Date of Award

5-2025

Document Type

Thesis

Degree Name

Master of Science (MS)

College/School

College of Science and Mathematics

Department/Program

School of Computing

Thesis Sponsor/Dissertation Chair/Project Chair

Aparna Varde

Committee Member

Pankaj Lal

Committee Member

Stefan Robila

Abstract

This study investigates using satellite-based wind data and machine learning models to support offshore wind energy planning and prediction. Sentinel-1 Level 2 data was utilized to analyze wind patterns. K-means clustering reveals that 64.6% of the wind originates from the southwest, while higher speeds are more frequently recorded from the northwest during winter. Principal Component Analysis (PCA) was employed to reduce dimensionality and interpret key patterns in wind speed and power output at the Jersey-Atlantic onshore wind farm. Artificial Neural Networks (ANN), a temporal ANN, and a Convolutional Neural Network (CNN) were developed to forecast turbine power output. The CNN, which captured spatial dependencies across turbines, achieved the best overall performance, with R2 scores exceeding 0.93. Results indicate that model performance is influenced by turbine location and wind direction, reflecting the role of wake effects. This research illustrates how spatial data and machine learning can enhance wind energy forecasting and inform more effective wind farm design.

File Format

PDF

Available for download on Saturday, May 22, 2027

Share

COinS