ESROP – Physics-Informed Neural Network Engine for Perovskite Solar Cells

Perovskite solar cells are a promising next-generation energy technology, yet their widespread commercial adoption is hindered by extreme sensitivity to manufacturing conditions, leading to unpredictable performance. This project aims to overcome this reliability challenge by developing a “smart” predictive engine that integrates artificial intelligence with fundamental physics. By leveraging a massive dataset of nearly 500,000 device measurements, interns will construct a modular system that automatically categorizes solar cells into distinct performance regimes and trains specialized “physics-Informed” models for each category. This innovative approach ensures that AI predictions remain consistent with physical laws, offering a significant improvement over traditional tools that struggle with complex hardware data. The project benefits the National University of Singapore by delivering a robust, automated tool to accelerate the optimization of solar materials, while enabling the University of Toronto to deepen its capacity in Scientific Machine Learning, fostering international knowledge transfer that supports the global transition to clean energy.

Faculty Supervisor:

Arthur Chan

Student:

Partner:

National University of Singapore

Discipline:

Engineering

Sector:

Education

University:

University of Toronto

Program: