Evaluation study of predicting the dynamic mooring tension of offshore floating photovoltaic array using machine learning

Estimating mooring tension on offshore floating photovoltaic (OFPV) platforms is critical for ensuring the safety of the platform and has significant implications for its operation and maintenance. This study develops machine learning models for predicting mooring tensions in OFPV systems based on three neural network architectures (backpropagation neural networks (BP), gated recurrent units (GRU), and long short-term memory networks (LSTM)). The training data were computationally generated using OrcaFlex software, where the motion data of the OFPV platforms and corresponding mooring tensions served as training datasets for the three machine learning models. Through comparative analysis of prediction accuracy under various environmental parameters, the LSTM model demonstrated optimal performance in both computational efficiency and training economy. This comparative study provides valuable references for mooring tension prediction in OFPV array.