We’re Not Saying We’re the Avengers, But...
In the 2019 blockbuster "Avengers: Endgame", it takes genius-billionaire-playboy-philanthropist Tony Stark as long to perfect time travel as it takes most people to mow their lawn. Critical to his success is his Artificial Intelligence (AI) assistant, which is able to almost instantly turn his offhand musings into workable, visually pleasing solutions.
While a time heist isn’t on ARPA-E’s agenda, we are excited to announce the project selections for our first AI/machine learning-focused program: Design Intelligence Fostering Formidable Energy Reduction (and) Enabling Novel Totally Impactful Advanced Technology Enhancements (DIFFERENTIATE).
ARPA-E and the DIFFERENTIATE team selected projects that incorporate AI and machine learning into the energy technology and/or product design processes. Their goal is to enhance the productivity of energy engineers and help them develop much-needed, next-generation energy technologies. Movie magic aside, AI and machine learning are proven to help speed up the design process for new technology.
So, you may ask yourself, how do AI and machine learning help design more efficient, lower carbon energy systems? AI makes it possible for machines to learn from experience, adjust to new inputs, and perform like humans. Machine learning is a core part of AI, and it centers around computer algorithms that improve automatically through experience.
The way these technologies help design and development is summed up in the old adage, “if at first you don’t succeed, try, try again.” It’s just that when machine learning is involved, the number of tries required to successfully train a new algorithm may be prefixed by kilo, mega, or even giga.
At its core, machine learning is an application of AI that enables a system to “learn” from experience. Machine learning algorithms take in huge amounts of data, and “learn” that particular inputs result in specific responses. The trick is to build these algorithms in such a way that they can “learn” to predict a response for an input that hasn’t been specifically programmed.
In order to organize the anticipated efforts, the DIFFERENTIATE program has adopted a simplified engineering design process framework and used it to identify several general mathematical optimization problems that are common to many engineering design processes. It then conceptualizes corresponding machine learning tools that could help engineers execute and solve these problems in a manner that would dramatically accelerate the pace of energy innovation.
ARPA-E is excited to open the next phase of energy technology design and innovation using AI and machine learning. The DIFFERENTIATE projects include:
National Renewable Energy Laboratory
End-to-End Optimization for Battery Materials and Molecules by Combining Graph Neural Networks and Reinforcement Learning
National Renewable Energy Laboratory
INTEGRATE – Inverse Network Transformations for Efficient Generation of Robust Airfoil and Turbine Enhancements
Northwestern University
Adaptive Discovery and Mixed-Variable Optimization of Next Generation Synthesizable Microelectronic Materials
Iowa State University
Context-Aware Learning for Inverse Design in Photovoltaics
Massachusetts Institute of Technology
Machine Learning Assisted Models for Understanding and Optimizing Boiling Heat Transfer on Scalable Random Surfaces
Massachusetts Institute of Technology
Global Optimization of Multicomponent Oxide Catalysts for OER/ORR
University of Michigan-Dearborn
Machine-Learning-Enhanced Automated Circuit Configuration and Evaluation of Power Converters
Carnegie Mellon University
Predicting Catalyst Surface Stability Under Reaction Conditions Using Deep Reinforcement Learning and Machine Learning Potentials
Julia Computing, Inc.
Accelerating Coupled HV AC-Building Simulation with a Neural Component
University of Maryland
Invertible Design Manifolds for Heat Transfer Surfaces (INVERT)
Los Alamos National Laboratory
Machine-Learning-Based Well Design to Enhance Unconventional Energy Production
University of Texas at Austin
Learning Optimal Aerodynamic Designs
IBM Research
Model-based Reinforcement Learning with Active Learning for Efficient Electrical Power Converter Design
Carnegie Mellon University
High-Fidelity Accelerated Design of High-Performance Electrochemical Systems
Stanford University
Energy Efficient Integrated Photonic Systems Based on Inverse Design
University of Missouri
Deep Learning Prediction of Protein Complex Structures
United Technologies Research Center
LENS: Learning Enabled Network Synthesis
United Technologies Research Center
MULTI-LEADER: MULTI-Source LEarning-Accelerated Design of High-Efficiency Multi-Stage Compressor
GE Research
IMPACT: Design of Integrated Multi-physics Producible Additive Components for Turbomachinery
GE Research
Pro-ML IDeAS: Probabilistic Machine Learning for Inverse Design of Aerodynamic Systems
Princeton University
MILSPICE: Machine Learning based SPICE Modeling Platform for Power Magnetics
Lawrence Berkeley National Laboratory
Deep Learning and Natural Language Processing for Accelerated Inverse Design of Optical Materials
Pacific Northwest National Laboratory
Machine Learning for Natural Gas to Electric Power System Design
For more information, visit the DIFFERENTIATE program page here.