A surrogate model for capturing wave power farm dynamics using spatial-temporal attention

Area of Science:

• Ocean Engineering
• Renewable Energy Systems
• Computational Fluid Dynamics

Background:

• Wave energy converter farms face complex hydrodynamic interactions affecting performance.
• Accurate modeling of these farms in realistic sea states is computationally challenging.
• Existing models struggle with predicting interactions in novel array configurations.

Purpose of the Study:

• To develop a spatial-temporal surrogate model for predicting the motion of multiple interacting wave energy converters.
• To assess the model's ability to generalize to unseen farm layouts.
• To leverage experimental data for training and validation.

Main Methods:

• Utilized a transformer encoder architecture for the spatial-temporal surrogate model.
• Integrated incident wave time series and device coordinates into a unified representation.
• Employed self-attention mechanisms to capture temporal and spatial dependencies.
• Leveraged the SWELL dataset for training and validation.

Main Results:

• The model accurately predicted device responses in three irregular sea states.
• Predictions showed high agreement with experimental measurements.
• The model successfully predicted array responses for a farm configuration not seen during training.

Conclusions:

• An attention-based spatial-temporal surrogate model shows significant potential for wave energy farm dynamics prediction.
• The model can generalize to previously unseen array configurations.
• This approach offers a promising building block for simulating complex wave energy systems.