Wave-by-wave prediction for spread seas using a machine learning model with physical understanding
Jialun Chen, Paul H. Taylor, Ian Milne, David Gunawan, Wenhua Zhao
Abstract
Accurate surface wave predictions have the potential to greatly enhance the safety and efficiency of many offshore applications, such as active control of wave energy converters and floating wind turbines. However, real-time wave prediction becomes increasingly challenging when large directional spreading is considered. To address this challenge, the present study introduces a machine learning model that utilizes an Artificial Neural Network (ANN) for predicting moderate directional spreading waves. Linear, short-crested wave time histories are synthesized numerically to assess the capability of our machine learning model. The ANN model demonstrates better prediction capability than a recently developed theoretical scheme (Hlophe et al., 2022), extending the prediction horizon by approximately one peak period into the future. Further, a quantile loss function is introduced to quantify the uncertainty, enhancing the practical value of the developed model in decision-making processes and engineering applications, such as the active control of offshore renewable energy systems.