How Priya Donti’s AI Helps Power Grids Absorb More Wind and Solar — and Cut Costs

Priya Donti builds AI to help grids integrate more wind and solar

Priya Donti develops machine-learning algorithms that help electric power systems forecast and manage variable renewable energy—like wind and solar—while respecting the physical and engineering constraints of real-world grids. Her work aims to make grids more reliable, lower operating costs, and enable cleaner generation to replace polluting plants.

Early influences and education

Raised in Massachusetts with regular visits to relatives in India, Donti saw firsthand the disparities between nations. A high-school lesson on how rising global temperatures amplify inequality deepened her interest in climate justice. She matriculated at Harvey Mudd College intending to study materials science and chemistry to improve solar cells, but computer science coursework shifted her focus to algorithms and software.

Research path and Climate Change AI

Donti discovered work from U.K. researchers showing how machine learning could help electricity systems manage renewable generation. That insight brought together her technical skills and climate goals. While pursuing a doctorate at Carnegie Mellon University, she helped launch Climate Change AI, a nonprofit that connects researchers, policy experts, and practitioners and shares tools and resources for tackling environmental problems.

Practical algorithms for real power systems

Utilities face a core challenge: wind and solar output fluctuates and is less predictable than traditional thermal plants. Donti’s models explicitly incorporate the physical laws and engineering limits that govern power systems, allowing operators to forecast renewable production more accurately and manage the network more reliably.

She developed an algorithm that runs faster than many existing methods while reducing operational costs. By producing solutions that respect true physical limits—rather than relying on crude approximations—her approach helps grid operators make better, cheaper decisions. The technology has attracted interest from network operators exploring cleaner energy integration.

Synthetic data to unlock research

Real grid data is often confidential for competitive or security reasons, which can slow research. Donti creates synthetic datasets that mimic real-world grid characteristics without exposing sensitive details, enabling other researchers and developers to build and test grid-management tools securely.

Impact and teaching

Improved network management can lower household electricity bills and allow more renewable energy to displace polluting generators. Donti is on the faculty at MIT and will teach a spring course on applying AI to environmental challenges—choosing MIT in part for its collaborative culture.

"I knew that there would be an ecosystem of people who really cared, not just about success metrics like publications and citation counts, but about the impact of our work on society," she said.

Donti’s combination of rigorous engineering, practical datasets, and collaborative organizing aims to accelerate the safe, economical integration of renewables—and to make clean energy more accessible.