Microsoft is a partner in the work, along with universities and national labs around the country.
The C3.ai Digital Transformation Institute announced on Thursday $4.4 million in awards to researchers using everything from offline reinforcement learning and large-scale simulations to soil samples and seaweed to find solutions to climate change. Cross-disciplinary teams from universities and research groups want to use artificial intelligence to make electricity grids more resilient, improve wildfire forecasting and develop workable options for carbon sequestration.
The institute put out its first call for papers last spring to identify ways to use AI to address COVID-19. This year the focus was climate change.
Eric Horvitz, Microsoft’s chief scientific officer, is a board member at the institute and participated in a press call about the grant winners. He said he was impressed by the proposals that take on some of the hardest problems the world faces.
“We need these kinds of audacious yet technically and scientifically sound projects and these projects need to be pursued by teams known for deep technical thinking and creativity,” he said.
Ali Hortacsu, an economics professor at the University of Chicago, is using supply and demand data from the electricity grid to build a model and understand how the grid responds during events such as the winter storm that hit Texas in February. Hortacsu’s goal is to determine how investments in the grid could reduce similar failures in the future.
Zido Kolter, a computer science professor at Carnegie Mellon University, is also focused on the power grid but his area of expertise is offline reinforcement learning. He will use the funding to build a simulation that can incorporate strict safety constraints. Kolter and Sergey Levine, University of California, Berkeley, are the lead investigators on this project.
Claire Tomlin’s goal is to build smart seaweed growing structures that can move through the ocean independently and use ocean currents as a power source.
“We want them to stay in nutrient rich areas, stay away from ships and use the least amount of energy,” she said.
Seaweed absorbs carbon dioxide and could help with carbon sequestration efforts. Tomlin is an electrical engineering and computer science professor at University of California, Berkeley.
During a press call about the awards, an audience member asked why the Institute was taking an open source approach with the technology developed by the research groups.
Tom Siebel, chair of the institute and chairman and CEO of C3.ai, said that to get participation from university researchers, the effort couldn’t be about helping a company make money; the effort had to be about helping society.
“All of the science goes into the public domain, and it was always set up to be that way,” he said. “It’s the reason for active participation of some of the most brilliant minds in the world.”
The institute issued a call for proposals in February and received 52 submissions. The reviewers awarded 21 grants to research proposals for improving resilience, sustainability and efficiency through such measures as carbon sequestration, carbon markets, hydrocarbon production, distributed renewables and cybersecurity.
The institute awarded a total of $4.4 million in cash to researchers. Research teams also will gain access to up to $2 million in Azure cloud computing resources, up to 800,000 supercomputing node hours on the Blue Waters petascale supercomputer at the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, up to 25 million computing hours on supercomputers at Lawrence Berkeley National Laboratory’s National Energy Research Scientific Computing Center, and unlimited access to the C3.ai Suite.
Researchers got $100,000 to $250,000 for each project for one year of operations. The list of winners is listed below by project title, principal investigator and affiliation:
These projects apply AI, machine learning and advanced analytics to support sustainability initiatives for energy consumption and greenhouse gas emissions:
- Learning in Routing Games for Sustainable Electromobility (Henrik Sandberg, KTH Royal Institute of Technology)
- AI-Driven Materials Discovery Framework for Energy-Efficient and Sustainable Electrochemical Separations (Xiao Su, University of Illinois Urbana-Champaign)
AI for Carbon Sequestration
These efforts apply AI/machine learning techniques to increase the scale and reduce costs of carbon sequestration:
- Optimization of Agricultural Management for Soil Carbon Sequestration Using Deep Reinforcement Learning and Large-Scale Simulations (Naira Hovakimyan, University of Illinois at Urbana-Champaign)
