Autotune isn’t just for music anymore.
Researchers from the Oak Ridge National Laboratory (ORNL) have developed a new building energy modeling tool that automates much of the collection and adjustment of building parameters, cutting down the time investment and expertise needed. Tentatively named after the popular note correction practice in the music industry, Autotune accurately estimates energy use and evaluates potential energy efficiency improvements while remaining relatively hands-off for users.
Its power lies in streamlining data collection. With traditional energy modeling, experts must specify various building parameters – up to 3,000 of them – in order to manually generate a model, a notoriously time-consuming practice. The tax rebates and utility incentives that require energy modeling often require an error rate below 30% to calibrate building models to monthly utility bills.
However, Autotune uses about 150 of the most important parameters and couples them with machine learning algorithms, allowing it to learn from the optimization paths it recommends and determine which ones are successful. This allows it to avoid strategies that haven’t worked for later projects when similar parameters are introduced. The software has routinely calibrated test models to error rates below 1%.
“We had to use supercomputing resources to create all the metadata used to train the software, but the version that will be available to the public doesn’t require all of these high-performance resources,” notes researcher Jibonananda Sanyal. “We commonly run the software on a laptop.”
The application is currently in beta testing. The team anticipates a public release in September 2015.