By Markku Vainamo, ABB Global Product Manager for Generators: https://new.abb.com/motors-generators/generators/data-center-generators
Backup generators are vital to ensure 100% uptime and reliability for data centres.
But not all generators are created equal. Markku Vainamo, ABB Global Product Manager for Generators, outlines the most important technical factors to consider when specifying a generator for a data centre.
What do continuous data centre power (CDCp) ratings really mean?
It is important to recognise that data centres cover a wide diversity of sizes and power demands – everything from racks of a few kilowatts (kW) to multiple megawatt (MW) hyperscale data centres. That means the demand for availability also differs greatly. This demand for reliability is often expressed by TIER levels I to IV. Availability (or uptime) is typically specified as follows:
- TIER I: availability of 99.671%
- TIER II: availability of 99.741%
- TIER III: availability of 99.982%, unlimited operation hours
- TIER IV: availability of 99.995%, unlimited operation hours
These may seem like quite high uptime percentage ratings. However, the true picture emerges when you convert the figures into actual downtime. Because we then have a range of acceptable periods of downtime from 144 minutes to just two minutes per month (from TIER I to IV respectively) and this is where the difference comes in.
By understanding the need for uptime, it is then possible to make the correct choice for backup equipment. In practice, TIER I-II can be accommodated by standard back-up gensets suitable for a couple of hundred hours of operation a year. In contrast, TIER III-IV requires gensets rated for continuous operation.
In reality, most data centres receive their main power supply from a largely reliable source, such as a connection to the national grid. This is the reason why many manufacturers have specially designed their equipment with ratings known as “continuous data centre power” or CDCp. This ensures there are no limitations on average power, unlimited operational hours and 110% overloadability for between one and 12 hours. The way this compares with the ISO standby rating is shown in Figure 1.
Defining genset performance
Having defined the TIER levels and ratings according to the data centre’s needs, the next step is to look more closely at the genset design to make the correct choices to ensure optimum performance. The most important characteristics are start-up time; the excitation and control system; block load-ability; leading power factor.
The start-up time of a generator set depends mainly on the size and type of the machinery driving the generator combined with its starting capability. In general, a combustion engine is superior to a turbine and liquid fuel performs better than gas.
For data centres and other mission-critical applications, there is an often-quoted start-up time requirement of 10 to 30 seconds. Typically, this is achievable by liquid fuel engines or pre-heated gas engines (“hot start condition”). There are alternatives on the market, such as rotary UPS systems. However, for plants where there is only an occasional need for backup starts, the constantly rotating system is often seen as less attractive.
Excitation and control
Once the genset is starting to gain some speed, the excitation and control by the automatic voltage regulator (AVR) kick in. There are several excitation methods available, such as shunt+boost; permanent magnet generator (PMG); auxiliary winding regulation excitation principle (AREP); self-excitation etc.
The most prominent method is PMG for 1-3 MW sets, while AREP is for smaller, low voltage and non-critical systems. AREP has its advantages, mainly cost, but the biggest downside occurs should there be any insulation issues in the stator. Therefore, AREP in mid to high voltage stators is often seen as a major concern. The shunt+boost system is equal to PMG in terms of technical performance but it comes without extra rotating components.
When the excitation circuit and control are properly designed and tuned, the voltage will be set to have an acceptable overshoot to support faster settling to the rated voltage and the fastest possible voltage build-up. The best performance is obtained when incorporating shunt+boost or PMG together with PWM (pulse-width modulation) control AVR. Even so, the system must be fully tuned and validated to achieve the best performance. Figure 2 shows measured ramp-up and voltage build-up graphs for a 3 MW, 11 kV, PMG-excited, liquid fuel genset. It takes less than 10 seconds to reach full speed and full voltage ready to supply power.
Another important role of the excitation and control is realised when the genset is being loaded, especially during dynamic operations such as a sudden increase or decrease in load or during frequency fluctuations. Meeting grid codes also brings requirements like fault-ride-through. This, in certain cases, can apply to gensets that are also intended to supply power for loads outside the data centre, such as “balancing” or “peaker” applications. This is an interesting and fast-growing application as it enables asset owners to open up additional revenue streams that offer a greater return on the investment in their backup systems.
Furthermore, using backup gensets to support loads outside the data centre is a potential solution to help facilitate the global energy transformation to using more environmentally friendly renewable energy resources. This is especially the case in the EU, where there is increasing interest in utilising the large fleet of backup gensets to stabilise and support the grid when the output from wind and solar power plant is low.
Operating as a balancing power resource has slightly different design principles compared to the original standby resource and the typical solution is to use higher (often gas-powered) genset ratings. Currently, many manufacturers are testing new fuels including hydrogen and other gas blends, as well as synthetic fuels, to limit CO2 emissions. A balancing power genset might need to be started up from zero speed several times per day, so it needs to be designed with this operational pattern in mind.
Leading power factor
Last, but not least, is the “historical weighting” of backup generators linked to the leading power factor (PF). In data centres, this occurs at times when the UPS is out-of-order and the load on the genset is highly capacitive – from the IT equipment. If the leading PF (or negative kVAR) limit is exceeded, the generator approaches its instability limits and it could trip offline. This is definitely an undesirable event when the data centre is operating on backup power.
The historical solution has been to oversize the generator to ensure its operation over a certain kVAR range. However, this added cost can be avoided through careful generator design to allow extensive leading PF operation. In addition, the parameters, especially reactance values, must be designed properly to “push” the stability limit “left”, as shown by the purple line in Figure 3.
This example of a 2.5 MVA, 10.5 kV generator shows that by enabling a leading PF of 0.9 (instead of 0.95) there is an additional +300 kVAR of room to play with (as shown by the red arrow).
Figure 3 – Extensive leading PF operation avoids generator oversizing.
OPEX will become more important than CAPEX
This article has outlined the main design characteristics to consider when selecting a mission-critical generator for a data centre. They have a main influence on the capital expenditure (CAPEX) for the project, and currently, the major players in the industry are keen to reproduce existing validated concepts as far as possible to benefit from the lowest unit cost. Scalability is covered simply by selecting the required amount of identical gensets.
The focus on CAPEX is likely to change in the near future as asset owners place an increasing emphasis on their operational expenditure (OPEX). These costs will differ widely between applications and locations. However, the main drivers will always be quality and serviceability. It is also possible that a range of features will be adopted from base load generators, like predictive maintenance and remote monitoring. The likely outcome is the original unit cost will no longer be the critical parameter for data centre operators when specifying their critical backup generators. Instead, they will be more inclined to look at the total cost of ownership (TCO) of their generator solutions and how they might even be a source of revenue rather than a cost.