Machine Learning Approaches for Predicting Power Generation in Wave Energy Converters

The potential of wave energy as a renewable energy source is significant, and therefore the development of accurate methods for predicting power output from Wave Energy Converters (WECs) is a necessity. This study employs a dataset from the UCI Machine Learning Repository. Five regression techniques were employed to this dataset for analysis: Support Vector Regression (SVR), Random Forest Regression (RFR), LASSO Regression, Ridge Regression, and XGBoost Regression. Model evaluation was conducted based on two key metrics: mean absolute percentage error (MAPE) and the coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>). Cross-validation was conducted to ensure consistency, and the results demonstrated that Random Forest Regression and XGBoost Regression outperformed other models due to their capacity to handle non-linear, multivariate datasets. The Random Forest Regression model achieved a mean absolute percentage error (MAPE) of 10.93%, a coefficient of determination (R2) of 0.98, and a mean absolute error (MAE) of 0.0408. The relatively inferior performance of linear models underscored the complexity of the dataset. This study demonstrates the feasibility of accurate wave energy converter (WEC) power output predictions, thereby providing valuable insights for optimising WEC systems. Future work will concentrate on optimising model performance through the application of advanced hyperparameter tuning techniques and the investigation of additional evaluation metrics.