Real-time prediction of fish cage behaviors under varying currents using deep neural network

Abstract

This paper presents a Deep Neural Network (DNN) model for rapid and low-cost prediction of fish cage behavior under varying currents. We employ a numerical model of the fish cage created in Orcaflex and a set of current profiles from the water surface to the bottom of the cage (0-30 m). A DNN model is trained on a subset of simulated results and evaluated on a separate dataset. Our findings demonstrate that the DNN model can provide real-time, model-free predictions of fish cage behavior comparable to those of the simulator, with improved computational efficiency and robustness. The method is demonstrated to be suitable for digital twin applications, offering near-instant updates on cage behavior and valuable insights for ensuring the safety and stability of fish cage structures in challenging ocean environments.