Within the next five years, the AE firm SmithGroup wants to be able to incorporate the highest level of energy efficiency into every project it designs and builds. But the challenge is selecting the right energy model from literally thousands of options.
“The number of buildings we need to touch, and the pace we need to do it, exceed what an individual could do in a lifetime,” says Stet Sanborn, engineering lead in SmithGroup’s San Francisco office.
To speed this process, SmithGroup has partnered with the U.S. Department of Energy’s Oak Ridge National Laboratory (ORNL), whose Automated Building Energy Modeling software suite, better known as AutoBEM, has simulated the energy use of 123 million structures, or 98 percent of U.S. Buildings.
AutoBEM was developed using high-performance computing to process layers of imaging data with information about each building, such as its size, use, construction materials, and HVAC technologies. The goal was to create a digital twin of the nation’s buildings, says Joshua New, ORNL’s project leader.
Sanborn tells BD+C that SmithGroup got involved with this technology collaboration project 18 months ago, after responding to the lab’s request for proposal. After six months of planning, SmithGroup provided information for every building type it handles, including climate zones, footprints, orientation, and envelope construction. The modeling also took into account lighting and HVAC variations.
This project generated more than 200,000 building iterations and 250,000 energy models. The computation, which took less than an hour to complete, was equal to the output of one employee working full time for 365 years.
Using AI to spot trends
AutoBEM’s information was used to train an artificial intelligence tool that will allow SmithGroup to pre-simulate the energy impact of every design possibility for any building.
Sanborn says that these data sets are still pretty rudimentary, so they should benefit from ORNL’s plans to update its information this year. “We’re looking for much better resolution, so we can eventually have real-time and predictive feedback,” he explains.
In 2023, SmithGroup plans to train the AI bots to spot trends and choose the best energy-efficient iteration for a given project. Sanborn says his team has already been surprised by how the data has highlighted “the interaction of things” like glass U-values and wall R-values. More refined data, he predicts, should make these data sets more accessible and actionable. (Sanborn says that SmithGroup already has pre-simulated climate data for every market it builds in.)
The first new designs for which this energy modeling is likely to be applied could be for a civic or higher education building, says Sanborn, where no restrictions to sharing information exist.
ORNL’s partners such as SmithGroup and Google—which is using AutoBEM to improve its free Environmental Insights Explorer tool—have committed to sharing data sets created by using AutoBEM. Some data sets have already been posted, and Sanborn thinks that open-source access is important because “SmithGroup can’t build every building, as much as we’d like to. We don’t want to hold a secret sauce or limit everyone’s ability to drive efficiency in response to what is really a climate emergency.”
SmithGroup participated in the funding of AutoBEM, whose development, expansion and collaborations are also funded by DOE’s Office of Electricity, the Energy Efficiency and Renewable Energy’s Building Technologies Office and the National Nuclear Security Administration. The research team leveraged supercomputing resources from Argonne National Laboratory.
Related Stories
| Jun 30, 2011
Balancing cost and energy performance in net-zero buildings
BD+C Editorial Director Robert Cassidy talks with The Weidt Group's David Eijadi, FAIA, about the cost of producing net-zero buildings.
| Jun 29, 2011
New leadership role for architects in net-zero design
BD+C Editorial Director Rob Cassidy talks with RNL Design's Tom Hootman, AIA, about the changing role of architects in net-zero designs.
