Article first appeared in Puget Sound Business Journal
The average HVAC system in a U.S. office building accounts for about 40 percent of the overall energy bill. At Microsoft, HVAC costs are 70 to 80 percent of the company’s total energy costs due to the need to cool power sucking and heat producing data servers.
A pilot program at the 500-acre Microsoft campus in Redmond is using artificial intelligence to cut the energy consumption by 15 percent at a test chiller plant, the electrically voracious machines that cools most large office buildings and data centers.
“We are very confident that this has potential,” says Mohan Reddy Guttapalem, Microsoft’s senior facilities manager for the Puget Sound. “We see a great opportunity.”
A single large chiller plant can use several million kilowatt hours per year, costing hundreds of thousands of dollars. There are an estimated two dozen of the very large chiller plants that help cool Microsoft and other companies in the Puget Sound area.
Microsoft’s pilot program combines a cloud-based machine learning program and data analytics software and builds on the Energy Smart Buildings technology launched on the Microsoft campus in 2011. The ESB program collects real-time data from all disparate HVAC devices and sensors and presents it in a unified cloud-based platform for analysis, monitoring and repairs.
But the star of the chiller plant study is machine learning, a mushrooming field of computer science. Sometimes described as machine-driven pattern recognition, machine learning uses advanced statistical methods to make predictions given a set of conditions.
The fact that your chiller plant keeps your office cool on a hot day is no indication that it is working well.
“The problem that we’re all struggling with in this industry is that HVAC systems are at the same time both over-engineered and inefficient,” says Bert Valdman, CEO of Seattle-based Optimum Energy. “They’re designed to perform best at peak load.”
Read full article (Here):