- Affordable Gigaton-Scale Carbon Sequestration: Navigating Autonomous Seaweed Growth Platforms by Leveraging Complex Ocean Currents and Machine Learning (Claire Tomlin, University of California, Berkeley)
AI for Advanced Energy and Carbon Markets
The goal of this work is to enable dynamic, automated, and realtime pricing of energy-generation sources:
- Quantifying Carbon Credit Over the U.S. Midwestern Cropland Using AI-Based Data-Model Fusion (Kaiyu Guan, University of Illinois at Urbana-Champaign)
- The Role of Interconnectivity and Strategic Behavior in Electric Power System Reliability (Ali Hortacsu, University of Chicago)
Cybersecurity of Power and Energy Infrastructure
These projects use AI/ML techniques to improve the cybersecurity of critical power and energy assets, along with smart connected factories and homes:
- Private Cyber-Secure Data-Driven Control of Distributed Energy Resources (Subhonmesh Bose, University of Illinois at Urbana-Champaign)
- Cyberattacks and Anomalies for Power Systems: Defense Mechanism and Grid Fortification via Machine Learning Techniques (Javad Lavaei, University of California, Berkeley)
- A Joint ML+Physics-Driven Approach for Cyber-Attack Resilience in Grid Energy Management (Amritanshu Pandey, Carnegie Mellon University)
Smart Grid Analytics
These researchers are applying AI and other analytic approaches to improve the efficiency and effectiveness of grid transmission and distribution operations:
- Scalable Data-Driven Voltage Control of Ultra-Large-Scale Power Networks (Alejandro Dominguez-Garcia, University of Illinois at Urbana-Champaign)
- Offline Reinforcement Learning for Energy-Efficient Power Grids (Sergey Levine, University of California, Berkeley)
Distributed Energy Resource Management
This work applies AI to increase the penetration and use of distributed renewables:
- Machine Learning for Power Electronics-Enabled Power Systems: A Unified ML Platform for Power Electronics, Power Systems, and Data Science (Minjie Chen, Princeton University)
- Sharing Mobile Energy Storage: Platforms and Learning Algorithms (Kameshwar Poolla, University of California, Berkeley)
- Data-Driven Control and Coordination of Smart Converters for Sustainable Power System Using Deep Reinforcement Learning (Qianwen Xu, KTH Royal Institute of Technology)
AI for Improved Natural Catastrophe Risk Assessment
These projects apply AI to improve modeling of natural catastrophe risks from future weather-related events (e.g., tropical storms, wildfires and floods):
- AI for Natural Catastrophes: Tropical Cyclone Modeling and Enabling the Resilience Paradigm (Arindam Banerjee, University of Illinois at UrbanaChampaign)
- Multi-Scale Analysis for Improved Risk Assessment of Wildfires Facilitated by Data and Computation (Marta Gonzalez, University of California, Berkeley)
Resilient Energy Systems
This work addresses how the use of AI/ML techniques and markets for energy and carbon introduce new vulnerabilities:
- A Learning-Based Influence Model Approach to Cascading Failure Prediction (Eytan Modiano, Massachusetts Institute of Technology)
- Reinforcement Learning for a Resilient Electric Power System (Alberto Sangiovanni-Vincentelli, University of California, Berkeley)
AI for Improved Climate Change Modeling
These projects use AI/ML to address climate change modeling and adaptation:
- Machine Learning to Reduce Uncertainty in the Effects of Fires on Climate (Hamish Gordon, Carnegie Mellon University)
- AI-Based Prediction of Urban Climate and Its Impact on Built Environments (Wei Liu, KTH Royal Institute of Technology)
- Interpretable Machine Learning Models to Improve Forecasting of ExtremeWeather-Causing Tropical Monster Storms (Da Yang, Lawrence Berkeley National Laboratory)
The C3.ai Digital Transformation Institute is a research group focused on accelerating the benefits of artificial intelligence for business, government and society. The Institute works with scientists to conduct research and train practitioners in digital transformation, which “operates at the intersection of artificial intelligence, machine learning, cloud computing, internet of things, big data analytics, organizational behavior, public policy and ethics.”
C3.ai DTI was founded in March 2020 by C3.ai, Microsoft Corporation, University of Illinois at Urbana-Champaign, University of California, Berkeley, Carnegie Mellon University, Lawrence Berkeley National Laboratory, Massachusetts Institute of Technology, National Center for Supercomputing Applications, Princeton University, and University of Chicago. The Institute is jointly managed and hosted by University of California, Berkeley and University of Illinois at Urbana-Champaign.
C3.ai is a consulting company that specializes in enterprise AI and unifying data, deploying models and deploying applications.