in their buildings. Many design considerations are needed to retain the value of the UPS system. Collaborate with your local UPS vendors and seek qualified peer reviews to make the best decisions for your next UPS design!