Integration of Machine Learning for Enhanced Wave Energy Converter Power Output Estimation
O. Mert Gursel, Paul Yanik, Hayrettin Bora Karayaka · 2024
This study aims to estimate the power output of a wave energy converter using machine learning algorithms. It builds upon previous research that developed a model to estimate power outputs from wave height, wave period, and resonance condition. The resonance condition is defined by the phase alignment between wave excitation force and wave energy converter power take off device. A dataset was created with the simulation model and used to train and test a neural network for power output estimation. The neural network achieved a maximum error rate below 1% with randomly generated test data not used in training. A heatmap was generated to identify the optimal network topology for the lowest error rates based on different configurations. The approach successfully integrates machine learning for accurate estimation of wave power outputs, offering improvements over traditional methods and enhancing the efficiency of wave energy systems. These results are also applied in a wave energy converter's maximum power point tracking as validation of the approach.