There are numerous options to consider in the area of data center power system design, and every choice has an impact on data center efficiency and availability. The data center is directly reliant on the critical power system, and a poorly designed system can result in unplanned downtime, excessive energy consumption, and limited growth.
When considering options, the uninterrupted power supply (UPS) system configuration, UPS module design, efficiency options, and the design of the power distribution system should be top of mind.
INCREASE UTILIZATION RATE TO IMPROVE EFFICIENCY
Most businesses need to consider having some level of redundancy in their UPS system to mitigate the impact of downtime, eliminate single points of failure, and provide for ongoing maintenance.
A concern often raised in discussions about redundancy is utilization rates. A 2N UPS system that has the highest availability unfortunately offers the lowest utilization. Each bus of a 2N system can only be loaded up to 50% so that the one bus can provide full load in the event that other bus is not available. Many business critical data centers use 40% as the peak loading factor on each bus in this configuration to allow for variations in IT power draw and provide a cushion for immediate expansion capability. By moving 2N system utilization from 40% to 45%, an enduser could save as much as $33.5k on an 1,100 kW UPS system. To realize this savings through high utilization, find a trusted UPS supplier that can provide a maximum protection UPS that you are comfortable loading close to 50%. Customers have expressed concern that they don’t trust all UPS suppliers to be able to support 100% load. Find a UPS supplier you can trust who offers a UPS that can provide for the full load even during a stack-up of adverse conditions, such as high and low line conditions, temperature to 40°C, blocked filter, fan failure, and altitude.
DON’T GAMBLE ON AVAILABILITY – FAULT ISOLATION MATTERS
There has been growing interest in using transformer-free UPS modules in higher power, three-phase mission critical power backup applications (e.g. 200 kW to 5 MW). In general, well-proven transformer-based UPS systems are highly robust and excel at providing the highest capacities and availability while simplifying external and internal voltage management and fault current control. The latest transformer-free designs offer better efficiency, smaller footprint, and improved flexibility while providing high levels of availability. Surprisingly, even though the data center industry demands both transformer-based and transformer-free topologies, not all UPS suppliers provide both, limiting customer choice.
Those choosing a transformer-free topology should know that removing transformers exposes the UPS system to faults that could reduce the availability or push the critical load to utility power more often than a transformer-based system. A potential exposure is DC ground fault. Shorting the positive or negative battery terminal to ground in a transformer-based system results in an alarm, but the UPS continues to provide protected power. Shorting the negative or positive battery terminal to ground in a transformer-less architecture at best results in a transfer to bypass and the load is exposed to unprotected power, and at worst, drops the critical load.
Do you consider transferring to bypass during one of the most common UPS system faults robust? Don’t put your career on the line for it.
Potential cost savings of increased availability: According to the “Ponemon 2013 Cost of Downtime Study,” $690,000 per occurrence.
There is the misperception that transformer-based UPS systems are “old technology.” But in reality, modern transformer-based UPS systems deploy the latest digital signal processing (DSP)-based controls and energy optimization features to provide the best availability for business-critical applications and at efficiencies that can meet or exceed transformer-less offerings.
One such energy optimization method is active inverter eco-mode, which provides the majority of the critical bus power through the continuous duty bypass and offers a potential savings of over $22.5k on an 1,100 kW UPS system running 920 kW in active inverter eco mode 80% of the time. This technology keeps the inverter active and always ready to assume the load in the event of an outage, while other UPS systems that do not deploy the latest active inverter intelligent eco-mode could have a notch in the output waveform going in and out of eco-mode. These systems have to perform an interrupted transfer to turn off the bypass before turning on the inverter. Not the most ideal situation.
Data center managers are always looking for ways to reliably scale their capacity. Some UPS topologies force them to add hardware to increase capacity of a system that has already been factory witness tested and commissioned. Adding equipment at this stage could pose a risk to availability.
Ideally UPS units would provide capacity on demand by using software to increase capacity without adding to the system footprint or put the system at risk with the introduction of unproven hardware. UPS units with this software functionality can start with current capacity requirements and then easily scale up to a larger capacity with a simple software key as power needs change. These types of modules are scalable in increments, such as scalable from 40 to 80 kVA, from 80 to 120 kVA, or from 160 to 200 kVA.
Some row-based, modular UPS units allow quick power capacity increases with the addition of internal power core hardware assemblies. These core assemblies allow the system to expand for capacity or redundancy in standard capacity increments within a single cabinet. It is important to look for power cores that incorporate distributed intelligence and scalable power in a common assembly and allow configuration of a completely redundant power and control system. When power requirements change, data center managers can easily add capacity without increasing the system footprint. This approach allows for right-sizing of the UPS, resulting in improved energy efficiency and reduced power expenditures.
Savings can be significant. If the end state is 600 kW, it could cost up to $52,000 more to start with a 300 kW module and later add another 300 kW module, vs. starting with 400 kW and employing software capacity increases from 400 kW to 500 kW to 600 kW.
UPS systems equipped with intelligent paralleling can improve the efficiency of redundant UPS systems by idling UPS modules that are not required to support the load while maintaining redundancy and taking advantage of the inherent efficiency improvement available at higher loads.
For example, a redundant multi-module UPS system configured to support a 500 kVA load using three 250 kVA UPS modules can support the load with only two modules operating at higher load levels, while maintaining redundancy with the third, standby module.
This feature is particularly useful for data centers that experience extended periods of low demand, such as a corporate data center operating at low capacity on weekends and holidays.
A 1,200 kW system of 3x 400 kVA UPSs could save 5 kW of energy when running at one-third load with intelligent paralleling enabled vs. running the full system without the benefit of intelligent paralleling.
WEIGH SAFETY RISKS AND HIDDEN COSTS OF ALTERNATIVE DISTRIBUTION VOLTAGES
Normal 3-wire distribution for enterprise data centers is 480 V. There has been much discussion about going 400/230 V 4-wire direct from the UPS to the server. While this configuration looks good on paper it has some significant limitations:
• Fault current can be much higher in these direct-to-the-server configurations
• This configuration also strands capacity in the gear
• Requires higher ampacity buses and can increase wiring costs
Before going to this extreme, consult a trusted data center adviser to understand the costs and risks associated with this architecture, and for alternate suggestions to meet data center goals.
Higher efficiency and utilization can be achieved without going to the extreme by moving from 208 V to 240 V distribution with a potential annual savings of $17k per year for an 1,100 kW system running at 920 kW load.
Do your research. Due diligence on the latest UPS technology and efficiency optimization modes will help you choose improved critical power systems with the highest availability and new levels of utilization and efficiency.