Optimization of parameters of the OWC wave energy converter device using MLP and XGBoost models

In the present work, we have studied the performance of a breakwater-integrated quarter-circle-shaped front wall OWC device under the influence of irregular incident waves. Firstly, the boundary value problem associated with the hydrodynamics of OWC device is handled for solution using the dual boundary element method (BEM). To examine the complex relationships between all input features and the target variable in a time-efficient manner, supervised machine learning models are developed. Here, two different models: (i) multilayer perceptron (MLP) model based on an artificial neural network, and (ii) a tree ensemble model, namely the XGBoost model are developed. The submergence depth of the front wall of the OWC device, chamber length, rotational speed, and diameter of the turbine blade are considered as input attributes, and the average annual power generated by the OWC device is considered as the output attribute. The MLP model is employed to optimize these input parameters, leveraging the insights provided by the XGBoost model to maximize the annual average power generation. From the dual BEM based numerical results, and using the Latin hypercube sampling technique, 3750 samples were generated to train, validate, and test the machine learning models. Using the XGBoost model with the support of accumulated local effect plots, we find four specific regions of the input space in which the annual average power extraction will be maximum. Hereafter, an extended input database is generated with twenty equally spaced levels for each parameter and the dataset is passed through the developed MLP model to find the optimized values of the parameters of the OWC device which maximizes the power generation. It is obtained that the maximum power generation is attained for y(0)/h= -0.65, r/h = 3, 2.8 <= D <= 3 and {70 <= N <= 80 boolean OR 105 <= N <= 116}.