Diagram decorative image

Don?t Keep Your Servers as Comfortable as You

Part 1: Optimizing the Performance of Your Chilled Water System
Many people think if their mechanical systems follow ASRAE 90.1 – the most widely adopted energy standard by code – they are running at their highest efficiency. However, ASHRAE 90.1 is only a standard for minimum energy performance, and by simply following it, they may be only meeting this minimum while overlooking opportunities for further efficiency improvements. Without the temperature demands for comfort cooling, data centers have a unique opportunity to improve upon the efficiency benchmarks provided by these energy standards by utilizing independent chilled water systems.

First, let’s take a look at the heart and largest energy consumer of a chilled water system, the chiller. We can start by looking at the key variables affecting chiller performance – load and lift.

What is load?
Chiller load, usually measured in tons, is the amount of heat the chiller needs to remove from the space and reject to the atmosphere.1 It is important to remember that heat and temperature are two separate things. In cooling, heat can be thought of as the amount of energy that needs to be removed to maintain a specific temperature.2

What is lift?
Lift is a technical term dealing with pressure differentials within the chiller. It is essentially determined by the difference in temperature between the evaporator side and the condenser side of the chilled water system. For every degree in temperature that this difference is reduced, whether on the high or low side, the chiller performance can be improved by up to 2%.3

How can chilled water systems be optimized?
With a basic understanding of load and lift, we can look at the opportunity to improve efficiency in your data center by segregating comfort cooling loads (office space) from the process cooling loads (data center space). Often times, these loads are cooled by the same system. In this situation, the performance of the cooling system is driven by the more strict cooling demands of the two – that of comfort cooling. Comfort cooling requires colder temperatures for employee thermal comfort, and these cooler temperatures dictate the chilled water set-point of the system.

The goal of data center cooling is to keep the server equipment happy and minimize downtime due to thermal overload. Per ASHRAE, the temperature of air supplied to these environments can be up to 80?F. This is significantly different from most comfort cooling applications where the system is returning temperature from the space at 75?F. Keeping the data center at this higher temperature only reduces the lift on the chiller if it is allowed to operate at a higher set-point associated with the higher temperatures in the space. In other words, a chiller making warmer water for data center cooling can operate much more efficiently than a chiller that needs to make colder water for comfort cooling. Therefore, we are missing a huge opportunity in overall efficiency if the chiller is operating based on the office space requirements.

This combined cooling system concept can be very inefficient because the average percentage of office space supporting data center facilities is generally only around 10%. Let’s say we have a 100,000 ft2 data center facility with 10,000 ft2 office space. General office space cooling loads are typically somewhere between 5 and 7 Watts/ft2, and typical medium density data center cooling loads are somewhere between 125 and 150 Watts/ft2. As you can see, the process cooling load out-measures the comfort cooling load by a significant amount (over 200 times in this typical example – 15,000kW vs. 70kW). If the chiller is working to meet the more strict cooling demands of the office space, even though the load is only 0.5% of the total demand on the system, the data center cooling efficiency suffers dramatically. The efficiency for the entire plant is decreased to meet the demands of that small office space.

Per today’s standards for data center temperature requirements, we have seen chillers, such as the ones at Data Realty in South Bend Indiana, operate up to 36% (or 0.239 kW/ton) at full-load, and up to 40% (or 0.180 kW/ton) at part-load improvement over ASHRAE 90.1 standards. This was only possible by segregating the two cooling load requirements into separate cooling systems.

With separate cooling systems for the data center and office space, each system is optimized for its intended purpose. According to Commonwealth Edison’s website, chillers alone account for 33% (on average) of the typical data center’s energy usage – outweighing the IT server equipment itself by 10%. Reducing the chiller energy consumption by 40% like Data Realty has a net impact on the overall energy consumption of most data center’s bottom line of 13%. And as an added bonus to your increased efficiency, utility companies such as ComEd offer attractive rebate programs for implementing such systems.

1 The heat can be rejected through either an air-cooled condenser or a water-cooled tower.
2 The amount of chilled water load on the system is directly proportional to the UPS output, and thus cannot be optimized to improve the performance of the system. However, there are ways to apply the equipment to better handle the variations in load and optimize the system. These include the use of variable frequency drives, the addition of part-load machines or steps of compression, and the integration of control sequences to properly stage the equipment at part-load conditions.
3 The condensing side (higher in temperature) can be reduced by using water or evaporative cooled condensers. The evaporator temperature (lower in temperature) can be increased by raising your chilled water supply temperature set-point.