How is CFD modeling keeping data centers cool?
March 2023
Thermal management is a growing concern for our data center clients around the globe. Using digital wind tunnels in conjunction with computational fluid dynamics (CFD) can optimize the cooling performance of data centers – which in turn maximizes efficiency and sustainability.
Modern data centers equipped for data storage, high-performance computing, and telecommunication consume an immense amount of electricity for operations. Data centers account for approximately 2% of the total electricity use in the US, according to the US Department of Energy, and 1-1.5% of global electricity use, according to the International Energy Agency (IEA).
Continued demand to increase power loads and run sophisticated applications are leading to growth in heat generation that needs to be dissipated. As a result – and to be resilient and agile – data center owners are increasingly concerned with thermal management.
“The removal of heat is as crucial on the inside of a data center as it is on the outside. Thus, how equipment is cooled inside the data hall has always been a critical piece of any facility’s design,” says Jens Chr. Bennetsen, Head of Center of Excellence for Advanced Simulations at Ramboll.
But as power and heat loads rise, efficient cooling of data centers is becoming more challenging. Ensuring that cooling takes place in a sustainable and energy-efficient way requires an in-depth study of how cooling performs at multiple operational conditions. Failing to achieve the required cooling can result in decreasing the data center's capacity by as much as 30-40%, while increasing energy consumption within the cooling system.
One important tool to reveal the physics behind the cooling and the way it works is computational fluid dynamics (CFD) methods using Ramboll’s validated Digital Wind Tunnel (DWT).
CFD is revolutionizing modern data centers
CFD is particularly beneficial as it allows running virtual tests during all design stages – well before construction and commissioning start, thereby providing insights into how cooling systems will work under different environmental conditions for the complete site. This ensures that the cooling design will be able to meet demands, thus providing the capacity required and helping to reduce the total energy consumption of the whole data center facility.
As a result, the tool effectively models and analyzes a data center and before the facility is built, provides detailed design insights into how key aspects and critical equipment will perform. CFD can also be used to optimize retrofit design activities or troubleshoot challenging installations.
Today, designing larger data halls requires efficient layout and operation of the chiller banks to ensure capacity for year-round operation. The combined effect from heat ejected from the chillers and heat stemming from backup power containers, gensets, and substations can shift the thermal loads that must be considered.
Seeing the full picture with different operational and environmental conditions also allows for rethinking the cooling approach and understanding any implications early on.
Strategic decision making with a hyperscale data center company
Ramboll partnered with a hyperscale data center company to study and verify concepts for the design of a mission critical facility. The project helped ensure that full consideration was given to the airflow needed for the proposed chiller plant at peak ambient temperatures. Doing this during the design stage – instead of after – safeguarded that design improvements (i.e., fixing identified issues before ordering and commissioning) could be implemented with minimal cost and avoid costly retrofitting.
CFD offered the best tool to model the facility, proposed installation, and extreme design conditions to ensure that the equipment could operate without compromise during peak ambient conditions, considering the local environment and potential wind and building wake interactions.
“We could assess the variation of the ambient wind environment with multiple wind directions and speeds to capture the annual variation and possible impact on the facility’s operation throughout the year,” Jens Chr. Bennetsen says. “The high-fidelity simulations and subsequent investigations, analysis, and optimization that became possible were a critical part of the design stage and supported the successful delivery of the project."
"In addition, our Digital Wind Tunnel adopts a technology platform with high-end racing, like Formular 1, making it possible to do multiple design iterations in an agile way and within the project timeline," he adds.
Initial results from the first design iteration were found to be unfavorable in a small number of conditions, with recirculation possible between the intake and exhaust sections of the genset acoustic enclosures. In addition, thermal contribution from the backup power containers, gensets, and substations resulted in a temperature elevation that surpassed 46 °C (116 °F). Because of the simulation outcomes, the proposed design was changed to mitigate these issues, and subsequent simulations were produced to demonstrate the improved airflow pattern that increased the fresh air supply substantially and lowered the temperature at the chiller to 34-36 °C (93-96 °F).
“Without design optimization using evidence-based simulations like CFD, the heat rejection can compromise the operational requirement. This means the facility could potentially struggle at high ambient temperatures, resulting in lower overall efficiency and lesser data center operational capacity due to a need to scale down operations/throughput when sufficient cooling would be absent,” Jens Chr. Bennetsen says.
Summary
To ensure that data center designs are meeting today’s requirements for sustainability and lowering energy consumption, it is essential to capture the full building layout of the campus to avoid missing the full picture and to derisk the investment and operational cost. Failing to meet the design threshold can lead to a significant increase in energy use and economical losses. Growth and demand for data center capacity naturally enforce the problem at hand above handling heat rejection.
Want to know more?
Sign up to receive updates, insights, and perspectives from our experts related to data centers and mission critical facilities.
|
About the author Dr. Jens Chr. Bennetsen specializes in advanced flow dynamics, computational fluid dynamics (CFD), and mathematical and numerical modeling. He has extensive experience in advanced turbulence fluid flow modeling for steady and unsteady flows related to the built environment, civil engineering, risk and safety, industrial fluid flows and process engineering, and the oil and gas sector. |