Priya Donti, an assistant professor at the Massachusetts Institute of Technology, is developing new artificial intelligence algorithms designed to manage and stabilize electric power grids that rely on fluctuating renewable energy sources like solar and wind. Her work aims to create more efficient and reliable systems for integrating clean energy at a large scale.
Donti's research focuses on moving beyond simple energy forecasting, which is a common application of machine learning, to actively balancing the grid in real-time. This involves creating AI models that understand the physical constraints of power systems, a critical step for ensuring grid stability as the world transitions away from fossil fuels.
Key Takeaways
- Priya Donti's research at MIT focuses on using machine learning to improve the stability and efficiency of power grids with high levels of renewable energy.
- One of her innovations is a grid optimization tool that operates 10 times faster than existing methods by incorporating the physical realities of the grid.
- Donti is also creating synthetic datasets to train AI models, overcoming the challenge of private or secure data in the energy sector.
- She co-founded the nonprofit Climate Change AI to build a community of experts applying AI to solve climate-related problems.
The Challenge of Variable Renewables
Integrating renewable energy sources into national power grids presents a significant challenge. Unlike traditional power plants that provide a consistent output, sources like solar and wind are intermittent. Their energy production varies based on weather conditions, creating fluctuations that can destabilize the grid if not managed properly.
Grid operators must constantly balance energy supply with demand. The unpredictable nature of renewables makes this balancing act far more complex. Machine learning is already used to predict solar power generation, but Donti's work addresses the next crucial step: using those predictions to make optimal decisions for grid management.
"Machine learning is already really widely used for things like solar power forecasting, which is a prerequisite to managing and balancing power grids," says Donti. "My focus is, how do you improve the algorithms for actually balancing power grids in the face of a range of time-varying renewables?"
A New Approach to Grid Optimization
A central part of Donti's work is the development of advanced algorithms that can optimize grid operations while respecting the system's physical laws. Traditional methods often rely on approximations, which can lead to inefficiencies or potential instabilities.
Faster and More Accurate Solutions
One of Donti's key breakthroughs is a solution that allows grid operators to optimize for cost and efficiency while adhering to the actual physical constraints of the power lines and equipment. According to initial findings, this method works 10 times faster and is significantly cheaper than previous technologies. While it has not yet been deployed in a live grid, it has captured the interest of industry operators for its potential to improve performance.
This approach uses deep learning models specifically designed to understand and incorporate the physics of electric power systems. This ensures that the AI's recommendations for managing energy flow are both efficient and safe for the grid's infrastructure.
Personal Motivation for a Global Problem
Priya Donti's dedication to this field is rooted in personal experience. Childhood visits to India highlighted global inequities, and a high school class on climate change revealed how environmental issues could worsen these disparities. This inspired her to combine her interests in computer science and public policy to address these interconnected challenges.
Overcoming Data Scarcity with Synthetic Information
A major obstacle in developing AI for power systems is the lack of public data. Much of the detailed operational data from power grids is kept private due to proprietary reasons or national security concerns. This makes it difficult for researchers to train and test new machine learning models.
To solve this problem, Donti and her research group are working to create high-quality synthetic data. This artificially generated data mimics the properties and complexities of real-world power grids without revealing sensitive information.
Creating Realistic Benchmarks
The goal is to produce datasets and benchmarks that are challenging enough to drive innovation in the field. "The question is," Donti says, "can we bring our datasets to a point such that they are just hard enough to drive progress?" By providing these tools to the broader research community, she aims to accelerate the development of more robust AI for energy systems.
Building a Collaborative Community
While pursuing her PhD at Carnegie Mellon University, Donti co-founded the nonprofit Climate Change AI. The organization's mission is to connect experts from different fields—including computer science, policy, and energy systems—to foster collaboration and share resources. It serves as a hub for education and community-building to help accelerate the use of AI for climate action.
Educating the Next Generation of Innovators
Donti joined MIT in September 2023, drawn to its focus on applying technology to solve major societal problems. Her role extends beyond research into educating future leaders in the field. She holds the Silverman Family Career Development Professorship in the Department of Electrical Engineering and Computer Science (EECS) and is a principal investigator at the MIT Laboratory for Information and Decision Systems (LIDS).
Next spring, she will co-teach a new class called AI for Climate Action. The course will be taught alongside Assistant Professor Sara Beery, whose work is in AI for biodiversity, and Assistant Professor Abigail Bodner, who focuses on AI for climate science.
Her work has earned significant recognition, including:
- MIT Technology Review’s "35 Innovators Under 35" in 2021
- Vox’s "Future Perfect 50" in 2023
- The U.S. Department of Energy Computational Science Graduate Fellowship
- The NSF Graduate Research Fellowship
Through her research, teaching, and community-building efforts, Donti is contributing to the critical technological advancements needed to enable a global transition to a sustainable energy future.