GSA Completes Study Of Machine Learning In Federal Buildings, Sees It As 'The Future'

Federal real estate officials believe they have found a technology that "makes the human better," and they are looking to implement it in large buildings across the government's portfolio.

The General Services Administration found it could save millions of dollars in energy costs by implementing smart systems that control energy use in real time at certain sites, according to the results of a study backed by the Department of Energy and shared first with Bisnow.

"This is probably the best implementation of machine learning we’ve seen," Kevin Powell, director of the GSA's Center for Emerging Building Technologies, told Bisnow. "It really does point to the future of how you can lighten the load, so to speak."

The study, which began shortly before the pandemic and finished this summer, implemented Prescriptive Data's Nantum OS technology across four GSA properties that represented a range of building types under the federal government's control.

The report's authors, who are all researchers from the National Renewable Energy Laboratory, found that the energy savings across a tranche of its largest buildings — about 500 buildings over 100K SF — could net the GSA millions of dollars if Nantum operates as well as it did at the four test sites.

"We absolutely achieved and demonstrated success," Powell said. 

The GSA's Proving Ground program solicited proposals for cutting-edge technologies that claimed to assist in managing energy load for commercial buildings and make them more responsive to the electrical grid. Nantum was among four technologies that the GSA accepted for use in studies in 2020.

Ultimately, four buildings were chosen to fully implement Nantum OS: the federal courthouse in Austin, Texas; the Dallas Terminal Annex; the Harvey Wiley Federal Building in College Park, Maryland; and the Bureau of Alcohol, Tobacco, Firearms and Explosives headquarters in Washington, D.C.

The results showed that while Nantum provided more than 10% energy savings at one of the buildings that has high energy usage, it was unable to surpass that threshold across all buildings. 

The GSA also recommended implementing Nantum's "unified user interface," a dashboard showing all connected building systems at a glance, throughout its portfolio, saying it "was universally appreciated by staff who said it made their jobs easier."

"I think every single person who experienced this from a user satisfaction point of view felt this was very helpful in their day-to-day operation of buildings," Powell said.

Prescriptive Data was spun out of an in-house initiative from Rudin Management Co. sponsored by the state of New York in 2016 after achieving more than 30% energy savings in some buildings.

The firm has signed deals with Blackrock and Brookfield Properties in the years since, and in April, it committed to bringing Nantum OS software to 30M SF of JPMorgan Chase's real estate holdings. 

But Sonu Panda, the firm's CEO, said getting selected for and achieving positive results through a study with the federal government was the most rigorous test Nantum has faced so far.

"The fact that the federal government has these demonstration programs that are, no pun intended, designed to illuminate what I would call lighthouse successes for the sake of other people's learning was the kind of thing that made me think that we simply could not pass up the opportunity," Panda said.

Nantum's system got high marks from building operations professionals, who appreciated the software's dashboard of building systems.

The four buildings that fully implemented Nantum's smart systems achieved between 5% and 11% energy savings. The GSA estimates that if it is able to achieve just 5% annual cost savings on its energy usage by implementing the technology, it could save money at 90 of the 504 buildings it anticipates would benefit most from an upgrade.

Those savings would gross the GSA about $4.5M annually. Powell said that is the floor for total savings, and buildings with older energy systems or higher energy costs would see a heightened benefit from implementing smart energy systems.

"This is a technology that I would say works pretty broadly," Powell said. "Obviously, if you have high energy costs, this is gonna have great payback."

The GSA is also responsible for a broad portfolio of smaller buildings that fall outside the scope of the study. Powell said those smaller buildings, especially ones smaller than 50K SF, tend to have fewer controls available for heating, cooling and other building systems.

But as the technology develops, and especially as the need for grid-responsive buildings increases, Powell said the GSA could take a look at how to smarten up its whole portfolio.

"Once we pick the low-hanging fruit of these large buildings and making the big impact on them and there are fewer pieces of large fruit, we really would like to get to the many pieces of smaller fruit that are higher up in the tree," Powell said. 

The federal government has to go through a procurement process if it wants to expand the range of buildings using energy-saving technology. Because the study has ended, it has also had to turn off Nantum at the four test buildings.

Powell said that since Nantum was shut off, the energy savings that those buildings were enjoying had been largely erased.

"We’ve seen actually degradation of performance," Powell said. "The improved performance that was delivered while the system was operational ended when the system was no longer operational, which again, I think is a testimony to what it delivers."