Department of Energy Deep Learning

Part
01
of one
Part
01

Department of Energy Deep Learning

The Department of Energy finances and sponsors a variety of deep learning or machine learning projects. Some of these projects include the use of deep learning to interpret massive scientific data sets, a program aimed at using machine learning to increase the production of biofuel and a new method that can be applied to nuclear reactors to accelerate the speed of predictions and improve accuracy of diagnosis of instability inside the reactor.

1) Interpreting massive data sets

The US Department of Energy (DOE) has financed a $2 million machine learning project with the aim of improving the interpretation of huge scientific data sets. The program is name ASCEND which stands for Advances in Machine Learning to Improve Scientific Discovery at Exascale and Beyond. In this particular project, deep learning is used to help scientists and researchers explore the link that exists between data collected from experiments and properties of various materials. This would ultimately assist researchers in the process of designing improved materials.

The deep learning based technology in this case is a network that is able to make sense of data gathered from a huge number of inputs, and to improve its understanding by learning from the data gathered previously.

2) Increase in Biofuel production

Scientists from a DOE laboratory called the Lawrence Berkeley National Laboratory (LBNL) have used machine learning to assist them in the creation of a new technique to speed the creation of biofuel-producing microbes.

Using machine learning, the researchers have managed to find an algorithm that uses previous experiment data about proteins in a microbe that produces biofuel. Machine learning would then allow to learn how chemical reactions produce a wanted compound, using data from previous experiments.

The use of machine learning would also allow to facilitate the predictive powers of a computer algorithm by allowing it to be trained and to learn from previous operations.

The test of the technique found that the algorithm managed to teach itself about the variations observed in the concentrations of metabolites during the chemical reactions, and to predict with accuracy the amount of biofuel that could be produced.

3) Forecasting instabilities in nuclear reactors

The DOE has sponsored a project called the Fusion Recurrent Neural Network (FRNN) whose aim is to forecast dangerous instabilities in nuclear reactors, that can occur due to a disruption in tokamaks, devices that are used to capture fusion energy.

The FRNN is a new deep learning technique that can be used to accelerate the speed of predictions of potential tokamaks disruptions that can damage nuclear reactors. This breakthrough can have an important impact on nuclear reactors safety in the future, by helping to better understand complex systems.

In this particular case, deep learning has been used to train a computer to solve a complex issue with speed and accuracy, by processing various data until it finds the desired outcome.









