The rapid growth of AI has swiftly accelerated the demand for high-performance data centers. Designing an environment that can stay ahead of this high-functioning data processing requires a reliable infrastructure that maximizes uptime while minimizing high-intensity energy demands. While traditional methods for maintaining uptime have largely relied on heavy electric loads and water usage, today’s evolving energy grid constraints, conservation requirements, and worldwide decarbonization targets are requiring technology leaders to prioritize sustainability as much as building performance.
Across industries, data centers are one of the most energy-intensive types of buildings, consuming 10 to 50 times the energy per floor space of standard commercial office buildings and collectively using about 2% of the nation’s total electricity consumption. HVAC equipment is responsible for as much as 40% of electricity use within data centers, making building system optimization critical to achieving net-zero sustainability.
The latest innovations in HVAC and smart building technology make this outcome possible by cutting energy and water use to reduce carbon emissions and operational costs. For example, purpose-built systems comprised of mission critical air-handling units paired with air-cooled, magnetic bearing chillers, digital solutions and building automation technologies designed uniquely for data centers can significantly improve sustainability while maintaining an environment that maximizes reliability and uptime.
Purpose-built technology for an evolving industry
Historically, data centers have used HVAC equipment designed for comfort cooling and did not account for the unique dynamics present in data center environments. For example, chilled water setpoints within comfort cooling are typically around 44°F. However, innovations in chillers designed for data center applications make it possible for chilled water setpoints to operate from 70° to 80° and sometimes even higher, as server manufacturers have become more comfortable with processors and motherboards operating at higher temperatures.
Designed specifically for hyperscale and colocation data centers, air-cooled, magnetic bearing centrifugal chillers, for example, are optimized for increased temperatures inside the white space and the lifts that are prevalent in the data center design today. They can deliver chilled water temperatures that are upwards of 80° and cater to a low lift, resulting in greater energy efficiency.
While most data centers use air-cooled chillers that have free cooling coils to benefit from lower ambient conditions, air-cooled, magnetic bearing centrifugal chillers can operate at inverted conditions and provide free cooling without the need for additional coils. Free cooling coils that are added to the condenser of the chiller can create inefficiencies and lead to additional pressure drops.
At the same time, using a chiller that is lighter in weight and provides inverted-operation free cooling reduces the carbon footprint of the building itself in many dimensions. Eliminating the need for free cooling coils minimizes the equipment’s overall embodied carbon and reduces shipping and rigging weight. It can also eliminate the need for additional building structure that inherently has more steel in it to support the mass of higher-weight cooling equipment.
A friction-free, magnetic drive benefits uptime, as well. If power is interrupted, a typical chiller can take up to 10 minutes to restart. In comparison, some advanced magnetic bearing centrifugal chillers feature “quick start” technology that enables a much faster compressor restart time, achieving a return to full load in as few as three minutes after power is restored. Because air-cooled, magnetic bearing centrifugal chillers use a variable-speed drive, there is no inrush current. This means a fast, controlled return to full capacity and set point.
To further improve data center sustainability, air-cooled, magnetic bearing chillers produce notably less sound than many screw chillers, and some use R-1234ze, a refrigerant with ultralow global warming potential (GWP).
When connected to an AI-based solution, air-cooled magnetic bearing centrifugal chillers combined with high-efficiency mission critical computer room air handlers designed with electronically commutated motors (ECM) can match cold aisle temperature with the real-time load and optimize energy use from moment to moment. Having a dynamic chilled water set point and cold aisle temperature optimizes energy use without risk to data center uptime.
Establishing dynamic cold aisle temperatures
The temperature of the cold aisle determines how aggressively HVAC equipment and server fans must work and, therefore, how much power they consume to ensure the proper volume of air moves through servers to remove heat. A higher cold aisle temperature results in lower chiller power consumption. A lower cold aisle temperature results in a smaller volume of airflow being needed and less fan power consumption for the air-handling unit fans and server fans.
To prevent hot spots in a data center’s white space and ensure uptime, data center cooling strategies have historically favored lower cold aisle temperature and higher airflows. Oftentimes, cold aisle temperatures are cooled beyond what’s needed, serving as a form of airflow insurance when an application is close to the edge of requirement. Although this method can prevent overheating in some instances, it is most often wasted energy. The latest generation of servers can operate at high ambient conditions, enabling warmer cold aisle temperatures.
