Graduation Term

Fall 2024

Degree Name

Master of Science (MS)

Department

Department of Mathematics

Committee Chair

Mehdi Karimi

Committee Member

Gaywalee Yamskulna

Committee Member

Papa Sissokho

Abstract

Rapid advancements in data-driven deep learning models have led to their increased usage in recent years. Highly-effective models for conducting complex decision-making tasks have been developed. Given their success, data-driven deep learning models are being developed and improved in a variety of areas, one of which is renewable energy. In this thesis, we discuss the development of data-driven, deep learning models for optimizing wind farm layouts for maximum power output. We include a background for wind energy, the physics of fluid flow and power production, and two types of neural networks: graph neural networks and physics-informed neural networks. Additionally, we discuss the construction and training of neural networks. Ultimately, we provide a physics-informed graph neural network capable of performing wind farm layout optimization for maximum power production of a wind farm under varying conditions.

Access Type

Thesis-Open Access

Share

COinS