Save electricity while you sleep

A UPS system is all about maximum availability. The system must be constantly ready for operation and supply the load with clean, stable mains voltage at all times and under all circumstances. It is not for nothing that the VFI-SS mode, in which the output voltage is generated completely synthetically from the battery voltage via an inverter independently of the mains, is considered the ultimate in terms of output quality. Of course, constant operation also means constant power consumption. Despite enormous improvements in circuit technology, the efficiency of such systems is in the range of 90 to 97 percent. In some cases, it is also considerably lower if, as is common in many data centers, the system only handles half of the load as part of an A/B supply network and a second UPS system supplies the other half. Because UPS systems reach their maximum efficiency close to full load, this constellation is particularly inefficient.

No transfer time due to modularity

Due to the high availability requirements in data centers, where system failures are simply intolerable, UPS systems that work according to the VFI-SS principle are considered indispensable. This is at least partly due to historical reasons. In the past, the only alternative to VFI-SS were systems that required relatively long periods of time to generate the mains voltage from the inverter in the event of a fault. What was known as a backup or line-interactive UPS generally used the mains voltage to supply the load and only switched to inverter operation in the event of a serious problem. During the switching process, the load was not supplied with power. The so-called transfer time was in the range of a few tens of milliseconds - unproblematic for private use, but too risky for highly utilized servers. So there was no way around the VFI-SS UPS, formerly known as the online UPS.

However, developments in the UPS sector in recent years have been dominated by modularity. If the load to be protected is split across several modules, it becomes much easier to build redundancy at module level. For n+1 operation, a complete additional UPS system is no longer necessary, but only one additional module and for n+2 redundancy, two additional modules are sufficient. This opens up completely new possibilities for the operational readiness and availability of UPS systems. Because there are several modules supplying the load, the question is no longer "either or?" when it comes to energy efficiency versus supply guarantee. Current modular UPS systems allow numerous operating modes in which, depending on the power requirements and availability requirements, not all modules work continuously, but a balance is found. The result is drastically improved efficiency with consistently high availability.

In today's standard modular UPS systems in the medium and high power range up to two megawatts, modules are often used that are not actually needed. In normal operation, when both supply rails are working, half of the modules could actually be switched off, thereby saving energy and reducing wear on these UPS modules. However, if a fault occurs - for example, if one of the two supply networks A or B fails in the worst case - these UPS modules must be able to be reactivated quickly and without interruption to the load.

Hibernation mode saves when used correctly

Further developments in circuit technology have now made the necessary intelligent hibernation mode (hibernation) in the UPS possible. Users can configure numerous parameters that determine when and how many UPS modules remain in hibernation mode while the active modules supply the load. Of course, there are numerous requirement profiles in the data center, not every server hosts applications of the same importance and not every load situation requires the same level of redundancy reserve. The UPS system must therefore offer detailed setting options to ensure that hibernation mode does not have a negative impact on availability.

Among other things, it is essential to define the load reserve that the UPS system maintains as redundancy. If the modules each have a nominal load of 50 kW and the load is 100 kW, two modules are required for standard operation. With n+1 redundancy, a further module must be actively running, with n+2 two modules. If the overall system can handle a nominal load of 500 kW, i.e. has a total of ten modules, five of them are unnecessary even with n+2 redundancy and can switch to hibernation mode. As soon as the load changes, the UPS system can retrieve modules from hibernation mode and insert them into the system. The n+2 redundancy is maintained throughout the entire time. It would also be possible to put additional modules into hibernation mode when the load drops.

Advantages of the intelligent sleep mode

In addition to the energy savings from the idle modules, which can be considerable depending on the UPS system, users also benefit from the intelligent idle state in other ways. This is because modules that are switched off do not generate any waste heat, do not need to be cooled and are not subject to wear and tear in terms of operating hours. To prevent some modules from running constantly and others from not being used at all, it is important to use all modules as evenly as possible. A corresponding algorithm in the UPS system regulates the cycles between the idle and working phases of the modules. Users should be able to define the pause cycles for this rotation. Values between a few minutes and a few months make sense. In the default setting, many UPS systems change the modules at intervals of several minutes. For problem-free planned maintenance, it should be possible to remotely restore modules to normal operation in sleep mode using VPN-secured network access.

It is also crucial that the hibernation mode can be defined for the UPS system as a whole. Especially when several modular UPS systems are connected in parallel, all subsystems must be aware of the current status of all modules and the load and availability requirements. This means that complete subsystems can also be switched off, which has the greatest benefits in terms of energy efficiency because not only the power modules but also the surrounding infrastructure of the subsystem no longer consumes any power. In the event of a load change or a sudden load increase, the modules switch back to normal operation in such a short time that there is no interruption to the load supply. Of course, the UPS system must also be able to absorb the potentially large increase in load within a short period of time when restarting. In bypass mode, the overload current should be able to withstand up to ten times the nominal current; in inverter mode, a factor of five makes sense.

Battery hibernation: the batteries age much more slowly

Extend sleep mode to battery charge

This hibernation mode is not only a sensible state for the power modules. Hibernation mode also reduces wear and costs for the battery modules. As a rule, the trickle charge voltage is permanently applied to the batteries. With larger batteries, this leads to a continuous charging current that heats up the battery. This leads to very rapid ageing, especially if the battery is not air-conditioned or is already exposed to higher ambient temperatures. Intelligent battery management can also be used to extend rest periods to the batteries so that charging breaks of several days to several weeks can be set. During this time, the batteries lose virtually no capacity, consume no charging power and age significantly less.

It goes without saying that the batteries must always be monitored during this time. It is also important that the UPS system - depending on the type of battery used - can be programmed for trickle or equalization charging after the charging pause. In order to adapt the system optimally to the battery, it is necessary to set parameters such as charging voltage for trickle charging, charging voltage for equalization charging, charging current for trickle and equalization charging and the duration of equalization charging, trickle charging and charging pause independently of each other.

In modular systems, redundancy is not just limited to the power units, but must also include parallel communication. Redundant control units and - if several system cabinets are used - redundant parallel cabling in ring topology make sense. Last but not least, the user must also be able to operate the system if an operating unit or display fails.

Individual consideration necessary for optimum availability

Optimizing energy efficiency versus availability is no trivial task. For an optimum operating point, the load should be divided between as many of the available power modules as possible. The higher the number of power modules, the smaller the gradation and therefore the difference to the optimum operating point. However, the hibernation mode is then rarely or never reached and there is no savings effect. Installing more modules than necessary is also a way of distributing the load across many modules and still having reserves that can be switched to hibernation mode. However, more modules also mean higher acquisition costs and more effort for maintenance. In practice, it is important to find an economic compromise that is determined by the best reliability and lowest TCO (Total Cost of Ownership).

If this intelligent sleep mode succeeds in saving even half a percent of the annual losses, this means an annual saving of 35,000 kWh of energy with a load of 800 kW. This is roughly equivalent to the energy consumption of ten households with two to three residents.