The Future of Data Center Energy Use: Managing Complexity with AI
It’s one of the great paradoxes of our time: the world needs to build energy-intensive data centers to meet growing demand. Meanwhile, it’s become imperative to reduce carbon emissions to curb climate change. It is a challenge and an opportunity to introduce technologies that could transform energy use in buildings forever, across all sectors, including […]
It’s one of the great paradoxes of our time: the world needs to build energy-intensive data centers to meet growing demand. Meanwhile, it’s become imperative to reduce carbon emissions to curb climate change. It is a challenge and an opportunity to introduce technologies that could transform energy use in buildings forever, across all sectors, including machine learning and artificial intelligence.
Innovative HVAC approaches, including hyper-efficient chillers and free-cooling systems, are key considerations for improving data center energy usage efficiency (PUE). Regardless of the systems installed, precise control achieved through data center infrastructure management (DCIM) is critical to maximizing efficiency benefits.
But is DCIM’s potential to reduce energy consumption hampered by our human limitations? Tradition, processes, and assumptions are holding back energy efficiency improvements? In today’s typical data center, engineers optimize system performance based on what they know about the equipment and their experience or assumptions about how systems respond under specific conditions of time and place, including weather, occupancy, and utilization. But even if the best engineers make the programming decisions, human intelligence has its limits. Machine learning and artificial intelligence could be the deciding factor that brings us the next generation of energy efficiency in an industry determined to squeeze every kilowatt. AI has an infinite capacity to master system complexity and manipulate complex cooling system sequences to optimize energy usage in infinitely dynamic real-time conditions. You can instantly analyze hundreds of what-if scenarios to arrive at the best action. Sometimes AI-based decisions seem to outpace human logic.
Early adopters manage any risk by controlling the AI and reviewing the system’s decisions before implementing them. During a pilot program, the AI advised facility operators to run the chiller in a situation where they would normally use freecooling. Curious traders willing to give the AI the benefit of the doubt followed the advice. The engineers were then surprised to see that, under those specific conditions, it took less energy to run the chiller than it did to run the freecooling pumps.
We are now starting to see what this means at some test sites. These are important and interesting days, where experts representing various sectors of construction and technology work together to find solutions that can align environmental and IT goals to create a digital and climate-safe world.
Was this article useful to you?
0 Feedbacks