Using CFD to Evaluate Data Center Chiller Performance in Challenging Environmental Conditions

This webinar covers modeling the outdoor heat rejection equipment of a data center: air-cooled chillers and the generator sets placed near them. It is aimed at designers who must show that a rooftop or yard layout still delivers its rated cooling in hot, windy conditions.

What the webinar covers

The presenter starts with how an air-cooled chiller works: a vapor-compression cycle in which a compressor lifts the refrigerant pressure so heat can be rejected to outdoor air. The higher the outdoor air temperature, the more lift is required and the less capacity the chiller delivers. The capacity on a spec sheet assumes a good day with no wind and no recirculation.

The modeling section shows how the demonstration represents a chiller in CoolSim. Each unit is a block with inlets and outlets, with the heat load, number of fans and airflow taken from the spec sheet. A louvered screen is modeled as a partly open baffle, and round fans are replaced by square openings of equal area to keep the mesh simple. Units are grouped and cloned to lay out a bank quickly. The bounding box, ambient temperature, altitude, and wind speed and direction (taken from a wind rose) complete the setup.

Results are shown in the browser report: pathlines, inlet temperature plots and pressure slices. In the first example, exhaust from upwind units is pulled into downwind units by low pressure under the equipment, and downwind inlets run several degrees warmer than upwind ones. The second example is a larger layout of chillers with generator sets, chimneys on the chillers and a screen, where generator exhaust reaches some chiller intakes. The last part answers questions on model size, modeling generator sets, and validation.

Key points

• Chiller capacity falls as inlet air temperature rises. The presenter uses a rule of thumb of roughly 1 percent of capacity per degree F.
• The average inlet temperature at the midplane of the unit is the number to read from the model, for each wind direction and ambient temperature.
• Low pressure beneath the units, caused by their own fans, drives recirculation, and wind over a screen or parapet can worsen it.
• Generator exhaust is far hotter than chiller exhaust. It matters whether it reaches the chiller intakes.
• Chimneys and horizontal surfaces between units are options for reducing recirculation, and the model can compare them before anything is built.
• The presenter recommends modeling the interior first, then using its total heat load to drive a separate exterior model.
• The presenter advises validating against measurements wherever the site exists, and describes computational fluid dynamics (CFD) as guidance for engineering judgment, not a forensic tool.