Humidity Control in Data Centers (March 8, 2017)

Overview

The report, written by Vali Sorell of Syska Hennessy with Magnus Herrlin of LBNL as principal investigator, starts from the history of humidity control. Tight bands of roughly +/- 5% RH were originally imposed for paper printers, punch cards and magnetic media, not for the computers, and those devices were later moved into separate rooms. The remaining concerns are electrostatic discharge when air is too dry and corrosion or hygroscopic dust failure when it is too moist. ASHRAE TC 9.9's guidelines widened the recommended range from 40-55% RH in 2004 to 15.8 °F dew point up to 60% RH, capped at a 59 °F dew point.

A central argument is that dew point, not RH, should be measured and controlled. RH depends on temperature, so sensors in the hot and cold aisles read differently, while room dew point is nearly uniform when the cooling coils run dry and no moisture is added. The only real moisture source is outside air, so humidity control becomes a matter of conditioning a small outside air stream. The report notes that the ASHRAE guidelines refer only to air entering the IT equipment, and that legacy CRAC units often dehumidified while neighbouring units humidified, which it likens to driving with brake and gas pedals pressed together.

The control chapter gives a concrete sequence. Select CRAH or CRAC units for the highest practical supply air temperature so their coils stay above room dew point, and give all humidity control to a dedicated outside air system (DOAS). Use at least three temperature and RH sensor pairs, convert to dew point, discard any pair out of calibration and control on the average. Dehumidify above about 59 °F dew point and humidify below about 16 °F, with a 4 °F throttling range, for example 100% response at 61 °F and 0% at 57 °F for the 59 °F set point. A separate sequence is given for airside economizers.

Water is treated as part of the energy picture. A 1 MW load rejected by evaporation consumes about 3,500 lb/h (421 gal/h) in free-cooling mode and about 4,400 lb/h when chillers run, because of roughly 0.25 MW of extra heat; blowdown adds 10-20%. A water-cooled chiller plant therefore uses about 25% more water on site than direct evaporative cooling, while air-cooled or DX cooling shifts water use to power generation. The report also compares wetted media, ultrasonic and spray adiabatic humidifiers, which use far less energy than steam boilers or infrared units whose heat adds to the cooling load.

The benefits section stresses fewer hours of humidification and dehumidification, avoided short cycling from oversized equipment, slower rates of environmental change, and smaller electrical and generator capacity for humidifiers and reheat. The report is a guidance document built on ASHRAE editions up to 2015, and its closing note says that relying on excursions into the allowable range may require a documented risk analysis in some cases. It also frames the choice between airside free cooling with a large moisture load and minimal moisture without free cooling as mutually exclusive.

Key points

1. ASHRAE's recommended humidity range moved from 40-55% RH (2004) to a 15.8 °F dew point up to 60% RH, capped at a 59 °F dew point.
2. The 2004 edition recommended 68-77 °F and 40-55% RH; the 2008 edition recommended 64.4-80.6 °F with 41.9 °F dew point to 59 °F dew point and 60% RH.
3. Class A1 equipment is known to function up to 90 °F and A4 up to 113 °F, although all classes share the same recommended range.
4. Dew point is nearly uniform across the room when coils run dry and no moisture is added, so it is the reliable control variable; air at 80 °F and 50% RH holds almost twice the moisture of 60 °F air at 50% RH.
5. Recommended control: dehumidify only above about 59 °F dew point and humidify only below about 16 °F, leaving a wide dead band where no energy or water is used.
6. Use at least three temperature/RH sensor pairs, convert to dew point, and control on the average of pairs known to be in calibration, never on a single pair.
7. Suggested DOAS dehumidification throttling range is 4 °F around the 59 °F set point: 100% response at 61 °F dew point and 0% at 57 °F.
8. A 1 MW heat load rejected by evaporation consumes about 3,500 lb/h (421 gal/h) of water in free-cooling mode and about 4,400 lb/h with chillers running; blowdown adds 10-20%.
9. The owner must choose between free cooling with outside air and a substantial moisture load, or minimal moisture load with no outside-air free cooling.

Using it in design

• Write the humidity strategy around room dew point: set a wide dead band (about 16 °F to 59 °F dew point unless the IT owner agrees a higher lower limit) rather than a narrow RH band.
• Size a small DOAS to carry all humidification and dehumidification, and select CRAH/CRAC coils for supply temperatures high enough that they never dehumidify.
• Specify at least three combined temperature and RH sensor pairs in the same air stream, with the controller averaging calibrated pairs, and check them against each other periodically.
• Choose adiabatic humidification (wetted media, ultrasonic or spray) over boilers or infrared units, and budget for deionized water where ultrasonic or spray systems are used.
• Decide early between airside economizing and minimum outside air with waterside economizing, and compare site water against the embedded water of grid electricity before rejecting evaporative cooling.
• Agree the lower humidity limit and any ESD mitigation with the IT procurement team before sizing humidifiers.

Who should read it

Mechanical designers and operators who specify humidification, dehumidification and sensor layouts, and owners deciding between airside and waterside economizing.

Document contents

1. Executive Summary
2. Introduction
3. Why Data Centers Have Been Humidity Controlled
4. How Data Centers Are Humidified
5. Current Humidification Issues
6. ASHRAE Guidelines
7. Control Strategies
8. To Maintain a Minimum Moisture Level in a Data Center
9. To Provide Free Cooling
10. Use Dew Point Control
11. Benefits of a Wider Humidity Range
12. Summary
13. References
14. Abbreviations

Publication details

Authors

Vali Sorell; Magnus Herrlin

Publisher

Lawrence Berkeley National Laboratory

Year

2017

Document language

English

Source type

Government / public

Access

Citation only

Licence

Licence conditions not established — confirm the conditions with the publisher

Licence evidence

datacenters.lbl.gov

Citation