Wrangler Supercomputer Speeds through Big Data

Data-intensive supercomputer brings new users to high performance computing for science

Published on March 10, 2016 by Jorge Salazar


Save Money, Save Energy with supercomputers


Computer scientist Joshua New of Oak Ridge National Laboratory optimizes buildings to save energy

Saving energy saves money.

Scientists at Oak Ridge National Laboratory (ORNL) are using supercomputers to do just that by making virtual versions of millions of buildings in the U.S. The Wrangler data-intensive supercomputer is working jointly with ORNL's Titan in a project called Autotune that trims the energy bills of buildings.

Jibonananda Sanyal and Joshua New

Oak Ridge National Laboratory researchers Jibonananda Sanyal (left) and Joshua New (right) developed Automated calibration software on the Titan supercomputer for building efficiency studies. Credit: Oak Ridge National Laboratory.

Computer scientist Joshua New of the ORNL Building Technology Research and Integration Center is the principal investigator of the Autotune project, funded by the U.S. Department of Energy. Autotune takes a simple software model of a building's energy use and optimizes it to match reality.

"What we're trying to do is create a crude model from publicly available data," New said. "Then the Autotune project takes utility bill data, whether it's monthly electrical utility bills, or hourly bills from advanced metering infrastructure, and calibrates that software model to match measured data." New said that once Autotune calibrates the model well enough, it can be legally used in multiple ways including for optimal building retrofit packages.

Autotune optimizes or 'tunes' the simulation engine called EnergyPlus. "It can tell you practically anything you want to know about a building," New said. "That's as long as you have a software description of the building in the first place for EnergyPlus to run on."

The main hurdle in doing so is that there's over 3,000 'knobs' to adjust in the software model — its "parameter space" as New called it — to match 12 data points from monthly utility bills.

"That's what we're using supercomputers for," New said. "We're sampling the parametric space of inputs to figure out quantitatively how sensitive certain knobs are for affecting energy consumption for electricity, for natural gas, and for any other sensor data that we could collect or report from the simulation engine — and then use that to inform the calibration process so that it can create a model that matches the way the building works in the real world."

New uses the second fastest supercomputer in the world, Titan at ORNL, to do his large-scale parametric studies. "We now have the ability to run 500,000 simulations and write 45 TB of data to disk in 68 minutes. That's fairly large data by most people's standards," New said. He added that he hopes to scale out Autotune to run simulations for all 125.1 million commercial and residential buildings in the U.S. And he said it would take about two weeks of Titan running nonstop 24/7 to do it. "Right now we've run only eight million simulations. And we have 270 TB of data that we're combing through," New said.

Combing through all that data for material to strengthen the virtual building is right up the alley of the Wrangler data-intensive computer.

"Wrangler fills a specific niche for us in that we're turning our analysis into an end-to-end workflow, where we define what parameters we want to vary," New said. "It creates the sampling matrix. It creates the input files. It does the computationally challenging task of running all the simulations in parallel. It creates the output. Then we run our artificial intelligence and statistic techniques to analyze that data on the back end. Doing that from beginning to end as a solid workflow on Wrangler is something that we're very excited about. Wrangler has enough horsepower that we can run some very large studies and get meaningful results in a single run."

Autotune Diagram

Autotune is a set of automated calibration techniques for tuning residential and commercial building energy efficiency software models (Energy Plus) to match measured data. Credit: Joshua New, ORNL.

The main customer segment for Autotune has been energy service companies. EnergyPlus helps them more quickly create software descriptions of buildings, according to New.

"If you have a software description of a building, it would be a very quick analysis to figure out of the 3,000 plus things that you could do to your building, which would make it the most energy efficient, save you the most money, and give you the most bang for the buck," New said.

For example, some of those changes could be changing out the HVAC, adding insulation, changing windows, and sealing ducts.

Another use case is national policy, New said. Local, state, and federal governments are considering new energy-saving building technology that might not be immediately cost-effective, but they have incentive structures that can pay for part of it.

"Where do you set those values?" New asked. "If the government subsidizes technologies that don't work, then they're wasting taxpayer money. How do they set those values in an informed fashion so that we get the best energy savings?"

The ultimate goal, New said, is to bring down the energy bill of the U.S. "We're very excited to be helping industries make our country more energy efficient, increase our energy security and have a better and brighter future through enabling core technologies like the supercomputing center here at ORNL," New said.


Continue reading Wrangler Special Report >>