Many building owners today are looking at green retrofitting options to lower energy demand and costs. From single-family homes to skyscrapers, upgrades such as installing more efficient HVAC systems, installing new windows or green roofs, or switching to low-energy light bulbs can lower utility bills and, in many cases, eventually pay for themselves. A whole industry of energy consultants now helps owners plan for their retrofit, and scientists have created computer models to estimate a building’s energy demands and savings under different scenarios.
But an even more direct way to reduce the energy consumption of a city is to make new buildings more efficient before they are even built. On March 20th at the CI, Argonne researcher Leah Guzowski presented work on a new energy demand model, called EECALC, that she hopes will be equally useful for today’s buildings and the buildings of tomorrow. As part of the Urban Center for Computation and Data’s LakeSim platform, the model will help plan for the massive, 600-acre Lakeside Development planned for the South Side of Chicago.
“Lakeside is basically building a whole new Loop just ten miles south of the current Loop,” said Guzowski, who leads the LakeSim project with UrbanCCD director and CI senior fellow Charlie Catlett. “Developers are comfortable at the 30 to 50 acre scale, but at 600 acres, the current computational tools don’t scale up.”
Guzowski hopes to address those issues of scale with EECalc, an energy calculator developed by Argonne and the Georgia Institute of Technology. The motivation for EECalc was to build a faster, scalable model that worked with more data, but required less human labor to run. Researchers also wanted to add uncertainty to the model’s forecasts, since hard-to-predict factors such as weather, equipment faults and degradation, energy prices, and simple human error make it misleading to put an exact number on energy savings forecasts.
In one important step for scalability, the researchers re-examined the different kinds of data typically used by energy models to determine the most important information to collect. From an original pool of 45 inputs, the team identified 10 high-priority measures, including factors such as equipment plug loads, lighting power density, HVAC infiltration, and cooling.
The team then tested the model on a mixed-use concrete high-rise in the Chicago Loop, with 28 floors and 350,000 square feet of space. Using Bayesian methods, the model was calibrated until it came close to predicting the building’s utility bills each month, then run with a test case where the building replaced a 60% efficient boiler with a 90% efficient model. The output, factoring in uncertainties, provided a range of possible savings that could be used by the building manager to convince ownership that the near-term expense would produce long-term savings, Guzowski said.
While these early tests of the model on its own are promising, the ultimate plan is to incorporate EECalc alongside models for transportation, wastewater management, weather, and other sectors into a unified platform for LakeSim. Currently, the researchers are using software called CityEngine to conduct proof-of-concept tests for the Lakeside developers and designers. With EECalc built in, users can now can see energy consumption of different zones and different building designs, Guzowski said, though more work is needed to get accurate predictions at a neighborhood or city scale.
Ultimately, the hope is not just to assist the Lakeside designers in making their huge new development as energy-efficient as possible, but to develop a platform that can be used around the world, particularly in countries such as China and India that are rapidly urbanizing.
“China plans to move 100 million people into cities by 2020, that’s a massive urbanization in a very short period of time,” Guzowski said. “We know Lakeside is a great opportunity, but we think the platform is most useful for other parts of the world.”
You can see Guzowski talk more about LakeSim in the video for last year's City of Big Data event.
[LakeSim image by Bo Rodda]