Generator ratings and the implications for data centers
Understanding how generator ratings are determined is quite simple, but there is a long list of variables to consider. This reviews the fundamentals of generator ratings and a recent industry discussion about specific ratings for generators in data centers.
- Differentiate power generator ratings in data center applications.
- Understand power generator components.
- Consider impact of new emissions standards on generator ratings.
Recently, some generator manufacturers have created ratings specifically for products they supply to the data center market. This is in direct response to the technical white paper, “Data Center Site Infrastructure Tier Standard: Topology,” published by the data center organization Uptime Institute LLC. The paper states that for Tier III (99.982% reliability) and Tier IV (99.995% reliability) designs of data centers, the generator, not the utility, is to be treated as the primary source of power. It is important to note that the references to Tier III and Tier IV designations have nothing to do with EPA Tier 3 and Tier 4 emissions requirements for engines (which we’ll discuss later)—an unfortunate coincidence in designation that has led to a good deal of misunderstanding in the industry.
The Uptime Institute’s paper specifies that the generator must be “capable of supporting the site load on a continuous basis” and refers to ISO 8528-1 (International Standard for Reciprocating Internal Combustion Engine-Driven Alternating Current Generating Sets), in which four generator ratings are defined: standby, prime, limited running time, and continuous.
As a result, data center designers have specified generators with a “continuous” rating based on the peak demand of the site, which in turn has led to the installation of oversized generators that add unnecessary cost to the data-center design. Some generator manufacturers are even promoting and specifying a “continuous” rating for data center use on their generators to sell larger models and boost sales. We believe that oversizing of generators in the data center industry adds unnecessary cost and diminishes efficiency. When you consider the actual run time of these generators, oversizing does not provide value to the customer or improve reliability of the system.
Specifying a generator rating
Here’s why specifying a continuous rating (as the Uptime Institute recommends) for generators in the data center industry is unnecessary. First, consider that the main purpose of the ISO 8528-1 ratings standard is to classify the type of application for which a generator rating is to be used. In nearly every case, data centers are standby (emergency) applications. The generator is not the primary power source; it is used to back up a reliable utility, or in the case of Tier III and Tier IV data centers, multiple independent utilities. The only exceptions are areas where there are no utilities, and the generator is used as prime power.
Figure 1 shows the relationship between the load level and engine life relative to the rating. The highest rating for a generator with a given engine and alternator combination is the standby rating; the lowest is the continuous rating. This relationship is the same for all engines (and generators). Although the slope of the line may differ from generator to generator, the relationship of these ratings remains the same. The more hours you plan to run the generator at a given load, the lower the rating needs to be to maintain the same level of reliability.
When choosing the generator rating, you need to know how many hours per year you expect the generator to be used and what the load profile looks like. As stated above, the majority of data centers are stationary emergency (standby) applications; generator run time is low when backing up a utility. So even though the Uptime Institute specifies that the generator should be the primary power source for Tier III and Tier IV data centers, we know this is not the reality.
Another important factor to consider is the load profile over time. Figure 2 shows a standby application profile that has a varying load over time. Data centers have a varying load profile. In its white paper, the Uptime Institute provides generator rating definitions referenced to ISO 8528-1, which include a continuous rating. This does not consider the rest of the definition in ISO 8528-1, which states that the load is also nonvarying. Figure 3 shows a continuous-rating load profile. Continuous ratings are intended for peak shaving programs where a designated block of power is sold back to the utility or used on-site, but are not intended for applications where the load varies. A data center’s load profile is always variable, which fits the standby rating definition in ISO 8528-1.
Consider a worst-case scenario. In 2005, Hurricane Katrina created an extremely long utility outage. According to the Energy Information Association, it took 30 days to resolve 90% of the utility outages after the hurricane. When you consider that a month of continuous generator run time equals 720 hours and that the engines and alternators used in Kohler generators for large data center applications, for example, are designed for tens of thousands of hours, a clearer picture starts to emerge. The fact is that standby-rated generators can run an unlimited number of hours with varying load during the utility outage. (Unlimited hours are subject to the manufacturer’s recommended maintenance intervals.)
Another point to remember is that the main causes of a generator’s failure to start or to operate during an outage are not due to engine life, but to fuel contamination, low or dead starting batteries, or a poorly run maintenance program that did not follow the manufacturer’s recommendations.
Some generator manufacturers offer multiple standby ratings definitions that vary based on the average load factor and number of hours run. ISO 8528-1 states that the average power output shall not exceed 70% power in a 24-hour period, unless the engine manufacturer agrees to a level higher than 70%. The engines in some generators allow for an 85% average power level for large data center applications. Thinking back to Figure 1, this is a true statement of an engine’s durability. The generator can be specified using the manufacturer’s generator-sizing program; all that is required is to enter the planned site load profile.
Given all of the above, our recommendation is that data center designers fully consider the actual requirements of the generator during an outage. Specifying the correct size generator at the start and then following a structured maintenance and testing program will prove the most sensible course of action.