Wave excitation force forecasting using neural networks

Many wave energy conversion applications require future knowledge or forecasting of the excitation force values. Most converter (WEC) control strategies need to forecast time-series for harvesting maximization. The main aim this study is experiences by a two-body heaving point absorber WEC (as case study) using three neural network methods. calculated based on hydrodynamic characteristics considered device in frequency and time-domain simulations. nonlinear autoregressive (NAR) network, group method data handling (GMDH) Long Short-Term Memory (LSTM) are fitted elevation values force. performance examined methods evaluated various irregular incident waves that created different spectrums. Moreover, sensitivity analyses sampling period algorithms input parameters performed investigate accuracy generalizability discussed at conditions. Each set divided into training test sets. results show all satisfactory sets short-term ahead forecasting, but NAR provides relatively better agreement with target compared other