Abstract
This thesis presents an integrated approach to estimating and optimizing the power output of a Wave Energy Converter (WEC) using machine learning, followed by its adaptation to the wave characteristics of the North Carolina coast. Initially, a neural network model was developed to estimate a custom designed WEC’s power output based on wave height, wave period, and resonance conditions, defined by the phase alignment between the wave excitation force and the power take-off (PTO) system. A simulation-generated dataset was used to train and validate the model, achieving a maximum prediction error below 1% on randomly generated test data. A heatmap was generated to determine the optimal neural network topology, minimizing errors across various configurations. The trained model was applied to develop a Maximum Power Point Tracking (MPPT) strategy, confirming its reliability and practical application in real-time control scenarios. To enhance real-world applicability, the machine learning-based estimation framework was extended and integrated into a comprehensive wave resource assessment tailored for the North Carolina coastline. Historical wave data from July 2005 to December 2018, obtained through WaveWatch III and HYCOM models, were used to analyze 491 offshore locations. Sea states were segmented in three-hour intervals to capture statistical wave behavior, seasonal variability, and energy flux patterns. A Joint Probability Distribution (JPD) of significant wave height and peak wave period was constructed to guide the design and sizing of the Department of Energy’s viii Reference Model 3 (RM3) WEC, ensuring that it operates efficiently under the region’s most frequent wave conditions. A MATLAB/Simulink-based WEC model was developed in WEC-Sim design tool using a two-body point absorber and a slider-crank mechanism to convert heave motion into rotational mechanical power. A reactive control strategy was applied to synchronize the float’s motion with the incoming wave force, thereby maximizing power extraction. Site selection for deployment was based on mechanical power outputs across different WEC scales (full, half, and one-third), resulting in the identification of 12 optimal locations. Additionally, the impact of various array configurations was analyzed to determine the most cost-effective deployment strategies in terms of Levelized Cost of Energy (LCOE).To support MPPT development for this new WEC, a detailed dataset covering 432 unique sea state scenarios was constructed, varying significant wave heights (0.75–4.75 m), peak wave periods (3.5–12.5 s), and eight time shift values (−0.25 to 0.35 s) representing the timing offset between wave force and float velocity. Analysis showed that while in-phase alignment (0 s) yielded optimal output in certain scenarios, slight negative time shifts (−0.05 s, −0.15 s and -0.25 s) resulted in higher power capture in higher sea states, emphasizing the importance of dynamic phase adjustment. These findings demonstrate the potential of machine learning to enhance wave energy system performance and support the design of adaptive, cost-efficient WECs tailored to the wave spectrum of the U.S. East Coast