Optimization of Parameters of the OWC Wave Energy Converter Device Using a Machine Learning Approach

In this study, the performance of an L-shaped front wall OWC (Oscillating Water Column) device is investigated in the presence of irregular incident waves. To efficiently explore the complex interactions between input features and the target variable, two supervised machine learning models: (i) a decision tree model and (ii) an XGBoost model are considered. The input parameters selected for this analysis are: OWC Front wall Draft, OWC Duct Length, Chamber Length, ripple Number, and ripple amplitude of the sinusoidal seabed. The output attribute is the maximum power extracted by the OWC device. The XG Boost model is used to identify critical input regions where power extraction peaks, while the decision tree model further aids in understanding the parameter influences on power extraction. A dataset of 3125 samples was generated for training, validation, and testing. By employing accumulated local effect plots with the XG Boost model, five specific regions in the input space were identified that maximize annual average power extraction. The results indicate that optimal power extraction is achieved for a specific value of each parameter, thereby validating the influence of each on OWC device performance.