By optimizing the cold aisle temperature, a data center can consume the minimum power required to cool it. Currently, this temperature is a static number. However, recent research proves that a dynamic cold aisle temperature can deliver optimum cooling according to the ambient conditions in the data center load at any given time — and the technology to do it is available now.
For instance, when it’s very cold outside, there is an opportunity to use economization, or free cooling, to simultaneously lower the chilled water set point and lower the cold aisle temperatures. This reduces airflow and power consumption from the computer room air handler as well as server fans. However, when a data center uses a static chilled water set point and a static cold aisle temperature, this opportunity is lost.
On the hottest afternoons of the year, the chiller power consumption is highest because the lift on the chiller is high. Chiller lift refers to the difference in pressure between the refrigerant in the condenser and the refrigerant in the evaporator. At higher lifts, the compressor consumes higher amounts of power to drive the thermodynamic cycle. The lift may be reduced by raising the chilled water set point and the cold aisle temperature for a few hours in the afternoon. This reduces the power consumption by the chiller compressor(s). The industry uses the term temporary excursion from ASHRAE Thermal Guidelines for Air Cooling of IT Equipment.
The pairing of an air-cooled, magnetic bearing centrifugal chiller and mission critical computer room air handlers with an open source digital platform and building automation system can drive cold aisle temperatures that suit data center loads at any given moment. A dynamic chilled water set point, and a dynamic cold aisle temperature overall, helps optimize a data center’s power consumption without compromising the uptime of the data center. This method of continuous optimization could lead to the best real-time energy efficiency of the data center while providing cold aisle temperatures that help maintain uptime.
Data-driven energy use
As part of a digital platform, an AI-based solution can be either an advisory or a supervisory function sitting on top of the building management system (BMS). In this capacity, it helps ensure data center personnel are able to properly evaluate the real-time data center loading and real-time data center requirements around the ambient conditions as well as understand historical loading patterns or trends. Equipped with this valuable information, facility managers can be certain the system is operating as efficiently as possible.
Data-driven technology can inform a chilled water reset strategy to help reduce energy use during peak demand periods in data centers that experience high ambient temperatures. A chiller’s power consumption depends on lift, and lower lift means less energy use. For example, the chilled water setpoint can be adjusted to a higher temperature for four or five hours in the afternoon when demand is high to improve system energy efficiency in conjunction with a slight ramp-up of the high-efficiency ECM fans in the computer room air handlers.
This chilled water reset deviates from standard conditions and isn’t permitted by some service-level agreements. To improve overall efficiency and data center sustainability, it’s important to include chilled water reset for a set number of hours per year in service-level agreements.
Using historical trends, an intelligent cooling system can anticipate and prepare for the next loading change. For example, if a data center consistently generates a lot of heat around 8 a.m., the system can be automated to gradually ramp up capacity starting at 7 a.m. rather than running at 100% capacity at 7:59 a.m. This gradual ramp-up minimizes system spikes, improves energy efficiency, and can even extend equipment life.
Intelligent digital services, like that provided by an AI-based solution, connected equipment, and building automation technology can make data centers smarter and more sustainable. These solutions allow facility managers to continuously monitor equipment health and energy consumption in real time while automating key processes. Some solutions also offer easy-to-read dashboards that display trends and notify assigned personnel when set parameters deviate from assigned values. That allows facility teams to address issues, identify opportunities for energy savings, and drive outcomes that matter most.
Improving energy efficiency today and in the future
Data center regulations are advancing as rapidly as big data itself. The industry is facing strict sustainability goals and aggressive deadlines to reach them. Meeting these outcomes can result in improved uptime, reliability, and ROI. Falling short can result in regulatory incompliance and lost investor recognition.
Purposefully designed HVAC systems can improve energy efficiency, reduce water usage, and enhance data center performance. Air-cooled, magnetic bearing centrifugal chillers combined with mission critical computer room air handlers driven by artificial intelligence can optimize HVAC performance by operating with higher chilled water set points and higher cold aisle temperatures informed by real-world, white space conditions. This shift in data center building design presents an opportunity to create smarter, more sustainable data center infrastructure that can grow and evolve based on data-driven performance to yield continuous improvements in sustainability, both today and in the future.