Sources
Sources

Quotes
  • "Today, the U.S. Department of Energy’s (DOE’s) Advanced Research Projects Agency-Energy (ARPA-E) announced up to $20 million in funding to accelerate the incorporation of machine learning and artificial intelligence into energy technology and product design processes."
  • "The Design Intelligence for Formidable Energy Reduction Engendering Numerous Totally Impactful Advanced Technology Enhancements (DIFFERENTIATE) program seeks to enhance energy innovation by incorporating artificial intelligence and machine learning into energy technology development."
  • "DOE-fueled artificial intelligence is being utilized across all sectors, from strengthening cybersecurity and national security, increasing energy efficiency, optimizing grid security and resiliency, and developing innovative health solutions. "
  • "The DIFFERENTIATE program is the latest example of DOE paving the way towards the New American Energy Era."
  • "In order to organize these efforts, DIFFERENTIATE identifies six general mathematical optimization problems that are common to many design processes. It then conceptualizes several machine learning tools that could help engineers execute and solve these problems in a manner that dramatically accelerates the pace of energy innovation. "
  • "Up to $5 million of the $20 million allocated by ARPA-E under this funding opportunity will be made available specifically for small businesses under ARPA-E’s Small Business Innovation Research (SBIR) program, with the $15 million balance available to all applicants."
  • ""
Quotes
  • "The DIFFERENTIATE program seeks to enhance the pace of energy innovation by incorporating machine learning into energy technology development processes. By doing so, this program aims to enhance the productivity of energy engineers in helping them to develop next-generation energy technologies."
  • ""
Quotes
  • "PNNL is a Department of Energy Office of Science national laboratory. We perform research supporting the missions of several U.S. federal agency sponsors including the Department of Energy, the National Nuclear Security Administration, the Department of Homeland Security, the National Institutes of Health, the Department of Defense, the Nuclear Regulatory Commission, and the Environmental Protection Agency. "
Quotes
  • "Researchers from the U.S. Department of Energy’s Lawrence Berkeley National Laboratory have created a new method using machine learning to accelerate the design of microbes that produce biofuel."
  • "To speed up the production of biofuels, the scientists developed a computer algorithm that begins with abundant data about the proteins and metabolites in a biofuel-producing microbial. However, the algorithm does not contain information about how the pathway actually work and instead uses data from previous experiments to learn how the pathway will behave."
  • "By using machine learning, scientists can train a computer algorithm to make predictions of a pathway using data from related systems."
  • "The algorithm taught itself how the concentrations of metabolites in these pathways change over time and how much biofuel the pathways produce by analyzing data from two experimentally known pathways that produce small and large amounts of biofuels."
Quotes
  • "On Earth, the most widely used devices for capturing the clean and virtually limitless fusion energy that powers the sun and stars must avoid disruptions. These devices are bagel-shaped tokamaks. Massive disruptions can halt fusion reactions and potentially damage the fusion reactors. By applying deep learning—a powerful version of the machine learning form of artificial intelligence, researchers have a new code, Fusion Recurrent Neural Network (FRNN), to reliably forecast disruptive events."
  • "Such predictions are a crucial for large future reactors. Researchers can also use the code to make predictions that could open avenues for active reactor control and optimization."
  • "Support for this work comes from the Department of Energy (DOE) Computational Science Graduate Fellowship Program of the DOE Office of Science and National Nuclear Security Administration; Princeton University’s Institute for Computational Science and Engineering; and from Laboratory Directed Research and Development funds that the Princeton Plasma Physics Laboratory provides. "
Quotes
  • "In an effort to improve consumer education and access to renewable energy technologies, PowerScout reportedly uses AI to model potential savings on utility costs using industry data. The company reportedly leverages data analytics to identify “smart home improvement projects” based on the unique features and energy usage in a client’s home. PowerScout’s algorithm appears to match clients to potential hardware installation providers in an online marketplace format to ensure competitive rates. "
  • "Essentially, the AI acts as a marketplace advisor, providing recommendations to help clients make informed decisions regarding to make renewable energy technologies purchases for their homes. We imagine that this use of AI is similar to the recommendation capabilities seen in other marketplace businesses (which we’ve covered in greater depth in our recommendation engine use-case article). The development team claims the platform has collectively overseen the installation of solar capacity roughly equivalent to powering 250,000 homes as of March 2017."
Quotes
  • "Using the Titan supercomputer, a research team led by Robert Patton of the US Department of Energy’s(DOE’s) Oak Ridge National Laboratory (ORNL) has developed an evolutionary algorithm capable of generating custom neural networks that match or exceed the performance of handcrafted artificial intelligence systems. Better yet, by leveraging the GPU computing power of the Cray XK7 Titan—the leadership-class machine managed by the Oak Ridge Leadership Computing Facility, a DOE Office of Science User Facility at ORNL—these auto-generated networks can be produced quickly, in a matter of hours as opposed to the months needed using conventional methods."
  • "The research team’s algorithm, called MENNDL (Multinode Evolutionary Neural Networks for Deep Learning), is designed to evaluate, evolve, and optimize neural networks for unique datasets. Scaled across Titan’s 18,688 GPUs, MENNDL can test and train thousands of potential networks for a science problem simultaneously, eliminating poor performers and averaging high performers until an optimal network emerges. The process eliminates much of the time-intensive, trial-and-error tuning traditionally required of machine learning experts."