Power Generation Prediction and Optimization of Parameters of Piezoelectric Wave Energy Converter Device

In this work, we have studied the power generated (p gen ) by a piezoelectric wave energy converter(PWEC) device attached to a solid impermeable wall under the influence of regular incident waves. The Boundary Element Method(BEM) is employed to tackle the hydrodynamic problem associated with the power generation by the PWEC device. The submergence depth(d/h) and length(l/h) are the input attributes of the model, and the response variable is the power generation (P gen ) by the PWEC device. To predict, infer and to optimize the power generation, a supervised machine learning(ML) model, namely the XGBoost model, is developed. The input database is generated using the Latin Hypercube sampling(LHS) technique, and the corresponding target variable is obtained from the solution of the hydrodynamic problem solved using the BEM. Following the power generation prediction, the optimization of the input parameters is carried out by identifying the specific regions in the input space using the interpretable ML approach, namely the Accumulated local effect(ALE) plots developed based on the XGBoost model. Further, We create a new dataset with parameter values specific to these regions, which is then fed into the developed XGBoost model to find the predicted power output, thus facilitating the optimization of the input attributes of the PWEC device. The study reveals that the maximum p gen by the PWEC device occurs when 1.3≤ l/h ≤1.8 and d/h ≤−0.2.