Semester of Graduation

Fall 2022

Degree

Master of Civil Engineering (MCE)

Department

Department of Civil and Environmental Engineering

Document Type

Thesis

Abstract

The adoption of large-scale solar photovoltaic (PV) energy systems is needed now more than ever, yet solar project development and construction faces unquantified risk as it relates to wind damage. Lack of research, poor research methodology, and incorrect experimental assumptions have led to questions of whether wind loads on solar panels are properly codified, understood, and mitigated for. While the simulation of wind and its interaction with structures is possible with computational fluid dynamics (CFD), accurate simulation proves computationally demanding. Herein, state-of-the-art CFD approaches and machine learning (ML) algorithms are leveraged in two separate studies to determine and predict aerodynamic wind loads on solar panels respectively.

In the first study, a CFD model is validated based on experimental data and employed to investigate aerodynamic wind loading at various solar panel tilt angles. Wind loads determined from the CFD simulations are compared with those determined from recent editions of the ASCE 7 standard. From the results of the first study, code revisions and the optimal stow position of solar panels during wind events are recommended.

In the second study, several RANS based CFD solutions are used to train and subsequently employ a ML model to predict velocity and pressure distributions on a solar panel. This study serves as an investigation into the integration of ML and CFD for the purpose of speeding up CFD solutions without sacrificing accuracy. Results from the second study show the effects of ML hyperparameters on solution accuracy and demonstrate the significant time-savings achievable with ML compared to CFD simulations without the integration of ML.

Date

11-2-2022

Committee Chair

Aly, Aly M.

DOI

10.31390/gradschool_theses.5678

Available for download on Wednesday, October 31, 2029

Share

COinS