Towards a Physics Informed Neural Network Approach for Modeling and Prediction of Renewable Energy Production: Use Case of Solar Cells

Abstract

This work presents an innovative approach for modeling and prediction of renewable energy production of solar cells by integrating principles of materials physics with advanced techniques of machine learning, notably Physics-Informed Neural Networks. This novel approach offers the possibility of enhancing solar cell performance by gaining deeper insights into fundamental physical interactions such as electrostatic potential distribution. By modeling the solar cell using partial differential equations and predicting the energy production using the PhysicsInformed Neural Network, this work aims to unlock new avenues for solar cell design, thereby contributing to the advancement of solar energy and environmental sustainability. The present work, studying the phenomenon at thermodynamic equilibrium, has provided a well-performing model that has been able to accurately capture the distribution of the electrostatic potential. In future work, we intend to model the phenomenon under thermodynamic non-equilibrium conditions.